SharpXI is in beta — updated daily. Spot something off? Tell us →In beta — updated daily. Tell us →
SharpXI© Join the waitlist
Track record

How the Premier League model actually performs

A model is only worth its calibration. Every number below is an out-of-sample prediction for a 25/26 Premier League fixture, made before kick-off from earlier data only, then checked against what actually happened. No cherry-picking, no hindsight.

Read this first. This page measures one thing: whether our probabilities are honest — do the events we call 30% happen about 30% of the time? It is not a profit record and not closing line value. Profit and CLV need real bookmaker odds, which this project doesn't have a live feed for yet; when it does, that scoreboard goes here too. Until then we publish what we can actually verify, and we grade it against a baseline so a confident-looking model can't hide.

114,252
predictions graded, out of sample
11/14
markets show a real edge over guessing
380
25/26 fixtures scored as-of kick-off

Every market — does the model call it straight, and beat guessing?

Two honest questions per market. First, are our percentages true — when we say something happens 30% of the time, does it? The bar shows it: grey is what we said, green is what actually happened; the closer they sit, the straighter the call. Second, is the model genuinely sharper than just predicting the season average for everyone? That's the badge.

✓ Sharp — clearly beats guessing (the model genuinely knows more here).  Slim edge — beats it, but only just; near the base rate.  ≈ No real edge — no better than guessing: the noisiest markets, where even a sharp model has little to add. The % is skill over a base-rate baseline — not a claim about beating a bookmaker. We leave the weak ones in, because a record you only show when it flatters you isn't one.
MarketWe said → it happenedSharper than guessing?
said 44%happened 52%
✓ Sharp · +23%
said 22%happened 24%
✓ Sharp · +22%
said 12%happened 11%
✓ Sharp · +17%
said 22%happened 24%
✓ Sharp · +15%
said 46%happened 50%
✓ Sharp · +15%
said 49%happened 52%
✓ Sharp · +14%
said 20%happened 20%
✓ Sharp · +14%
said 21%happened 21%
✓ Sharp · +9%
said 8%happened 7%
✓ Sharp · +8%
said 7%happened 5%
✓ Sharp · +8%
said 8%happened 7%
Slim edge · +4%
said 14%happened 14%
≈ No real edge
Team & match markets
said 56%happened 56%
≈ No real edge
said 58%happened 57%
✗ No edge
Show the full numbers — predictions, Brier score, baseline

Brier score measures per-prediction accuracy (lower is better); the baseline is what you'd score by predicting the same base rate for everyone. Skill is how far the model beats that baseline — the maths behind the badge above.

MarketPredictionsModel avg ActualBrierBaseline Skill
To be carded 9,521 14% 14% 0.121 0.122 +0.7%
1+ shot on target 9,521 22% 24% 0.156 0.185 +15.4%
2+ shots on target 9,521 7% 5% 0.047 0.051 +7.9%
1+ shot 9,521 44% 52% 0.192 0.250 +23.0%
2+ shots 9,521 22% 24% 0.143 0.184 +22.3%
3+ shots 9,521 12% 11% 0.079 0.095 +17.1%
1+ foul committed 9,521 49% 52% 0.214 0.250 +14.2%
2+ fouls committed 9,521 21% 21% 0.150 0.165 +8.9%
3+ fouls committed 9,521 8% 7% 0.060 0.063 +3.7%
1+ foul won 9,521 46% 50% 0.213 0.250 +14.7%
2+ fouls won 9,521 20% 20% 0.139 0.161 +13.5%
3+ fouls won 9,521 8% 7% 0.057 0.062 +8.2%
Team & match markets
Match Over 2.5 goals 342 56% 56% 0.242 0.246 +1.7%
Match Over 3.5 cards 342 58% 57% 0.254 0.245 -3.5%

Calibration, up close

The aggregate can look right while the details are wrong, so here's the finer view. We sort every prediction into groups by how confident the model was — all the ~20% calls, all the ~30% calls, and so on — then check how often each group actually happened. The model is well calibrated when the two line up: when the things it calls 30% happen about 30% of the time. In each row, grey is what the model said and green is what actually happened — the closer they are, the better. Tap any market to open it.

What the model said What actually happened
To be carded
Model saidhow often it happenedActual
6%
9%
15%
15%
24%
20%
34%
25%
43%
26%
1+ shot on target
Model saidhow often it happenedActual
4%
5%
15%
16%
24%
24%
35%
37%
44%
52%
53%
65%
62%
82%
2+ shots on target
Model saidhow often it happenedActual
3%
2%
14%
11%
24%
21%
34%
34%
1+ shot
Model saidhow often it happenedActual
1%
1%
16%
18%
26%
29%
35%
42%
45%
52%
55%
61%
65%
76%
74%
89%
82%
96%
2+ shots
Model saidhow often it happenedActual
4%
4%
15%
15%
24%
23%
35%
37%
45%
53%
55%
67%
63%
78%
3+ shots
Model saidhow often it happenedActual
3%
2%
14%
10%
25%
23%
34%
34%
44%
48%
54%
58%
1+ foul committed
Model saidhow often it happenedActual
2%
2%
16%
23%
26%
34%
35%
41%
45%
49%
55%
59%
65%
66%
74%
75%
82%
85%
2+ fouls committed
Model saidhow often it happenedActual
4%
6%
15%
16%
25%
24%
34%
34%
44%
38%
54%
56%
3+ fouls committed
Model saidhow often it happenedActual
4%
4%
14%
11%
24%
16%
34%
28%
1+ foul won
Model saidhow often it happenedActual
7%
15%
15%
16%
25%
28%
35%
36%
45%
49%
55%
60%
65%
70%
74%
78%
83%
87%
2+ fouls won
Model saidhow often it happenedActual
5%
5%
15%
14%
25%
25%
35%
36%
44%
45%
54%
56%
64%
61%
3+ fouls won
Model saidhow often it happenedActual
3%
2%
14%
11%
24%
20%
34%
32%
44%
36%
Match Over 2.5 goals
Model saidhow often it happenedActual
47%
44%
56%
56%
64%
66%
Match Over 3.5 cards
Model saidhow often it happenedActual
46%
54%
56%
58%
64%
56%
74%
57%

Where the model earns its keep

The shots and fouls markets are the bulk of what the tool prices, and they're what the model does best: it beats a base-rate baseline by up to 23% on them — scored using only matches played before each game, so nothing is fitted with hindsight. That's most of the board, and the part the tool is built to price.

We show the ones that don’t, too. To be carded, Match Over 2.5 goals, Match Over 3.5 cards came out no better than guessing the league average this season — the noisiest markets on the board, where even a sharp model has little to add. We leave them on the page because a track record you only show when it flatters you isn't one.

The scoreboard that's still missing. Calibration proves the probabilities are sound. It doesn't prove they beat a bookmaker — for that you need to consistently beat the closing line, and that needs a live odds feed. We explain why that's the real test in the closing line value guide, and it's the next thing this page will show.

Method: player markets scored as-of each kick-off using only prior matches (the recency-weighted rate never sees the future); team markets fit on earlier seasons and evaluated on 25/26. Regenerated from source data on 2026-07-28. Full method on the methodology page.

The scoreboard above is free, and it always will be. The paid tool is that same model, pointed at this week's prices — the edge, and the confidence, on every leg.