Independent home, draw and away probabilities published before kick-off, with the original call kept on the match page through live play and final review.
2 upcoming0 live0 reviewed
Demo league record
Performance in this competition
Settled
6
Top-pick hit rate
66.7%
Brier score
0.173
Current board
Published match probabilities
Open a match to see its timestamp, model version, probability table and the factors behind the call. Fixture times are shown in WAT; readers in Kenya can add two hours for EAT.
Arsenal vs Aston Villa prediction
EPL
Premier League
Sat, 11 Jul · 15:00 WAT · 17:00 EAT
Pre-match
A
Arsenal
Home
vs15:00 WAT17:00 EAT
A
Aston Villa
Away
Model call
home
Confidence54%
Home54%
Draw26%
Away20%
What moved the model
Arsenal's home Elo is 84 points stronger after season regression.
Villa have taken four points from their last five away matches.
The rest-day gap gives Arsenal a small late-match advantage.
Chelsea vs Newcastle United prediction
EPL
Premier League
Sat, 11 Jul · 17:30 WAT · 19:30 EAT
Pre-match
C
Chelsea
Home
vs17:30 WAT19:30 EAT
N
Newcastle United
Away
Model call
home
Confidence44%
Home44%
Draw28%
Away28%
Competition-specific evidence
How Premier League predictions are evaluated
This England Premier League board publishes one locked home-draw-away probability set per covered fixture, then keeps that same pre-match forecast beside the final result.
League-level hit rate and Brier score are kept separate from the global record, so a strong result elsewhere cannot disguise weak Premier League calibration.
Competition
Premier League
Country
England
Forecast
Home · Draw · Away
Evaluation
Hit rate · Brier score · Calibration
Explore
Other covered leagues
Accountability
The record stays public
Demo model record
The numbers, including the misses.
Results are settled from the final score. No deleted picks, no edited probabilities, no hiding behind a winning streak.
Hit rate
54.2%
13 of 24 settled picks
Brier score
0.188
Lower is better
Hits / misses
13 / 11
54.2% observed
Calibration gap
14.8 pts
Weighted across buckets
Probability check
When we say 60%, how often does it happen?
Predicted vs actual
Model bucket
Predicted
Actual
Difference
Matches
Below 40%
37%
20%
-17 pts
5
40–49%
45%
71%
+26 pts
7
50–59%
55%
50%
-5 pts
8
60%+
65%
75%
+11 pts
4
Demo notice: accuracy figures on this preview are illustrative and are not a claim of live model performance.