Premier League · several engines · one identical exam

The engine room

Most sites show you one model and ask you to trust it. We run several, put them through an identical exam, and show you the marks — including the ones we would rather not publish. Pick whichever you want to follow.

This leaderboard is the Premier League's. The other four leagues each run a single engine rather than a field of challengers, so there is nothing to rank — their held-out record, their calibration and their week-by-week ledger are on the record, one league at a time.

Over 1900 held-out matches the best engine scores 0.1982 against the closing market's 0.1949 — closer to the sharpest benchmark there is than any public model we know, and every week of it is on the record, including the bad ones.
Algorithm of the week
xGB Momentum

Best ranked-probability score over the last 8 weeks of matches (since 2026-03-29). This changes as form changes — which is rather the point of running more than one model. The full week-by-week record is on the weekly record.

Every model is fitted on 2016–2020 only, then run forward week by week across 2021–2025 without ever seeing those seasons during tuning. Same data, same refit schedule, same metrics — the only thing that differs is the algorithm.

1

Closing market

benchmark — not selectable

Not ours: the betting market's closing prices, the strongest public predictor in football. The bar every engine is judged against.

RPS
0.19491
95% 0.1889–0.2007
Accuracy
55.2%
1900 matches
Log loss
0.9597
Brier
0.5700
2

xGB Core

The house engine. Reads the quality of every chance a side creates and concedes — not just results — adjusted for opponent and venue, weighted toward now. The champion until another engine beats it where it counts.

RPS
0.19824
95% 0.1924–0.2045
Accuracy
54.5%
1900 matches
Log loss
0.9707
Brier
0.5769
Season record — since this engine started running
Season RPS
0.20810
Season accuracy
49.0%
Live RPS
0.20810
Graded live
100
Reconstructed
0
GameweeknRPSAccuracy
GW1100.229270.0%
GW2100.189560.0%
GW3100.139130.0%
GW4100.241130.0%
GW5100.238340.0%
GW6100.209050.0%
GW7100.179760.0%
GW8100.247730.0%
GW9100.227760.0%
GW10100.179560.0%
Opponent-tier breakdown

Draws: caught 0 of the 27 that actually happened.

Top 6 involved · called 26 of 51 · RPS 0.1861
Mid-table involved · called 31 of 65 · RPS 0.2061
Relegation contenders involved · called 26 of 51 · RPS 0.2164
Every tier pairing seen so far (6 of 6)

Tiers from last season's final table (2025/26). A promoted club with no row in that table counts as a relegation contender here.

MatchupnCalledRPS
Mid-table vs Mid-table157/150.2108
Mid-table vs Relegation contenders2513/250.2287
Mid-table vs Top 62511/250.1808
Relegation contenders vs Relegation contenders93/90.2712
Relegation contenders vs Top 61710/170.1693
Top 6 vs Top 695/90.2322
Top 6
Arsenal, Aston Villa, Bournemouth, Liverpool, Man City, Man United
Mid-table
Brentford, Brighton, Chelsea, Everton, Fulham, Leeds, Newcastle, Sunderland
Relegation contenders
Burnley, Crystal Palace, Nottingham Forest, Tottenham, West Ham, Wolves
See xGB Core's calls on the board →
3

xGB Consensus

The engines vote. Disagreement between them is information, and the vote is settled before the season starts — never adjusted in hindsight.

RPS
0.19830
95% 0.1927–0.2044
Accuracy
54.4%
1900 matches
Log loss
0.9704
Brier
0.5769

Against the reigning champion engine over the same 1900 matches: worse by 0.00006 RPS (p = 0.8549) — inside noise, so we do not claim a difference.

Season record — since this engine started running
Season RPS
0.20770
Season accuracy
49.0%
Live RPS
0.20690
Graded live
80
Reconstructed
20
backfilled, not recorded live
GameweeknRPSAccuracy
GW1↺ reconstructed100.237570.0%
GW2↺ reconstructed100.184260.0%
GW3100.137730.0%
GW4100.228840.0%
GW5100.237440.0%
GW6100.209250.0%
GW7100.179260.0%
GW8100.253020.0%
GW9100.224260.0%
GW10100.186060.0%
Opponent-tier breakdown

Draws: caught 0 of the 27 that actually happened.

Top 6 involved · called 25 of 51 · RPS 0.1871
Mid-table involved · called 31 of 65 · RPS 0.2064
Relegation contenders involved · called 27 of 51 · RPS 0.2143
Every tier pairing seen so far (6 of 6)

Tiers from last season's final table (2025/26). A promoted club with no row in that table counts as a relegation contender here.

MatchupnCalledRPS
Mid-table vs Mid-table157/150.2120
Mid-table vs Relegation contenders2514/250.2250
Mid-table vs Top 62510/250.1844
Relegation contenders vs Relegation contenders93/90.2693
Relegation contenders vs Top 61710/170.1693
Top 6 vs Top 695/90.2285
Top 6
Arsenal, Aston Villa, Bournemouth, Liverpool, Man City, Man United
Mid-table
Brentford, Brighton, Chelsea, Everton, Fulham, Leeds, Newcastle, Sunderland
Relegation contenders
Burnley, Crystal Palace, Nottingham Forest, Tottenham, West Ham, Wolves

Includes 20 reconstructed calls above — backfilled, not recorded live.

See xGB Consensus's calls on the board →
4

xGB Current

Ratings that move with every match instead of being rebuilt from a window. The quickest engine in the room to accept that a side has changed, and the least attached to what they used to be.

RPS
0.19892
95% 0.1932–0.2050
Accuracy
53.9%
1900 matches
Log loss
0.9723
Brier
0.5782

Against the reigning champion engine over the same 1900 matches: worse by 0.00068 RPS (p = 0.2622) — inside noise, so we do not claim a difference.

Season record — since this engine started running
Season RPS
0.20810
Season accuracy
48.0%
Live RPS
0.20560
Graded live
81
Reconstructed
19
backfilled, not recorded live
GameweeknRPSAccuracy
GW1↺ reconstructed100.246770.0%
GW210.0967100.0%
GW2↺ reconstructed90.188344.4%
GW3100.137130.0%
GW4100.218840.0%
GW5100.237250.0%
GW6100.210740.0%
GW7100.178960.0%
GW8100.258420.0%
GW9100.221360.0%
GW10100.193160.0%
Opponent-tier breakdown

Draws: caught 0 of the 27 that actually happened.

Top 6 involved · called 25 of 51 · RPS 0.1887
Mid-table involved · called 29 of 65 · RPS 0.2073
Relegation contenders involved · called 28 of 51 · RPS 0.2134
Every tier pairing seen so far (6 of 6)

Tiers from last season's final table (2025/26). A promoted club with no row in that table counts as a relegation contender here.

MatchupnCalledRPS
Mid-table vs Mid-table156/150.2133
Mid-table vs Relegation contenders2514/250.2229
Mid-table vs Top 6259/250.1881
Relegation contenders vs Relegation contenders93/90.2685
Relegation contenders vs Top 61711/170.1704
Top 6 vs Top 695/90.2250
Top 6
Arsenal, Aston Villa, Bournemouth, Liverpool, Man City, Man United
Mid-table
Brentford, Brighton, Chelsea, Everton, Fulham, Leeds, Newcastle, Sunderland
Relegation contenders
Burnley, Crystal Palace, Nottingham Forest, Tottenham, West Ham, Wolves

Includes 19 reconstructed calls above — backfilled, not recorded live.

See xGB Current's calls on the board →
5

xGB Momentum

Short memory on purpose: what have you done lately? Streaks and slumps count for more here than anywhere else in the room.

RPS
0.20047
95% 0.1943–0.2070
Accuracy
54.2%
1900 matches
Log loss
0.9790
Brier
0.5825

Against the reigning champion engine over the same 1900 matches: worse by 0.00223 RPS (p = 0.0316).

Season record — since this engine started running
Season RPS
0.22770
Season accuracy
44.0%
Live RPS
0.22340
Graded live
81
Reconstructed
19
backfilled, not recorded live
GameweeknRPSAccuracy
GW1↺ reconstructed100.234250.0%
GW210.26450.0%
GW2↺ reconstructed90.258844.4%
GW3100.190520.0%
GW4100.280520.0%
GW5100.246550.0%
GW6100.214950.0%
GW7100.222750.0%
GW8100.260130.0%
GW9100.208650.0%
GW10100.159180.0%
Opponent-tier breakdown

Draws: caught 0 of the 27 that actually happened.

Top 6 involved · called 23 of 51 · RPS 0.2132
Mid-table involved · called 30 of 65 · RPS 0.2137
Relegation contenders involved · called 24 of 51 · RPS 0.2296
Every tier pairing seen so far (6 of 6)

Tiers from last season's final table (2025/26). A promoted club with no row in that table counts as a relegation contender here.

MatchupnCalledRPS
Mid-table vs Mid-table156/150.2376
Mid-table vs Relegation contenders2512/250.2274
Mid-table vs Top 62512/250.1856
Relegation contenders vs Relegation contenders93/90.2937
Relegation contenders vs Top 6179/170.1990
Top 6 vs Top 692/90.3166
Top 6
Arsenal, Aston Villa, Bournemouth, Liverpool, Man City, Man United
Mid-table
Brentford, Brighton, Chelsea, Everton, Fulham, Leeds, Newcastle, Sunderland
Relegation contenders
Burnley, Crystal Palace, Nottingham Forest, Tottenham, West Ham, Wolves

Includes 19 reconstructed calls above — backfilled, not recorded live.

See xGB Momentum's calls on the board →
6

xGB Ladder

A long-memory strength ladder: every result moves both clubs up or down, big upsets move them further. Slow to convince, hard to fool.

RPS
0.20075
95% 0.1950–0.2069
Accuracy
53.6%
1900 matches
Log loss
0.9781
Brier
0.5823

Against the reigning champion engine over the same 1900 matches: worse by 0.00251 RPS (p = 0.0268).

Season record — since this engine started running
Season RPS
0.20550
Season accuracy
51.0%
Live RPS
0.20600
Graded live
81
Reconstructed
19
backfilled, not recorded live
GameweeknRPSAccuracy
GW1↺ reconstructed100.189960.0%
GW2↺ reconstructed90.218244.4%
GW210.1189100.0%
GW3100.151940.0%
GW4100.259030.0%
GW5100.222740.0%
GW6100.199560.0%
GW7100.173070.0%
GW8100.242130.0%
GW9100.234060.0%
GW10100.174870.0%
Opponent-tier breakdown

Draws: caught 0 of the 27 that actually happened.

Top 6 involved · called 27 of 51 · RPS 0.1897
Mid-table involved · called 32 of 65 · RPS 0.2039
Relegation contenders involved · called 27 of 51 · RPS 0.2074
Every tier pairing seen so far (6 of 6)

Tiers from last season's final table (2025/26). A promoted club with no row in that table counts as a relegation contender here.

MatchupnCalledRPS
Mid-table vs Mid-table157/150.2187
Mid-table vs Relegation contenders2514/250.2120
Mid-table vs Top 62511/250.1870
Relegation contenders vs Relegation contenders93/90.2553
Relegation contenders vs Top 61710/170.1753
Top 6 vs Top 696/90.2244
Top 6
Arsenal, Aston Villa, Bournemouth, Liverpool, Man City, Man United
Mid-table
Brentford, Brighton, Chelsea, Everton, Fulham, Leeds, Newcastle, Sunderland
Relegation contenders
Burnley, Crystal Palace, Nottingham Forest, Tottenham, West Ham, Wolves

Includes 19 reconstructed calls above — backfilled, not recorded live.

See xGB Ladder's calls on the board →
7

xGB Pattern

Learns the result distribution straight from dozens of form and context signals instead of modelling goals first. The most data-hungry engine in the room, and so far the least convincing.

RPS
0.20414
95% 0.1982–0.2103
Accuracy
52.8%
1900 matches
Log loss
0.9938
Brier
0.5909

Against the reigning champion engine over the same 1900 matches: worse by 0.00590 RPS (p = 0.0001).

Season record — since this engine started running
Season RPS
0.22980
Season accuracy
37.0%
Live RPS
0.22270
Graded live
81
Reconstructed
19
backfilled, not recorded live
GameweeknRPSAccuracy
GW1↺ reconstructed100.286730.0%
GW210.23940.0%
GW2↺ reconstructed90.230333.3%
GW3100.134850.0%
GW4100.221220.0%
GW5100.245130.0%
GW6100.266620.0%
GW7100.214260.0%
GW8100.269840.0%
GW9100.238540.0%
GW10100.189450.0%
Opponent-tier breakdown

Draws: caught 2 of the 27 that actually happened (called 12 in total, including wrong calls).

Top 6 involved · called 21 of 51 · RPS 0.2107
Mid-table involved · called 24 of 65 · RPS 0.2234
Relegation contenders involved · called 22 of 51 · RPS 0.2182
Every tier pairing seen so far (6 of 6)

Tiers from last season's final table (2025/26). A promoted club with no row in that table counts as a relegation contender here.

MatchupnCalledRPS
Mid-table vs Mid-table154/150.2723
Mid-table vs Relegation contenders2510/250.2262
Mid-table vs Top 62510/250.1914
Relegation contenders vs Relegation contenders92/90.2772
Relegation contenders vs Top 61710/170.1754
Top 6 vs Top 691/90.3308
Top 6
Arsenal, Aston Villa, Bournemouth, Liverpool, Man City, Man United
Mid-table
Brentford, Brighton, Chelsea, Everton, Fulham, Leeds, Newcastle, Sunderland
Relegation contenders
Burnley, Crystal Palace, Nottingham Forest, Tottenham, West Ham, Wolves

Includes 19 reconstructed calls above — backfilled, not recorded live.

See xGB Pattern's calls on the board →

Teams are not players

The model that best picks match outcomes is not automatically the one you want for “will he score?”. A match forecast depends on the difference between two sides; a player’s chance of scoring depends on the level of his own team’s expected goals. A model can get the difference right while running high or low on the level, and the match leaderboard would never show it. So we grade each model on players too, over 9,008 appearances.

ModelGoal BrierBiasTop-10 boardSeparation
xGB Core0.07503-0.98%27.5%2.23×
xGB Momentum0.07547-0.84%28.9%2.39×
xGB LadderCannot answer this question — it rates teams, not scorelines, so it has no expected-goals number to hand a player model. We leave the row empty rather than invent one.

Separation is how much more often the model’s top-ten board returns a goal or assist than everyone else — the number that actually matters when you are picking a captain. The two are close, and they disagree about which is best: the champion is better calibrated, while rolling form separates its top picks more sharply. Graded only on players who actually appeared — the projection is conditional on playing, and we do not claim to know the teamsheet.

Tuned on 2016, 2017, 2018, 2019, 2020 · evaluated on 2021, 2022, 2023, 2024, 2025 · generated 2026-08-21