All Premier League data continues to update in a single page here.

Hello!

I hope you are all doing well. A rare email inbox appearance from me to quickly give some updates and answer a couple of common questions thrown my way recently.

Future Projects: I’ve been putting a lot of time into some exciting personal work lately. If you are interested in making better decisions, then it will be useful for you. It’s not ready to share yet, but I’ll share more when I can. When I do, your support would genuinely mean a lot.

World Cup Fantasy: Some of you may remember the dedicated pages I built for the World Cup 2022 and Euros 2024. There is an archived page (with randomised data) to show what they looked like.

100,000s of people used those pages, and a lot of people have asked if I’ll do this again for the World Cup this summer. I’m open to it, but only if there’s enough genuine interest. It would be entirely free with no restrictions (like everything else here) and cover the basic stuff (e.g. team goals, clean sheets, win %, and player production baselines).

Please vote in the poll below if you’d actually use it frequently (please try to avoid your bias for ‘free!’ and answer as honestly as possible).

FPL ID: Some keep asking what my FPL ID and rank is this season. My ID is 9325733, and my rank is currently 1453. Unfortunately, I am still getting lucky!

Vibe-coding: A few people have sent me links to other sites that are using the new powers of vibe-coding to essentially directly replicate what is on here. I appreciate you flagging it. Honestly, I don’t mind if people reuse or build on my stuff. Everything I put out is free, and I have always encouraged others to take it in their own direction. My general advice would be to use these new tools to be revolutionary. There is so much interesting novel stuff happening in football analytics now. Here are some examples: 1 and 2 (and fantasy too!).

wishing you all good health. 💜

elevenify

Use a browser for best experience.

This site is a free personal project shared publicly for the benefit of others. Apply your own judgments and use if useful.

Everything here is free and always will be. If it helps you, share it with others.

Or alternatively, consider making a donation to the most effective charities, under the guidance of Give Well. Let me know if you do!

This page features Premier League projections from my model. All values represent average expected outcomes to set realistic expectations of likely performance. Updated every gameweek within hours of the final fixture.

Contents:

• Team Ratings

• Simulated PL Table

• Matchweek Predictions

• Team Predictions:

• Player Predictions

• Fast Fantasy Model

Subscribe now

Team Ratings

Simulated PL Table

Matchweek Predictions

Team Predictions

Note: Fixtures are mapped to the corresponding gameweeks as the fixture list for the season was announced. These numbers do not account for any ‘double/blank gameweeks’. You can see those fixtures above. Alternatively, you can see the full-season sheet here.

Player Predictions

Subscribe now

Fast Fantasy Model

March 2026 update: Fast Fantasy will return for the 2026/27 season. As we enter the final stretch of the current season, blanks, doubles, and chip strategies become more important, and Fast Fantasy’s more generic approach becomes less appropriate. At this stage, everyone’s squad is different, and player importance is highly team-specific. Use the expanded player baselines above to guide your decisions. Good luck!

Share

This site is a free personal project. Everything here is free and always will be. If it helps you, share it with others.

Or alternatively, consider making a donation to the most effective charities, under the guidance of Give Well. Let me know if you do!

5 January 2026 Update: Rúben Amorim has been sacked by Manchester United. This article was written in early September. I have updated the model screenshots below with the ratings as of today. No other changes have been made.

The phrase “give the manager time” is firmly entrenched in football psyche. This is partly justified because deciding whether to sack a manager is difficult: the kind of poor performance that leads to dismissal considerations usually results from a mix of bad luck and bad coaching.

The saying also touches on the widely-held belief that the feedback loop on a managerial decision won’t be clear for a long time, so we should abstain and give managers time. I don’t think that belief helps us understand things. You should never believe a feedback loop is so long that you can learn nothing from it.

You don’t hire a manager and then receive no new information for years. We’ve had a steady stream of information throughout Amorim’s tenure. We can grade and evaluate Amorim’s appointment now (even if it is true that more will become clear in the future). That is what matters and what decisions should be based on… and every moment you choose to keep the status quo, e.g. retain a manager, you are making an active decision, even if it feels like passively doing nothing.

United Under Amorim

As a reminder, my model uses raw performance data to generate estimates of team strength. These strengths are updated after each match based on how wrong its predictions were, taking into account match performance and the strength of the opponent. In my view, this is the best way to separate bad play from bad luck. There is no other metric or trend that better calibrates beliefs based on the strength of the evidence.1

Two things can be true about United & Amorim:

1. United are an actual and deserved mid-table side. Fixing that in Amorim's short tenure is ~impossible.

2. Equally, there is little evidence that United have improved under Amorim.

Some may argue that playstyle is an alternative explanation for the lack of improved results, that the system will suddenly “click”, and results will turn around rapidly. I think this is unlikely2 and does not change the reality that Manchester United are no more likely to win a Premier League match today than when Amorim was appointed.

Decision Landscape

So things are bad, most would agree with that. How do INEOS decide whether to sack Amorim or give him more time?

It’s helpful to start with a ‘base rate’ for our expectations: what data points do we have that help us understand what usually happens in a situation like this?

We shouldn’t limit ourselves to just Amorim’s data when forming our prior expectations. When the question is whether Amorim has shown enough evidence during his tenure to be considered a Premier League-level manager, we also have data from 20+ other Premier League managers in the same period. My model has made nearly 1,000 predictions and received outcomes for all of those data points. That is a lot of information we can use to make decisions and help us address the long feedback loop problem.

These data points suggest that when a team hasn’t improved (at all) within the time Amorim has been given, it is unlikely to improve later.3 I think there are some fair ‘reference cases’: Nuno quickly improved Nottingham Forest, Iraola steadily improved Bournemouth, Glasner revamped Crystal Palace, and Thomas Frank consistently improved Brentford. 4 Slot even won the league in his first season with essentially the same squad. 5

Subscribe now

Recommendations

Primary Recommendation:

I think INEOS should dismiss Amorim now. I feel comfortable enough to predict Manchester United are not going to substantially improve under Amorim any time soon,6 nor do I see the long-term vision.

Good decision process is walking away when you ought to, not when you want to.

INEOS have invested heavily in this project. There are many signals that they should stop, but they may view Amorim as “their guy.” It was their first major decision, and now that they’ve committed to him, it may be hard to sack him. Classic sunk-cost fallacy.

The project is well underway, and there is now ample evidence that it isn’t working. Rather than focusing on the resources and reputation already invested in Amorim and the squad, it’s better to view each new investment in the project as further waste. Wasted investments are not just retrospective; they also impact future planning. Manchester United are already dealing with this issue, carrying a squad built by multiple managers with different tactical ideas.

Which, in itself, is an added benefit of sacking Amorim quickly. Manchester United actually have a fairly decent squad, and one thing most of the players have in common is (ironically?) that they don’t really know how to play well in Amorim’s system. Players are ultimately more important than the manager when it comes to on-field performance, so there’s a strong case to be made that the quickest and most reliable way for United to become a better team is by replacing Amorim with a manager who can better utilise the current squad.7 8

Secondary Recommendation:

My secondary recommendation would be to give Amorim more time with a clear deadline and measurable expectation for concluding the project is a failure. For example: until the end of the season and Manchester United’s model projected goal difference should improve from 0.01 to 0.15.9

I stress that the club should only continue if they have their own compelling evidence that the expected value of Amorim in the future is there. If you keep working with something that won’t make you a better football club, every minute spent persisting with an underperforming strategy represents an opportunity cost that could be redirected to a higher-probability success path. Focus on the future expected value of the decision.

Good decision-making process means separating what we knew at the time we made the decision from what has been revealed afterwards. I do not know enough about the original decision process to evaluate whether hiring Amorim in the first place was a good decision.10 However, the ‘feedback’ from that original hiring decision has been plain for all to see, and we are now in a good position to know enough about the current decision regarding Amorim’s future.

In a world of messy outcomes tangled up in luck and skill, I don’t really believe we are in a timeline where Amorim’s outcomes are determined by bad luck.

Sometimes, it is helpful to rephrase the question:

If INEOS were making the decision today, with the information now available, would they start the Amorim project?

Based on the evidence, probably not.

All model data is already updated for the next gameweek in a single page here.

I’ve been sharing my hobby of building team ratings models for several years now. Almost every week, I’m asked whether my model “accounts for XYZ”.

It is a good and fair question!

But what does it actually mean?

Considering and Accounting

One of the biggest ‘myths’ implicit in a lot of these interactions is that data-driven models are entirely objective. That there is a ‘correct’ way to do things.

Models will always, to a certain extent, reflect the beliefs of their creators. This is true mechanically. Models are shaped by countless approximations, assumptions, and decisions made during creation. These decisions are central to the core question of ‘what wins football matches?’ and venture into smaller questions such as:

• How to account for red cards?

• What about predicting/rewarding penalties?

• How much does ‘finishing skill’ matter?

• Do own goals ‘count’?

• What about manager influence?

• Do we count incorrect decisions (e.g. offside goal given)?

• How to account for new transfers?

These are just a handful of many examples. People often ask me directly for the "answer" to questions like these. To me, at least, these are very difficult questions. There is no clear answer, and dealing with them is (currently) more of an art than a science. Over time, this art can evolve into a science as we gain insights into which assumptions are flawed and refine our approach.

The question “Does the model account for this?” is often a loaded question that jumps towards an answer to the questions above. The question is almost always implying an expectation that ‘the concept’ should be accounted for in a specific positive or negative way. In essence, the question already has the approximations, assumptions, conclusions and beliefs of the questioner baked in.

I think it’s therefore important to take a step back from these implicit assumptions and first look at the concepts in a neutral way. For this, I like to use the term consider.1

So the flow looks like this:

Concept → Consideration → Accounting

There is a concept that might affect the model (e.g., player absences, penalties, etc.). I then consider how this might affect things and determine whether it warrants attention, from which I move on to the actual doing of accounting for it in the model.

However, this does not mean that every concept automatically leads to a specific positive or negative adjustment. The jump from consideration to accounting is where the ‘values’ part steps in. Whether a concept ends up being accounted for in a particular way depends on the individual’s philosophy and personal beliefs about how football works. Simply put, the model considers everything, but it doesn’t necessarily account for everything (either at all or in the way you might expect).

Subscribe now

Case Study I: Rodri’s Absence

Let’s take a concrete example: Rodri missing the season for Manchester City.

• People asked: “Does the model account for this?”

• What they really meant was: “Have you made Man City worse because he is injured?”

On the surface, it seems a straightforward and fair question: if a key player is missing, does the model reflect a drop in team strength?

However, the question itself is somewhat flawed and circular. By assuming Manchester City will be worse without Rodri, the question already has a conclusion baked in. People asking it are not really prepared to reckon with the full consequences of adjusting the team rating either way. The question I would ask back is ‘are you willing to accept a world where Man City are better without Rodri?’.

If the answer to that is no, which it often is, then any conversation about updating beliefs is dead on arrival. Logical consistency requires that we remain open to the possibility that outcomes could go either way. Teams adapt, change tactics or formations, and other players may step up, sometimes maintaining or even improving overall performance. If, in reality, Manchester City performs better without him, the logical conclusion would be that they might actually be stronger without Rodri, but that contradicts the questioner’s initial assumption.

Case Study II: Red Cards and Penalties

Another helpful example. I often receive comments on weekly team ratings updates that imply an update is incorrect because a red card or penalty was issued in a game. Penalties and red cards are difficult to predict, making them particularly challenging to incorporate into a model. It can be difficult to determine exactly how they should be considered and accounted for. Similar to the Rodri example, these events can create challenges with logical consistency (for example, what do you do if a team gets a red card and then performs better?).

A common approach is to simply dismiss red cards entirely, removing all game time after the event from the data, or to focus only on non-penalty metrics, essentially pretending that penalties never exist. Both of these approaches are the “considered but not accounted for” variety. Given the high degree of randomness in these events, you can see why this is appealing!

But ignoring actual events risks leaving the model unprepared for updating beliefs when there might be some signal, such as a team receiving red cards or penalties in consecutive games. At some point, what was considered “noise” may carry meaningful information. The key is to design a model that can sensibly consider and account for these rare events to some degree, while maintaining logical consistency and the ability to update predictions based on actual outcomes.

We want to avoid creating paradoxes.

What do you do if a team is performing better without their key player? What do you do if a team performs better after getting a red card? What do you do if a team loses a match but missed 3 penalties in the match?

If you don’t have fleshed out consideration to accounting flow response to these questions, then you’re probably creating some sort of paradox.

Subscribe now

Conclusion

So if these are ‘bad’ questions, what do good ones look like? As a general rule, I think we should start by highlighting the consideration whilst sharing more context about our personal beliefs on accounting for that. For example, today I was asked two similar questions:

• “Does this factor in Bruno’s positioning from this season or does it take data from last season.”

• “does this take into account bruno's new role as a cdm?”

For the reasons outlined in this post, I don’t really know how to engage with these questions. Yes, I am considering Bruno’s position, but that still doesn’t answer the accounting part. Perhaps a better question would be something like “What are your considerations around Bruno in your model? Those Bruno projections look higher than my expectations, as I think he will play CDM and get less production from that position”.

After doing this hobby for multiple years, I think I have almost definitely considered every concept at this point.

So if the question is simply, “Have you thought about this?”, then the answer is yes.

If the question is, “Am I accounting for this concept in the exact positive/negative way you think I should?”, then the answer is: I don’t know!

How concepts are ultimately accounted for in predicted outcomes depends on personal beliefs and philosophy. I try to be as transparent as I can and share all my outputs openly. That said, I can’t spend all my time explaining every nuance of my philosophy and approach behind this little hobby. If you’re desperate to find out exactly how something should be accounted for, the best way might be to experiment and build your own model :).

Subscribe now

All model data is already updated for the next gameweek in a single page here.

The start of the season is always tricky. We want to get excited and react, but deep down, we know the sample is far too small to support strong conclusions.

I jotted down some thoughts on why early-season analysis is so hard, and what helps me make sense of what’s happening in the Premier League.

I think we can broadly carve out two buckets of difficulty: the mathematical and psychological.

The Mathematical: A few games cannot reliably capture a team’s underlying strength. It is statistically negligible and tells us almost nothing about the season ahead (unless it’s combined with prior knowledge - spoilers for how we can do better!). Football outcomes are highly variable, influenced by chance, situational factors, and opponent-specific matchups. Extreme performances in the first few weeks almost always regress towards preseason expectations, and where they don’t, it is generally better to wait and be reinforced by subsequent evidence.

The Psychological: Human thinking doesn’t lend itself to useful reactions when dealing with the tiny sample of early-season games:

• Much of the information we process in daily life comes in small samples, so we tend to assume those samples reflect the larger population. This reflects a broader pattern: we often overestimate how consistent and coherent the world really is (e.g. expecting one match to mirror season-long performance).1

• We are also storytelling creatures. Building a coherent narrative from limited evidence feels persuasive, even when it isn’t warranted.

• Hindsight bias is powerful!! I sometimes get the impression that people have got bored with this concept as it has gained more awareness, but awareness doesn’t make it any less powerful or dangerous.2

• I get the sense that a limited sample amplifies any ‘availability’ bias we have. With less information to draw from, those results that stick in our minds can give an inflated sense of importance. Furthermore, the limited information might cause us to only build conclusions based on what is in front of us3 and not take the time to think about what evidence is missing from our analysis (e.g. injuries, fitness, training data, style-specific elements, etc.).4

Subscribe now

Here are some things that help me attempt to do a better job:

1. Anchor to your Priors and Base Rates: place early-season results within the context of some sensible pre-match expectations. Start with the most reliable long-run information, then update cautiously as new evidence arrives. In football, grounding judgments in long-term data or team strength models provides those priors, ensuring that predictions are stable and less distorted by short-term noise or recency bias.

2. Think in Probabilities: I think it is a good habit to get used to expressing reactions in probabilities rather than absolutes. For example, when a team performs well in a match and exceeds expectations, get used to saying ‘they are probably ~10% better than I thought’ rather than ‘wow, they are good now’. I also find that making a habit of probabilistic reasoning increases the subsequent chance of sensible revisions as more evidence accumulates.

3. Find Robustness Through and Across Metrics: we can obtain more ‘robust’ reactions through two sensible interpretations of metrics: (1) look for explanations through the good metrics available to you (e.g. xG over raw shots); and (2) by making sure the evidence for reaction aligns across multiple indicators/metrics (e.g. rather than just relying on a scoreline result for a belief that a team’s defence is getting worse, you could look across xG conceded, moving to a less possession heavy style, a change in goalkeeper, etc.).

4. Stress-Test Alternative Explanations: before revising expectations, come up with at least two other plausible explanations.

   Much love 💜

Subscribe now

This post shares screenshots from my latest model data, which I use to anticipate what might happen in the new season. Adjust for your own beliefs and use it if it’s helpful to you!

I don’t spam this data out via email every week because everything below is already in live-updating tables on a single page here. You can view at any time. For the best experience, use a browser (not the Substack app).

Team Ratings and Simulated Table

Here are my final team ratings for 2025/26.

A common reaction is disbelief about teams like United and Spurs, with many expecting a quick, dramatic turnaround. My model does incorporate off-season changes (including transfers), but their effect on ratings is smaller than market data. I put far more weight on on-field evidence before adjusting priors.

You may also be interested in reading my post here, which explores some of the differences with the market data.

Gameweek 1 Match Predictions

Subscribe now

6 Gameweek Horizon for Goals/Assists/Clean Sheets

Subscribe now

Player Predictions (Goals / Assists / Clean Sheets)

Note: (assuming each player starts, so adjust for your own beliefs, e.g., if you think Isak won’t start, you’d set him to 0 goals early on).

Subscribe now

Fast Fantasy

Fast Fantasy puts fantasy on “easy mode” for those short on time.

It’s not a predicted-points model. It’s a tool for idea generation. Pick likely starters, roll often, and adjust for your beliefs and chip strategy. Beyond a certain point, extra effort yields diminishing returns. Much of the heavy lifting happens in the sorting and filtering, which this model does for you.

I have made the decision to remove defensive contributions from these calculations because many people use them for non-FPL fantasy games. In any event, Defenders are less important than other positions, so rely on personal preference when selecting them.

If you are interested in being a better Fantasy manager, I highly recommend you read this Fantasy Frameworks article. For good fantasy results, this stuff is far more important than any of the data above.

Much love ✌️💜

All Premier League data is on one single page here.

With just 3 days to go until the Premier League kicks off, here’s the best guess we have right now of how the season might unfold.

Below is a visual comparison of my model’s predicted table versus Spreadex’s market predictions, as of August 12, 2025. My model will get a final update Friday, but no major changes are expected.

It’s worth reminding ourselves why market data is important when trying to predict the Premier League (with the usual caveat that all reference to market data on this website is solely for its utility in understanding the premier league, not for betting).

Spreadex markets pool the knowledge of many professional and informed participants and operate with minimal margin or hedging. This means their predicted points totals are carefully calibrated to be as accurate as possible. Plus, markets quickly absorb new information, making their predictions dynamic and up-to-date. All this adds up to a high predictive signal. For more on this, I recommend the video from Solio Analytics here.

Subscribe now

Let’s take a quickfire look at some differences that interested me:

Man City, Man United, Tottenham: these teams are prime examples of what I call the ‘trust me, bro’ effect often seen in market-based predictions over a summer. My model accounts for off-season changes like transfers, but I believe their impact on team ratings is much smaller than what the markets currently reflect. I care much more about seeing good on-field evidence to update my prior beliefs. There is very little on-field evidence to support the idea that (1) Manchester City can instantly return to challenging Liverpool and Arsenal so closely, or (2) that Manchester United and Tottenham can jump dramatically from their deserved 15th and 17th place finishes last season to 6th and 8th respectively.

Bournemouth and Crystal Palace: I was surprised to see the markets rank these teams this low. This represents the opposite side of the above effect. We have a full season of solid on-field evidence supporting Bournemouth’s and Palace’s strength. I don’t see enough off-field factors to justify such a steep hit to their inferred quality. Equally, I appreciate I may be too high on them.

One limitation is that Spreadex doesn’t provide team season total goals, so there’s no direct insight into attack versus defense strength or other specific reasons behind team performance projections. This lack of granularity makes it difficult to fully understand why shifts in team rankings occur. For example, a lower expected finish compared to last season might mean: (1) the team is perceived as weaker; or (2) it might simply reflect a stronger overall league which translates into a lower finish.

Brentford: lol don’t even get me started

The outlooks for Man City, Man United, Tottenham, Bournemouth, Crystal Palace, Brentford are the perfect cases of making sure you appropriately adjust for your own beliefs and philosophy.

Chelsea: Spreadex seems more optimistic about Chelsea closing the gap on the top three, while my model suggests any improvement will be more incremental than a big leap. This is amusing, as I was significantly more bullish on Chelsea than Spreadex during the Club World Cup! This also highlights an interesting observation: my model sees the league as more tightly bunched than Spreadex’s projections imply.

Here are some other predicted league table resources you may be interested in:

• Simon’s Premier League model here.

• Scott Willis frequently simulates the league with his model.

• The Transfer Flow predicted table and previews.

• Opta league table projections.

At this moment, these insights and resources offer the clearest and most reliable prediction of the Premier League season ahead.

If you missed it, I posted newsletter #01 of this season looking at somee player snapshot graphics (link here).

Much love ✌️💜

Subscribe now

All Premier League data is on one single page here.

I’ve been developing a way to clearly show how players may contribute to Team Ratings in a ‘fun’ and easy-to-digest format.

I’ve developed a ratings system that works for me. It’s built around three broad categories:

⚔️ Attack Rating: elements directly important for scoring goals (think about stuff like: goals, assists, threatening passes and runs).

🛡️ Defence Rating: elements directly important for preventing goals (think about stuff like: tackles, blocks, goal-line clearances).

⚖️ Support Rating: this is necessary to recognise all the other indirect contributions a player makes to support their team’s Attack and Defence Ratings.

These categories are chosen because they align with my belief that analysis, whether tactical, data-driven, or otherwise, should always connect directly to the actual effects on the likelihood of winning.

The ratings are scaled so the total equals 100, representing a player’s relative influence on Team Rating. They are not absolute values for better/worse comparisons.

For example, if two players from different teams both have an Attack Rating of 60, it means they contribute about the same relative importance to their own team’s attack. It doesn’t mean they are equally skilled or productive; it simply reflects that their impact on their team’s attack is similar.

To illustrate further, let’s walk through some examples:

Forwards

Benjamin Šeško and Viktor Gyökeres

As expected for players resembling a no. 9 striker, their ratings are heavily weighted towards Attack, with smaller contributions in Support and minimal in Defence.

Small differences hint at stylistic variations. For example, Šeško’s slightly higher Defence Rating aligns with what you might expect from a generational aerial talent.

Subscribe now

Midfielders

Jhon Arias and João Palhinha

I think Arias and Palhinha are an interesting comparison because both skew towards Support while contributing equally to their team’s Defence Rating.

This might feel odd for Palhinha, but under this ratings system, his work is less about directly stopping goals and more about keeping the team structure functioning (albeit with a defensive slant).

This is somewhat inferred also by the way we see Arias is much more likely to contribute to scoring goals compared to Palhinha.

Subscribe now

Defenders

Daniel Muñoz and Joachim Andersen

An interesting comparison because it really pulls out the differences you might see in defenders. Andersen maxes out on Defence and Support, reflecting his status as one of the league’s best passing centre-backs.

Muñoz, meanwhile, keeps a high Defence Rating while also being heavily involved in the final third (for a defender), a testament to his remarkable physical output. His p90 data from last season, available in the FotMob app, backs this up:

Bonus: Florian Wirtz

Sharing a last bonus one because it is interesting. There is a good chance this Attack Rating is not what you expected when thinking about Florian Wirtz. A couple of reasons why his rating might be more evenly distributed than we anticipated:

• My projection model suggests his role in that Liverpool team will lean slightly more towards creation than scoring.

• Wirtz does a lot to support the game generally. He touches the ball a lot and is happy to have it under pressure. He averages 64 passes attempted and 61 passes received per 90.

• Even after league adjustments, his defensive work should translate well enough. Last season, he ranked in the 97th percentile for StatsBomb’s possession-adjusted pressures.

That’s all for now. This concept is useful for me, and I hope it’s useful for others. If there’s interest, I can collect player suggestions and make another post in the same format.

Much love ✌️💜

NOTE: the data on this page been put together in response to popular demand! This is the first version of the new expanded 32-team club world cup format and there is limited data for many teams. Therefore, the data on this page is a healthy blend of vibes and analysis, combined for fun. They should not be expected to be as accurate as the Premier League predictions. I expect this page to be maintained throughout the whole tournament, but no guarantees!

Thanks for following along! If you are interested in more stuff like this, then check out my Premier League predictions and alternatively the work done by @SolioAnalytics.

Contents:

• Team Model Accuracy

• Next season plans

• Screenshots of Match Predictions and Top 10 Player Goal Involvement

Team Model Accuracy

Update: post-season some back-of-the-envelope analysis:

• The team predictions have managed a season mean average error (MAE) of 0.92 to Goals and 0.59 to xG.

• On MAE, this is slightly better than the Spreadex market predictions as frozen at FPL deadline.

• My model was closer to actual Goals 376 times compared to Spreadex 363.

• My model was closer to actual xG 383 times compared to Spreadex 356.

• There were 21 total draws (same predictions as market).

There is some interesting analysis to be done. I have an experimental ‘add-on’ to my model that adjusts for specific games to pick up on big ‘time-sensitive contextual’ events (e.g. injuries, rotation, etc.). I would like to backtest that if I get the time. The reality is I won’t get the time, but I hope all of this encourages people that they should still see building their own models as something worth doing.

Next season plans

I don’t expect to be playing fantasy again next season. However, I would be open to keeping elevenify predictions going, as the single best page for free Premier League predictions.

I’m exploring the best path forward to do this sustainably (e.g. less time commitment). Please reach out if you know anyone who wants to help with this.

Data Screenshots

As always, adjust for your own beliefs, overlaying whatever views you have about any final day of the season intangibles etc.

Have a good summer 💜

A list below of all decision-making blog posts in case any interest you (all short 3-minute reads). In addition to these, I am also considering repurposing notes into a kind of ‘52 decision tools’ list, where I share one framework each Monday morning as a decision technique to focus on applying that week - let me know if that interests you.

• is it more useful to feel shame or guilt?

• do trade-offs always exist?

• when to trust your gut

• can your make your own luck?

• how to detect a scam

• your personal risk appetite

• when to quit

Set of simulations for goal involvement:

Here is a collection of the 2-minute reads released on the decision-making blog so far:

On Deciding...

do trade-offs always exist?

I read a great article by Nicholas Gruen that explores how economics is often framed as the study of trade-offs (e.g. cost vs. quality). The idea that trade-offs are axiomatic is so central to economic thinking that it's rarely questioned. A consequence of assuming trade-offs are always necessary is that better solutions may be hidden…

Read more

a year ago · 4 likes · On Deciding...

First, thank you to everyone who checked out the new blog. No football stuff there, but the posts on personal risk appetite and making your own luck may interest anyone who plays fantasy. Upcoming general posts on trade-offs, scams, and feedback.

Second, unfortunately, the elevenify site will continue to have technical issues for the rest of the season. For your convenience, I have pasted below the following screenshots from offline working tables:

• GW35 Match Predictions

• A set of sims of the top 10 players for projected goal involvement for GW35

• An updated Fast Fantasy table for GW35-38

Apologies for the site issues - maybe Substack does not like my choice to make everything free! I am thinking about next season, and I invite people to reach out if they want to use my team ratings (or website!) for any cool projects. Many people have been using them for public projects/content that help everyone, and I am happy for that to continue. Enjoy the rest of the season!

On Deciding...

your personal risk appetite.

The purpose of this post is to clarify my thinking on determining risk appetite for any given decision…

Read more

a year ago · 8 likes · 1 comment · On Deciding...

elevenify site: You may have noticed the site’s currently a bit broken. Due to technical issues with Substack, the layout has reset, and player tables aren't updating properly. None of this is intentional. I probably can’t resolve this season. The model is still updating normally, and team-level data (goals, clean sheets, Win %) is available in the 6GW horizon table (GW33–38) here. Player data likely won’t return live this season.

New blog: I have created a new blog called “On Deciding…”. Most posts will be short, 200-500-word “mini-essays” written (for myself) to consolidate/clarify my learnings on decision-making topics. If that sounds interesting or you enjoyed Fantasy Frameworks, then you might like it. The first post, ‘when to quit’, is a good example of what to expect. No football stuff there, but I feel refining my decision quality improved my fantasy performance more than anything else.

GW34 Fantasy Data: I am on FH34 this week and given the site is not working, here is a list of the top options to choose from (xPts):

Goalkeepers:

• Pope (4.5), Alisson (4.3), Leno (4.3), Sanchez (4.3)

Defenders:

• Ait-Nouri (5.4), Trippier (5.0), Livramento (4.8), Burn (4.6), Van Dijk (4.4), Robinson (4.3), Cucurella (4.1)

Bonus: if you think Sessegnon starts RW and gets ~65 mins, then you’re looking at ~5.1. A Trent start and ~65 mins gets you ~4.7.

Midfielders:

• Salah (8.8), Palmer (6.6), Diaz (5.1), Kluivert (5.1), Semenyo (5.0), Mbeumo (4.8), Murphy (4.8), Mitoma (4.7)

Bonus: if Barnes / Gordon start, they are automatically the third-best midfielder here.

Forwards:

• Isak (7.4), Raul (5.8), Cunha (5.6), Welbeck (5.5), Evanilson (5.0)

If the deadline were right now, my XI would probably be: Alisson, Ait-Nouri, Trippier, Livramento, Salah (C), Palmer, Kluivert, Diaz, Isak, Raul, Cunha… GK choice, Sessegnon, and Diaz most likely possible changes.

Hope you all have a good weekend 💜

Extra love in advance to those of you who check out On Deciding… 💜

I’ve been thinking a lot about how our decisions usually start from a default.

The way we interface with the physical world generally has a consistent, long-term default, but our interface with the digital world changes constantly.

Sticking with the default settings on the tools we use feels comforting because there’s an assumption that the creators probably know best.

But there’s no reason to assume that the data we consume or the tools we rely on are set up to encourage the decisions we want.

I’m increasingly feeling that if I’m not designing my own decisions, then someone else is designing them for me.

Even if truly ‘neutral design’ is impossible, things could definitely be better.

Our subscriptions auto-renew, meeting invites assume 30-minute slots, boxes come pre-ticked, autoplay is enabled, and our search results are sorted by invisible priorities.

I’m trying to develop the habit of pausing with every new tool or dataset to ask: “Is this default actually right for me?”

Having the wrong default settings for your purposes can have a butterfly effect on the entire output and send you down the wrong decision path.

It feels like a good idea to try and reclaim some agency by investing more time up front in the settings of the tools we use.

I expect this will also have a causal effect on producing better outcomes over the long term.

elevenify.com

Subscribe now

If you want the football data, it is all in one place here - for free.

Fantasy games are about managing resources under constraints.

Finding the ratio between that resource allocation and expected return is the formula for success.

If you want to do this particularly well, try to locate a high output from investment in assets that diverge from crowd consensus. Taking a risk by ‘chasing differentials’ every week is the basic version of this. A more advanced version might be obtaining those same assets through strategic means at times when others cannot (e.g. patient utilisation of transfers/wildcards).

A good way to conceptualise your risk appetite is by the amount of investment you allocate to strategies (players or even chip usage) that diverge from the market. This is a useful definition in a fantasy context because it ties risk-taking to your distribution of outcomes. You understand allocating resources outside of the most-owned assets increases your chances of both benefit and harm.

The above is good fantasy play and is generally known to most people. We just aren’t very good at executing it. Indeed, most players lean too hard into the above, to the point that their play worsens because of it. Risk-seeking to their detriment. Harm-seeking.

So if all the above is what we know about good play - what is unknown? I feel like there is a lot more to learn. I am quite interested in working out what questions to ask, so I can learn what I need to know.

Expected minutes and chip plans seem to garner attention as a source of edge. I think there is a seductive mask to granularity in forecasting these. It isn’t obvious to me that the level of granularity has a strong causal link to good fantasy decisions. Or, at least not to the extent that the outcome is a net positive for fantasy decisions, considering the negatives being drunk on this granularity can bring. I don’t think we handle time horizons well in a fantasy context, nor do we adequately appreciate the switching costs that make many decisions almost irreversible.

How do we deal with that? My best guess is that the answer lies in something we might think of as robustness. I am never fully sure what that looks like, but in my head, I generally associate it with principles that look like (1) selecting assets that fare well no matter what the future holds; or (2) exploring a little less than you might think.

Optimisation models are probably going to develop better approaches to the above - but they probably won’t be helpful for us using them if we are not asking the right questions.

Share

elevenify.com is now entirely free.

Read on if you wish to find out why, otherwise enjoy the Premier League page which now has all the site data in a single page ‘hub’.

I've disabled paid features and issued pro-rated refunds for all subscribers.1

If you were a paid subscriber - thank you for your support.

This has always been a hobby. Paid subscriptions were just a way to justify extra requests I did not have time for (e.g. Euro 2024 project, PL player stuff, etc.). Paywalling is the only option to support creators on Substack, but I never liked it.

I’d love a setup where I can share everything freely and sustainably, allowing me to focus on what excites me: learning how to make better decisions.

I will take some time to evaluate how I can sustainably achieve this.

For now, all content will remain available to everyone and will continue updating immediately after each gameweek, just as it has all season.

Long-term, I’m exploring the best path forward. I may move to a platform with more flexibility for hosting new features (e.g., self-hosting on Ghost). If you know someone who could help with that, please reach out!

If you have any ideas on all of the above then please reach out. I want this to work for everyone.

much love

elevenify 💜✌️

This page is no longer in use. All data has been moved to one single page here.

Last week, I released Fantasy Frameworks, a dedicated resource of fantasy decision-making frameworks. I believe that if you practice these frameworks consistently, they will improve your fantasy management more than anything else.

This week is a perfect time to apply some. At the time of writing, there are six captaincy options with just a 5% difference in predicted points. How do you decide who to pick?

You may notice a slight shift in my focus. While my models and data will continue to passively update weekly, any active posts will focus more on decision-making. I don’t want you to feel like the ‘right’ answer is somewhere out there already produced for you in models/data, instead, I want to empower you to become better decision-makers.

I won’t run this site forever, but before I stop, I’m committed to helping you become a better and more responsible fantasy manager.

Subscribe now

This post will be most helpful if you’ve already read Fantasy Frameworks. This post takes a small selection of those frameworks and gives you an idea of how you might apply them to making your decision. So I recommend checking that out if you haven’t.

Blinding

One of the Fantasy Frameworks I discussed is Blinding, a method to reduce bias by hiding key information. Here are the anonymized predicted points for the top six captaincy options this week:

• A: 6.5

• B: 6.3

• C: 6.2

• D: 6.2

• E: 6.2

• F: 6.2

Pause for a moment and reflect. Think about how you feel about that spread of points and what your beliefs are, for example:

• Do you feel like 6.5 is sufficiently far ahead of the other options?

• Are C, D, E, and F equal? (They are all 6.2 so the answer should be yes).

Take note of your beliefs before moving on. The next section will include the same information but with the player names. It is very likely some thoughts might sway, for example:

• confirmation bias might make the 6.5 feel further ahead than your current belief

• a preference or dislike for certain players might make C - F start to feel less equal even though they are the same

When the options are this close, I think these biases are more likely to affect your decisions. That’s not inherently wrong, but you should recognize them and ensure you’re adjusting through a coherent process, rather than pure intuition.

Baselines

Here are the top 6 players for predicted points from my model for this week:

• Trossard: 6.5

• Odegaard: 6.3

• Semenyo: 6.2

• Isak: 6.2

• Salah: 6.2

• Palmer: 6.2

In terms of making our decision, we can take the above predictions as a kind of ‘base rate’ - our best guess starting point as to what we can expect from these players. The idea is that we will then forecast away from these base rates if we have a good reason to.

This principle would also apply if you were interested in other predictions such as predicted goals.

Subscribe now

Switching Cost

Fantasy Frameworks emphasizes that switching between players comes at a cost because your resources are limited. With the upcoming blank and a chaotic schedule, these costs are higher than usual.

Because of this, I don’t recommend making a transfer just for captaincy this week. This is also made easier by the options being so close so the gain would be minimal. From a captaincy decision-making perspective, you can immediately rule out any players you don’t already own.

Wisdom of the Crowd

Fantasy Frameworks suggests using multiple data sources to make better decisions. In this case, we can compare my projections with the FPL Review Free Model and take an average:

• Player: elevenify | Review | average

• Trossard: 6.5 | 6.2 | 6.35

• Odegaard: 6.3 | 6.5 | 6.4

• Semenyo: 6.2 | 5.1 | 5.65

• Isak: 6.2 | 5.7 | 5.95

• Salah: 6.2 | 5.5 | 5.85

• Palmer: 6.2 | 5.4 | 5.8

Averaging these models reduces the gap between the best options which might seem like it has made the decision harder. However, having multiple perspectives may increase your confidence in choosing the top option, even when the differences are small.

Reviewing market forecasts for players and matches can be another useful way to gather crowd-sourced insights that may justify adjusting your base rates.

Subscribe now

Practical Example

I’ve selected a few Fantasy Frameworks to show how they can be practically applied, though many others could also apply to your captaincy decision. Here’s a quick walkthrough of my process using my GW26 team (1FT, 0.2m ITB):

• Switching Cost: As noted earlier, I won’t make a transfer solely for captaincy. Additionally, none of the "Heuristics for Taking a Hit" apply. This allows me to eliminate Odegaard, Trossard, and Semenyo from consideration because I do not own them.

• Wisdom of the Crowd: After averaging multiple projections, I’m left with these three options: Isak (5.95), Salah (5.85), Palmer (5.80).

• Personal Policies: I have a Personal Policy that I will not chase ‘differential’ captains and should select from the restricted pool of top options. This confirms my choice is between Isak, Salah, and Palmer.

At this point, the frameworks have helped me narrow down my decision to three valid choices. Since the differences in their predicted points fall within a reasonable margin, there is no clear ‘correct’ pick and any of the three is a justifiable option.

If you want to go deeper, you could apply other frameworks. For example:

• Following the Befriend Your Resources framework might lead to the thought that models like elevenify may take time to adjust. You might reason that Nottingham Forest is slightly undervalued, which could lead you to discount Isak a bit.

• Another Befriend Your Resources theme might be adjusting your evaluation of Trossard, as you know we don’t have a good sample of him in his current role in the Arsenal team. You might be concerned that models are overestimating his contribution.

Finally, it is also sensible to finish your decision based on your risk appetite. For instance, because I have a low appetite for risk, I would likely captain Salah, expecting him to be the most owned option.

Summary shortcut process:

1. Get a starting point (e.g. xPts, predicted goals, etc.).

2. Apply any frameworks that might affect your decision.

3. Adjust if you have **good** reason.

4. Adjust for your risk appetite.

5. Choose the top option.

I hope this walkthrough gave you a clear, practical view of how to apply Fantasy Frameworks to your decision-making process.

If you found this useful and would like to see more, please consider sharing this post and letting me know. I always appreciate your feedback. 💜 ✌️

Share

I've been obsessed with decision-making for as long as I can remember. I am not an ‘expert’, but I consider myself a ‘decision science enthusiast’. I spent over five years at university researching decision-making, and now my day job revolves around helping people navigate uncertainty in business and compliance. I am enthralled by trying to understand what conditions lead to the best outcomes and designing strategies to improve decision success. Six years ago, this obsession naturally led me to one of the most fascinating games of making decisions amid uncertainty: Fantasy Premier League.

This experiment has gone quite well. At the time of writing, I have the best average rank over the six years I have played. Of course, a lot of that is luck, but perhaps there’s some signal buried in there. People often assume that any signal is linked to being an ‘analytics’ manager and the quality of data I use. Indeed, I was an early adopter of FPL Review and elevenify.com is popular mainly because I share my data and models. Naturally, people seek out the best projections and tools, but you might be surprised to hear that I don’t think having the best data is the most important factor in improving at fantasy games (excellent comment to devalue my website!).

The idea that models ‘play fantasy for you’ is completely wrong. Even with perfect data, you are still the one making decisions. You can’t remove the human element from the process, and having the best tools doesn’t automatically make you a logical, rational decision-maker. Sometimes, it does the opposite. There will always be aspects of the game that humans can navigate better than machines. Data is important, but what is more important is a proper marriage between the human and the data.

Fantasy will always be a game of incomplete information and luck. But that doesn’t mean we should be nihilistic about success. In the long run, fantasy is a game of skill. The best way to think about your actions is that they don’t guarantee a specific outcome, but they shape your distribution of outcomes. So, your goal is to find strategies that help you make the best possible decision given uncertainty. Refining your decision quality will improve your fantasy performance more than anything else. I believe techniques like those in this article would give you an advantage over others even if you only had access to worse data than they did. This article comes from the general notes I’ve kept over the years on improving my process, now expanded with specific applications to Fantasy Premier League.

My fantasy decision-making process generally starts with: Do I have a rule or heuristic I can apply? If I do, I use that as my foundation and build from there. This article contains 23+ frameworks, each with an application to fantasy decision-making. It will serve as a dedicated resource that I’ll continue to refine over time. Please message me or comment if you see any fantasy-specific examples when reading through this, and I will add them to the article.

It’s important to recognize that there is no reason to believe that fantasy games or the tools we use to help us are set up to encourage good decision-making by default. The hope is that these frameworks can act as a counterweight, creating a better system and a better environment for making strong decisions. I firmly believe that these skills can be trained and developed like muscles. Your goal is to implement and practice them as often as possible, and over time, you will get better. I truly believe that implementing these ideas correctly will improve your fantasy management more than anything else.

Switching Cost

Framework: The higher the switching cost, the more reluctant you should be to explore and try new options.

Fantasy application: In fantasy, the cost of switching between players is generally high because your resources are limited: one free transfer per week, hits required beyond that, and only two wildcards per season.

Beyond these tangible costs, there are also intangibles to consider:

1. Football is inherently unpredictable, so you don’t get immediate, clear feedback on whether a switch was good or bad.

2. Players are dynamic so there’s always a risk that both the player you bring in and the one you move out shift in unexpected ways, especially early in the season.

If these intangibles lead you to make a mistake, you often end up paying the tangible cost twice: once to bring a player in and again to transfer them out. The cost might be even greater if it involves expensive players or a shift in team structure.

From my experience, a conservative approach (exploring less than you might feel tempted to) is a solid rule of thumb. You want the moves you made 2–4 weeks ago to support the decisions you’re making now. Some might find this playstyle "boring," but I encourage you to see it differently: rolling transfers (exploring less) gives you more flexibility and freedom to make better decisions when they matter most.

Resources and related frameworks: (1) Free Rolls Framework; (2) Upstream Problem Solving Framework.

Understand The Meta

Framework: Try to understand the wider context of how games are played.

Fantasy application: If you understand the fundamental philosophies of a game, you can adapt your play to capitalize on opportunities. It’s not just about updating your assumptions about players and teams, you should also be reassessing the broader landscape (meta) as it evolves. The game has a fixed set of rules, but what is important within the game rules can shift over time. This applies both to the game of football itself and to the game of fantasy football:

In football, the meta could shift in a way that increases the average goals per game across the league, making clean sheets less common. In this scenario, you might feel dissatisfied with your defensive options (especially the expensive ones) and look for alternatives. Recognizing this meta-shift and adjusting your strategy accordingly might lead you to view defenders as less valuable. Instead of simply trying to find better defensive options, you might decide to allocate fewer resources to defence altogether, opting to play three at the back more often. You can make similar analogies for shifts in football rules that make (1) penalties more likely (penalty takers become more valuable), or (2) red cards become more common (certain players e.g., defenders might become less valuable). These are examples of how understanding broader meta-trends in football can help you make better decisions in a fantasy context.

In fantasy, it’s crucial to remember that it’s a game with a specific set of rules. While everyone is aware of the rules (e.g., forwards earn 4 points per goal, midfielders earn 5, etc.), few take the time to truly understand what those rules mean in practice. The Fast Fantasy Model I released serves as a useful heuristic, showing that, on average, defenders and goalkeepers score fewer points than other positions. As a result, they tend to be less important, and in most cases, your resources (e.g., transfers, budget) are better spent elsewhere.

Resources and related frameworks: (1) Fast Fantasy Model; (2) the rules page of the game you are playing; (3) any articles or statistical resources about football more generally.

Blinding

Framework: Limit conscious or unconscious bias by concealing key information.

Fantasy application: I structure my data with an extra page that removes team and player names to eliminate biases and preconceptions. This approach is especially useful for sensitivity analysis, as that process tends to highlight players with minutes risks and surfaces options I might not have otherwise considered. I don’t think this should be a core technique (as you still need to be engaged with the reality of your potential decisions) but I think it is a good check and balance you should be using infrequently. I typically check in this way first thing after a gameweek and then once or twice throughout the week.

Resources and related frameworks: search for blinding (or masking) in clinical trials.

Upstream Problem Solving

Framework: Solving problems before they happen rather than reacting to problems.

Fantasy application: The earlier Switching Cost framework emphasized that fantasy resources are valuable and limited. If you’re constantly using them just to fix problems rather than preparing for the future, it’s usually a sign of poor play. If this sounds like a habit in your game, you need to address the root cause. Bad habits creating these kinds of problems might include:

1. Buying injury-prone players.

2. Making unnecessary “luxury” transfers.

3. Bringing in players with minutes risks.

4. Constantly chasing the “best” player each week.

Eliminating these kinds of habits will help you manage your team more efficiently and reduce the need for reactive moves.

Resources and related frameworks: (1) Upstream by Dan Heath; (2) Switching Cost Framework.

Subscribe now

Wisdom of the Crowd

Framework: Collect multiple different independent data points and average them together.

Fantasy application: try to gather multiple data sources to inform your decisions. For example, you could aggregate: (1) starting lineup predictions; (2) team predictions; and (3) player projections.

A unique method to improve decision-making is leveraging your past self. If you document your decisions over time, you can refer back to them, creating a kind of "wisdom of the crowd" effect from your own historical thinking. Since opinions constantly evolve, reviewing past decisions can provide valuable perspective.

That said, always consider the quality of your data sources. For instance, averaging elevenify goal predictions with market-based forecasts is likely a solid process. However, adding mainstream pundit predictions might dilute accuracy rather than enhance it.

You don’t need to do this for every decision, but for major ones, it can be a powerful tool.

Resources for further exploration: (1) The Wisdom of Crowds By James Surowiecki; (2) Make Predictions Framework.

Kill Criteria

Framework: Determine criteria for changing your mind (e.g. what specific information do you need to find out and by what specific time before you can finalise your decisions).

Fantasy application: This framework is an excellent way to know when to cut your losses. Here are a few examples:

1. Choosing a chip strategy often requires waiting for more information over an extended period. While there’s always the possibility that a better opportunity arises later, delaying too long can cost you points. (For more on this, see the Over-Patience Framework.) To avoid indecision, be clear on what specific information you need (such as cup fixture results that determine double and blank gameweeks) before committing to a strategy.

2. If you’re unsure about a player (e.g., minutes risk, penalty duties), define in advance what evidence will confirm or refute your belief. For example, if they get benched without any clear explanation or miss out on a penalty they were expected to take, that could be a signal to move on.

3. To avoid rushed decisions on your week-to-week transfers, determine what you need to see before making a move, such as waiting for a midweek fixture or holding off until a press conference for injury updates.

Resources and related frameworks: (1) Quit by Annie Duke; (2) Over-patience Framework.

Personal Policies

Framework: Establish a hard rule with yourself to counteract unwanted behaviour.

Fantasy application: This framework was inspired by something I heard Adam Grant discuss on a podcast about how establishing personal policies can improve workplace decision-making. I think this concept is useful in the context of this article as a kind of beginner framework. By now, you’ve probably realized that applying all these frameworks consistently is hard. It’s a constant challenge. As a first step, it may help to identify a specific struggle or flaw in your play and create a personal policy to address it. For example:

• If you tend to make knee-jerk transfers, set a policy to never move early in the week.

• If you struggle with price-change FOMO, create a rule that you must list other valid reasons before making a transfer.

• If you constantly chase differential captains, consider a policy of always captaining your most expensive player.

Similar to Effort Allocation Framework principles, these policies might sometimes prevent you from making the absolute best decision at a given moment. However, over the long run, they will help you avoid costly mistakes while you improve other areas of your game. Once your decision-making improves, you can gradually move beyond rigid rules and make more nuanced choices.

Resources and related frameworks: (1) Rethinking the Workplace with Adam Grant; (2) Effort Allocation Framework.

Subscribe now

Befriend Your Resources

Framework: Fully understand the resources that you are using.

Fantasy application: Understanding your resources is crucial for making informed decisions and avoiding mistakes caused by incorrect assumptions.

1. Metrics like expected goals (xG) and expected assists (xA) require context. You should account for factors like sample size, playing time, and finishing ability to avoid misjudging a player’s true potential.

2. Projection models are a result of their inputs and settings. It is helpful to take a step back and think about the transparent and hidden aspects of any given tool. For example, you might be able to see things like minutes and penalty share but you might be unable to see positional assumptions. Understanding how these models work helps you evaluate their reliability and make better-informed choices. A key part of this process is reading the FAQs or documentation that accompany the tools you use. These often explain how the models are built, their assumptions, limitations, and what the different parameters mean. You can’t expect good results from a tool if you don’t understand how to use it properly.

Resources and related frameworks: any FAQs or manuals accompanying your resources.

Scales of Likelihood

Framework: Learn to express your beliefs in scales of likelihood.

Fantasy application: Cromwell’s rule advises against assigning absolute probabilities of 0 or 1 unless you are absolutely certain of an outcome. I think this is a useful principle to keep in mind. Humans aren’t naturally wired to think probabilistically, so developing that skill will give you a significant edge. The goal is to train yourself to see the world as a probability distribution rather than a series of binary outcomes.

This struggle with probability is especially apparent at the margins. In fantasy, player starting likelihoods and penalty-taker probabilities are good examples. Many people treat a 10% chance of a player taking a penalty as effectively zero, which is a mistake. Every possibility, no matter how small, should factor into your decision-making (for instance, see here a meme I made about Flekken’s expected assists months before he got an assist doing the exact free-kick routine I modelled).

A helpful next step is to express percentages as likelihoods. A 90%, 99%, or 99.9% chance of winning all feel like "almost certain" outcomes, but flipping the perspective clarifies the difference:

• A 90% win probability means a 1-in-10 chance of failure.

• A 99% probability means 1-in-100.

• A 99.9% probability means 1-in-1,000.

Similarly, instead of just seeing a clean sheet probability of 30%, reframe it as: “If this game were played ten times, this team would keep a clean sheet in three of them.” This mental shift encourages more critical and rational thinking about probabilities, helping you make clearer, more informed decisions.

Resources and related frameworks: (1) Thinking in Bets by Annie Duke; (2) Bayesian Thinking Framework.

Over-patience

Framework: Waiting too long to make a decision and favouring optionality can cause you to miss opportunities that outweigh the benefits of options.

Fantasy application: I believe this framework represents the biggest weakness in my own play. I generally consider myself a patient, analytics-driven manager, and I think that type of manager often overvalues optionality, which can lead to leaking a lot of points in the process.

To counteract this, I’ve started focusing more on when to "lock in" a strategy, similar to the philosophy behind the Kill Criteria Framework. This is especially relevant for chip strategies, which I now approach as a kind of optimal stopping problem. I’m not saying they are optimal stopping problems, but I find the thought experiment useful in the context of thinking about patience.

Resources and related frameworks: Kill Criteria Framework.

Planning Fallacy

Framework: Don’t get too optimistic that your future plans will materialise.

Fantasy application: We are generally too optimistic about the likelihood our future plans will materialise. Be careful about committing too rigidly to a plan that extends too far into the future. Small shifts in the football landscape can create butterfly effects that alter the optimal decisions day by day.

Always consider what could go wrong (injuries, red cards, unexpected team changes, new key information) and ensure you have a viable backup plan. If a single event can completely derail your multi-week strategy, then your planning wasn’t robust enough. A practical way to test this is by stress-testing your plan: create a hypothetical scenario where something goes wrong every week and evaluate how you would respond. This exercise helps you build more flexible, resilient strategies that can adapt to changing circumstances.

Generally, I consider future plans to be 'very ‘speculative’ a lot earlier than you might imagine. Even just 2-3 weeks ahead is still significantly abstract. Always take a step back and remember that you are only ever actually making a move in the current gameweek - the process will start fresh next gameweek.

Resources and related frameworks: (1) Wiki; (2) PwC Article; (3) general reading on optimism bias.

Share

Drunk on Models

Framework: Not doing whatever your model tells you.

Fantasy application: This one is particularly meta because, in a way, many other frameworks serve as antidotes to being drunk on models. It’s crucial to remember that models are, first and foremost, tools for idea generation, not absolute truth.

Always take a step back and sanity-check their suggestions. Pay close attention to the assumptions and simplifications being made, as well as any outside factors or intangibles that the model might be missing.

Ultimately, take responsibility for your own decision-making. Models can guide you, but they shouldn’t own your choices.

Resources and related frameworks: (1) Befriend Your Resources Framework; (2) consider the model-specific elements of every other Fantasy Framework.

Teamwork

Framework: You make better decisions and improve at an activity when participating in a group.

Fantasy application: Find a trusted person or small group to discuss your decisions with. Having the right decision group can improve your process in ways you simply can’t achieve alone. A good group can challenge your thinking, play proper devil’s advocate, correct errors, and share valuable insights. For this to work effectively, a few key requirements:

1. Everyone in the group should be there for good-faith collaboration, genuinely engaged in the decision-making process and share the same objective. Including you! Good-faith means actually thinking through responses, not just giving surface-level advice like "just roll." This is why smaller groups tend to be more effective as there’s more accountability and depth of discussion.

2. Be mindful of groupthink and confirmation bias. A good decision group should challenge your ideas, not just reinforce what you already believe. Make sure there’s space for differing opinions and real debate.

Resources and related frameworks: your friends!

Make Predictions

Framework: Make predictions, document them, and evaluate accuracy.

Fantasy application: This might seem obvious (after all, playing fantasy is already about making predictions) but you need to be more explicit about it. Put yourself out there. Put it on the line.

• Who is the penalty taker?

• Who will start this week?

• When will that fixture be rearranged?

Write them down, track them, and evaluate your accuracy. To build a real feedback loop, you need actual data on what you're getting right and wrong. The goal is to identify your strengths and weaknesses so you can improve.

Make sure to focus on relatively predictable things (you won’t learn much from trying to guess a random red card or penalty event) but you can refine your ability to assess things like team selections and set-piece takers. Fantasy football is inherently difficult because it’s a noisy, unpredictable game. Even the best managers are often wrong. The key is to learn from being wrong and gradually refine your process.

Resources and related frameworks: (1) Wisdom of the Crowd Framework; (2) FS The Knowledge Project Ep. #68; (3) Alliance for Decision Education Decision Journal Article.

Free Rolls

Framework: Look for decisions with only upsides and no downsides.

Fantasy application: True free rolls are rare in fantasy (mostly due to the high switching cost) but you should always be looking for small, free wins with no risk. This ties into the concept of decision stacking: making low-impact, easy-to-reverse decisions that help inform your high-impact, harder-to-reverse decisions. A classic example of decision-stacking in life is renting before buying a home or dating before marriage.

In fantasy, the goal is to identify small, low/no-risk choices that make your big decisions easier and more effective. By training yourself to find these free advantages, you make your overall decision-making process smoother and more robust. Here are some fantasy-specific examples:

• If two players have similar projected points, but one has a 10% penalty share and the other has a 90% penalty share, it might be better to bet on the 10% player. If the assumption about penalty takers is wrong, there's more potential upside in your favour.

• Choosing a player who fits multiple potential chip strategies (e.g., someone who covers a blank gameweek but also has a strong bench boost week).

• When you are forced into a move (e.g. replacing an injured/suspended player), check if you can gain other small wins, like picking someone who also helps navigate an upcoming blank or double gameweek.

• Doing proper due diligence on your ‘dead spots’ (e.g. choosing the 4m defender/goalkeeper with the most likelihood to play over a long horizon).

Resources and related frameworks: (1) Freerolls paper; (2) Switching Cost Framework.

Share

Path Exclusion

Framework: Think about what alternatives are foregone (or made impossible) when making your choices.

Fantasy application: In fantasy, the focus is often on the players you want, but it’s just as important to consider the opportunities you give up by making certain choices. Every decision comes with an opportunity cost, and recognizing these trade-offs can help you make better long-term decisions. Some fantasy examples are:

• Choosing a third player from a team means you lose the flexibility to bring in another option from that team later.

• Playing a chip at one point in the season means you can’t use it elsewhere, and it may limit how effectively you use other chips (e.g. the Assistant Manager chip locks out three full weeks, or using your final Wildcard early can make it harder to set up a Bench Boost later).

• Allocating your budget to specific players or positions can make it harder for you to rearrange your team. For example, choosing a 4.5m striker dead-spot may allow you to put funds elsewhere but if you need to get a striker back into that position you are forced to use two transfers.

• Any given expensive premium player generally means you forego other premiums and this could have knock-on effects (e.g. missing out on the best captaincy options).

Resources and related frameworks: reading generally on opportunity cost is relevant here.

Hyperbolic Discounting

Framework: People prefer outcomes that come sooner.

Fantasy application: This cognitive bias gets to the core of fantasy play in evaluating transfer options. Generally, people are too focused on what a player can provide them within the next single game week and it is better to look at the longer horizon of average.

Some analytics tools address this by using decay (a method of valuing the near future more than the distant future). This is done by applying a discount to predicted points further into the future (e.g., reducing their value by 10% per week). The logic is simple: the further ahead we plan, the more uncertainty we introduce (injuries, suspensions, tactical shifts, etc.). Let us take two players and imagine they have equal decay applied:

• Player A: GW1: 5 points , GW2: 4.6 points , GW3: 4 points

• Player B: GW1: 4 points , GW2: 4.5 points , GW3: 7 points

Total points over three weeks favour Player B, but due to hyperbolic discounting, many people will still prefer Player A because the good outcomes arrive sooner.

If you're considering the longer horizon and applying appropriate decay, be mindful that your own mental biases might further exaggerate the preference for immediate returns even when it has already been dealt with by decay. Recognizing this bias is all about making sure you have a logically consistent approach to the near and future horizon.

Resources and related frameworks: (1) Wiki.

Loss Aversion

Framework: People prefer avoiding losses compared to equivalent gains.

Fantasy application: This well-known bias relates to the human tendency to feel losses more strongly than gains. In fantasy, this can manifest in several ways:

1. You may feel pressured to buy a player simply because you don’t own them, fearing that you're missing out. In reality, you often have near-equal ways to get similar points, either from players already in your squad or through a different set of moves. It’s important to remember: you can’t own everyone.

2. The temptation to buy or sell a player just because their price is about to rise or fall can lead to rash decisions. While price changes matter, they shouldn’t dictate your transfers, often, waiting for more information (injuries, team news, etc.) is the better play.

3. Being biased against selling a player because you are too invested in the value you have built up in them even when there are better options and it is time to move on.

Resources and related frameworks: (1) Wiki.

Mere Exposure Effect

Framework: Preference for things because you’re familiar with it

Fantasy application: Although fantasy resets each season, certain ideas and concepts carry over and sometimes, that’s a problem. Your decision-making can be compromised in several ways:

• You may place too much trust in players you’ve had success with in previous seasons, forming an emotional attachment. This is where the Blinding Framework can help and also make sure you’re truly updating your priors using Bayesian Thinking rather than relying on outdated perceptions.

• Just because a chip strategy worked for you before (e.g., always wildcarding before a bench boost) doesn’t mean it’s the best approach every season. Be mindful of tunnel vision, and evaluate fresh options rather than defaulting to past habits.

• If you’ve had good or bad experiences with a well-known approach (e.g., "rotating goalkeepers", bench boosting in a double), you may develop an emotional bias for or against it, rather than assessing it objectively based on the current season’s conditions.

Resources and related frameworks: (1) Blinding Framework; (2) Bayesian Thinking Framework; (3) Wiki.

Share

Bayesian Thinking

Framework: Place great weight on your existing assumptions and update with new evidence as the season unfolds without completely discarding its existing assumptions.

Fantasy application: This update process involves adjusting existing assumptions based on how well new evidence aligns with or contradicts them. Crucially, the degree of adjustment should be proportional to the strength of the evidence: strong, consistent data warrants a significant update, while weaker or conflicting information should lead to only minor changes. In my opinion, this approach is one of the best ways to calibrate beliefs appropriately. It ensures that you’re neither overreacting to small sample sizes nor ignoring meaningful trends, leading to more balanced and rational decision-making.

In fantasy, I think we often overreact to new information. If a player starts five matches in a row but is then benched once, many tend to downgrade their expected minutes too much without considering the broader contextual pattern. Applying Bayesian thinking correctly across different scenarios is crucial: for example, a player starting five games before being benched once isn't the same as a player taking five penalties but missing the next. While the patterns appear similar, each requires a unique adjustment (e.g., is penalty-taking more stable than starting matches?). Put more generally - you should try to avoid overreacting to low-variance events but willing to update more quickly with high-variance events.

Resources and related frameworks: (1) Everything is Predictable by Tom Chivers; (2) Julia Galef’s videos referencing Bayes.

Effort Allocation

Framework: Allocate your decision-making efforts to those decisions that truly matter.

Fantasy application: We’re only human, with limited resources (time, energy, intelligence, etc.). Your decision-making suffers when you're tired, overwhelmed, or out of the loop. That’s why it's crucial to focus your attention on big decisions and streamline the ones that matter less.

For most fantasy decisions there is barely any correlation between effort expended and good outcomes. In fantasy, much of the heavy lifting happens in the sorting and filtering stage. That’s the primary value of my Fast Fantasy Model, it quickly highlights the best options so you don’t waste time on unnecessary decisions. Even if choosing between the final few options feels difficult, the reality is that the hard work is already done, you’re simply picking between good choices. At that point, you should probably just make a decision and move on.

If you frequently feel overwhelmed, consider setting some rules (similar to the Personal Policies Framework) to lighten your cognitive load. For example, always pick 4.5m goalkeepers and ignore pricier options. While this might not guarantee the perfect choice in every situation, it will likely improve your overall decision-making process by reducing fatigue, leading to better results in the long run.

Resources and related frameworks: (1) Fast Fantasy Model; (2) FS Blog Decision Matrix.

Bug-fix Your Decisions

Framework: Examine your wins and go beyond just identifying mistakes.

Fantasy application: Try this thought experiment: do you analyze a good outcome as deeply as you do a bad one? Most people instinctively dissect their mistakes, but they rarely scrutinize their successes. Don’t stop analyzing just because a decision worked out, go further and ask whether there was an even better choice you could have made.

Many of the frameworks in this article focus on good processes and on distinguishing between a good and bad decision. But don’t mistake identifying the mistake as the ultimate goal. You need to go beyond recognizing the error and ask: What should I have done instead? Sometimes, this process will even reveal that what you thought was a mistake wasn’t one at all.

Your fantasy evaluation should be so objective that an outsider wouldn’t be able to tell whether you had a good or bad gameweek. The goal isn’t just to celebrate wins but to refine your decision-making, even if that means realizing a so-called "good" decision wasn’t actually optimal. This requires being comfortable with the pain of turning subjective wins into potential losses. It’s not easy, but it’s necessary for long-term improvement.

Resources and related frameworks: general reading on outcome bias.

Heuristics for Taking a Hit

Framework: A checklist of heuristics to help me decide whether or not a hit is worth it.

Fantasy application: In the context of uncertainty as to whether a hit will be ‘mathematically’ worth it, the hope here is that thinking and acting by the below principles should help produce our best estimate of an optimal hit - or at the very least translate into more worthwhile hits. It is worth noting that, although this is framed in the context of taking a hit, all can be applied equally to transfers more generally.

• Captaincy: Does the hit allow me to gain the optimum captain for a GW(s)?

• XMins/Injury: Does the hit allow me to remove a player who has poor xMins? (injury, lack of fitness, rotation risk, transfer risk, suspension, manager comments, media stories, etc.).

• Team Value: Does the hit allow me to significantly increase my team value? Your team value can generally be mapped to an increase in points. As a general rule of thumb, this can be anywhere between ~0.3 to 0.5pts per gameweek per £1m.

• Excess of Options: having a significant number more moves of significantly higher value than you have free transfers is generally a good indicator you should consider taking a hit.

• Chip Strategy / DGW/Blank: Does the hit allow me to follow an optimal chip strategy or allow me to have better coverage in a DGW/Blank?

• Future/Extra Transfers: Does the hit allow me to bring a transfer forward which I will almost definitely have to make? (If your team is in such a bad way or you are a victim of poor planning then sometimes it might be worth taking a hit to bring a transfer forward that you know you will want to make anyway - this is especially true if doing so allows you to hit a better fixture with that player). Another consideration here is that if you know you will be priced out of a future move that you will likely make then you could be considered to be buying yourself ‘2 free transfers’ with a hit.

Resources and related frameworks: Thinking About Hits

Summary Checklist:

Below is a quick summary checklist that can be used as a reference for improving your play:

• Switching Cost: Can you justify this resource use?

• Understand the Meta: Have you considered how the broader landscape affects your decision?

• Blinding: Are you ensuring all options are considered?

• Upstream Problem Solving: Are you preparing for the future or just firefighting?

• Wisdom of the Crowd: Can you crowdsource aspects of this decision?

• Kill Criteria: Have you set clear criteria for changing your mind?

• Personal Policies: Does this decision break a rule you should follow?

• Befriend Your Resources: Do you fully understand the resources supporting your decision?

• Scales of Likelihood: Are you thinking clearly about probabilities?

• Over-patience: Are you waiting too long when you should act now?

• Planning Fallacy: Are you too rigidly committed to a long-term plan?

• Drunk on Models: Are you sanity-checking your model’s suggestions?

• Teamwork: Should you discuss your process with someone else?

• Make Predictions: Are you documenting and reviewing your decisions?

• Free Rolls: Is there a risk-free win available?

• Path Exclusion: What alternatives does this decision eliminate?

• Hyperbolic Discounting: Are you balancing short-term and long-term gains?

• Loss Aversion: Are emotional losses clouding your judgment?

• Mere Exposure Effect: Do you prefer this just because it's familiar?

• Bayesian Thinking: Are you updating your beliefs appropriately?

• Effort Allocation: Are you using your decision-making energy wisely?

• Bug-fix Your Decisions: Are you analyzing all decisions, even the wins?

• Heuristics for Taking a Hit: Consider captaincy, injuries, team value, chip strategy, and future transfers.