Fantasy Premier League
Regression candidates
Goals and assists are what the game pays for; expected goals and expected assists are what the chances behind them were worth. Over five rounds those two figures come apart, in both directions, and this page is the players furthest apart on each — read beside the same gap over the whole season, because whether a recent number is a change or noise is the question that actually matters.
Thierno Barry
FWD · EVE · 5.7m · owned by 8.3%
−1.65
Gap
2 returns over the last 5 rounds, against 3.65 expected behind them.
Gameweek 6 · 5-round window · feed of 01/10 11:39
Thierno Barry has the widest gap on the board: 2 goals and assists over 5 rounds from 3.65 expected, a shortfall of 1.65 involvements.
Over the whole season he is −1.65 the same way, so the last 5 rounds are a run rather than a week of it.
G−xG
−1.50
His goals over the last 5 rounds less the expected goals behind them.
A−xA
−0.15
His assists over the same 5 rounds less the expected assists behind them.
Mins
417
The minutes both gaps were read over. Under 180 a gap is a run of matches rather than a reading.
Of the 20 players your filters leave: 13 running level, 0 with no measured rounds behind them.
Running behind their chances
The board as your controls leave it — the same five that open the table below.
-
Ranked 1 Thierno Barry13.7 Ahead
- G−xG
- −1.50
- A−xA
- −0.15
- Season
- −1.65
Positive Full17 shots, 5 on target, 5 created in 5 matches
3.0- HUL (A)
- CHE (H)
- ARS (A)
- !
- COV (H)
-
Ranked 2 Gonzalo García13.4 Ahead
- G−xG
- −1.33
- A−xA
- −0.58
- Season
- −1.91
Strong positive Full15 shots, 2 on target, 3 created in 5 matches
2.8- IPS (A)
- HUL (H)
- COV (A)
- AVL (A)
- NEW (H)
-
Ranked 3 Igor Thiago13.1 Ahead
- G−xG
- −2.13
- A−xA
- −0.16
- Season
- −2.29
Strong positive Full13 shots, 4 on target, 2 created in 5 matches
3.2- AVL (A)
- LIV (H)
- HUL (A)
- NFO (H)
- BHA (A)
Running behind their chances
Their goals and assists over the last 5 rounds are under the expected figures behind them. That is a gap, and a gap is evidence about what has happened rather than a claim about what happens next.
- G−xG
- −1.50
- A−xA
- −0.15
- xGI/90
- 0.79
- Season
- −1.65
17 shots, 5 on target, 5 created in 5 matches
- G−xG
- −1.33
- A−xA
- −0.58
- xGI/90
- 0.62
- Season
- −1.91
15 shots, 2 on target, 3 created in 5 matches
- G−xG
- −2.13
- A−xA
- −0.16
- xGI/90
- 0.67
- Season
- −2.29
13 shots, 4 on target, 2 created in 5 matches
Running ahead of their chances
Their goals and assists over the last 5 rounds are over the expected figures behind them. Finishing runs like that are the most-owned players in the game about as often as anybody else.
- G−xG
- +0.65
- A−xA
- +2.94
- xGI/90
- 0.36
- Season
- +3.59
13 shots, 5 on target, 2 created in 5 matches
- G−xG
- +0.83
- A−xA
- +2.81
- xGI/90
- 0.59
- Season
- +3.64
11 shots, 4 on target, 7 created in 4 matches
What a gap is here, and what it is not
Every row on this page is a regression candidate, which means one thing: his returns over the last 5 rounds and the expected figures behind them are a long way apart. Nothing on this page says a goal is owed, and nothing here is backtested as a forecast — a gap can close because he starts scoring, because he stops playing, or because he was never the finisher the expected figures assumed.
Goals against expected goals and assists against expected assists are kept apart rather than added into one gap, because they are two different players' problems. The band is over both together; the two columns are each half of it.
The five bands are cut on the whole gap in involvements over the window: 1.75 or more either way is strong, 0.75 or more is a direction, and anything inside 0.75 is level. Totals rather than per-90 figures, deliberately — a gap divided by a substitute's twenty minutes would make him the most extreme finisher in the game.
The form column is expected goals plus expected assists per 90 over the window, which is the one figure here two players on different minutes can be compared on. It is blank under 180 minutes, because two matches is the least that averages anything.
A small sample is labelled rather than left out. A player with under 180 minutes in the window is marked thin and the gap beside him should be read as the couple of matches it is.
The season column is the same gap over the whole year, off Fantasy's own season totals — the figure the squad page prints for him. It is beside the window read rather than instead of it, because a recent gap against a season running the same way is a run, and one against a season running level is a departure.
Shots, shots on target and chances created come from this site's own match data over the same weeks, and are there only for the players it could match to one of its own. Where they are absent nothing is claimed either way — a dash is not nought shots.
A round the feed has published no expected figures for contributes nothing rather than a nought, and a player with no stored rounds at all is on neither list rather than credited with a gap of nothing.
Every ranking here is in expected points or in the weighted score for the rounds ahead, which are the units this model is measured in. The value column is those figures restated over a price — it is a way of scanning the board, and nothing on this site backtests it.
Bonus points are projected nowhere on this site, so they are in none of these figures.
The board is every player the game would let you field, the flagged included — whether a doubtful player is worth picking is a judgement the projection makes with his published chance of playing, not one made by hiding him.
What is measured here, and what is not
-
Measured The projected points
Replayed against gameweek 2 after it was played: an average miss of 1.617 points a player, against 1.842 for “predict what he has averaged so far”, over 456 players. Recorded Sunday, 6 September 2026 and refreshed by hand, so the date is part of the claim.
-
Not measured The form score
The gap between what a player has returned and the chances behind it, over the rounds stored. It is arithmetic over matches already played rather than a forecast, and no measurement here says whether a gap that wide closes.