How Vice Captain’s projections work
We build projections for four fantasy games, all by the same methodology. In the Premier League, where we also have something to compare against, Vice Captain’s projections sit at a level close to the world’s leading projection tools - and you do not simply have to trust us on that.
Here are three checks that show the difference between a real projection model and a guess that looks convincing.
The comparison was run with player minutes held equal across both models, to isolate the quality of the projection itself. 21 of the top 25 players appeared on both sides.
In other words: the comparison deliberately sets aside projected minutes - one of the hardest and most subjective parts of any projection model - and asks what happens when both models are handed the same assumption.
11 players times 90 minutes is 990 minutes.
So when you add up the projected minutes of an entire squad for a given fixture, the total has to be exactly 990. Not approximately.
Our models pass that check for every club in every gameweek, in every league we cover.
The official Premier League fantasy game and three more, each with a scoring table of its own.
Each game has different rules, and we compute each one separately against its exact scoring table - without assuming one game is "close enough" to another.
And there are two more important checks that did not make the three above.
In our Premier League model the bonus budget is 6.38 points per match - the average actually awarded across the 380 fixtures of the 2025/26 season - and not 6 points, as you might assume from the 3/2/1 split alone.
On top of that, the total projected goals distributed among a club’s players has to match the club’s own projected goals. If a team is projected to score 1.5 goals, you cannot hand its players a combined projection of 2.0. A model that does will systematically inflate the projections of attacking players, even if at first glance the numbers look reasonable.
A good projection is one you can check.
The scoring rules, exactly
Every fantasy game scores differently, which is why each model at Vice Captain is fitted separately to the rules of the game it serves.
For example, in the official Premier League fantasy game for 2026/27 a goalkeeper’s goal is worth 10 points and not 6, every three saves are worth a point, a point comes off for every two goals conceded, and there is an entirely new component - defensive contribution ("DefCon"): 2 points for a defender who reaches 10 defensive actions, and for a midfielder or forward who reaches 12.
A change like that affects the relative value of players in every position. A model that was not rebuilt around the new rules may carry on ranking players as though we were still in last season.
The same work is done separately for each of the four fantasy games we support.
What is actually hard here
The number of goals a team scored or conceded does not always tell us how good it really was.
A model that leans too heavily on last season’s results may learn the bottom line instead of the quality of the chances. A defence can concede little simply because its opponents missed a lot, or concede more than it deserved despite defending well.
Even across 38 gameweeks, those gaps do not always disappear.
That is why we do not train the model on the number of goals scored alone.
And the hardest part of an expected-points model is not projecting goals at all - it is projecting minutes.
You can judge perfectly well how much a team is likely to score, but if you are wrong about who starts, who comes off early and who is only in line for 20 or 30 minutes, the expected-points table will be wrong too.
A large share of the gap between different projection services comes not from who the better player is, but from how long each player is expected to be on the pitch.
That is also why the comparison above sets the minutes component aside: the aim is to test the projection model itself, with the minutes assumption identical on both sides.
Promoted clubs are a good illustration of the difficulty. They have no previous season in the same division to lean on, so you cannot simply treat them like another club with a full year of relevant data behind it.
And precisely because projected minutes are the hardest part, we do not require you to accept our estimate.
On the expected-points board you can change any player’s minutes, and the projection updates immediately.
If you think a particular player will start even though we expect otherwise, or that he is only in line for half an hour - just enter your own minutes and see straight away how it affects the expected points.
This is not only a convenience feature. It is a way of making room for your information and your knowledge exactly where even the best model carries the most uncertainty.
What goes in
The starting point for a significant part of the projections is the betting market.
Odds already price in a very large amount of information: team strength, home advantage, injuries, expected lineups, and new information entering the market through real money.
We do not copy the odds across as they are - we translate them into numbers that matter for fantasy.
Alongside them come official fantasy data such as prices and positions, predicted lineups and team news, player roles such as penalties and set-pieces, season statistics and the fixture list.
Where sources conflict, data with strong market backing generally carries more weight.
Ownership and effective ownership (EO) are shown as decision-support, but do not affect the expected-points calculation itself.
What you get
- Fixture difficulty (FDR) - a 1 to 5 rating based on the market’s read of the specific match, not on the club’s name or reputation.
- Projected goals and clean-sheet probability - per club and per fixture. Two of the most important base numbers the model rests on.
- Expected points (xPts) - goals, assists, clean sheets, goals conceded, saves, DefCon, minutes, bonus and cards, all translated into a single number you can compare players with.
- Goalscorer and assist odds - based on bookmaker player markets and shown as percentages.
- Predicted lineups - updated as news breaks, with the resulting minutes feeding straight into the expected-points calculation.
- Penalties and set-pieces - maintained separately, because these roles have a significant effect on a player’s attacking potential.
Limits, and who stands behind this
Projections are probabilities, not promises.
An injury, a red card, an early substitution, rotation or a surprise lineup can change entirely what actually happens. Football is a game with a great deal of variance, and in a single match it can be enormous.
The models are built and reviewed by Vice Captain’s founder, a long-time fantasy manager with proven experience and results. You can read more on the author page.
The data and the models do most of the work, and human review complements them where context is needed that numbers alone cannot supply.
And when we identify a component that is not yet good enough, the aim is to improve it - not to present a higher level of confidence than the data actually justifies.
Updates and corrections
Projections update at least once a day, and more frequently as a deadline approaches and especially around match days.
Every tool carries a timestamp showing when its data was last updated.
Each update replaces the previous data with the most current version, and material errors are corrected as soon as they are identified.
Seen something that looks wrong? Write to us and we will look into it.