Monthly report

Prediction Accuracy: September 2026

Every Best Pick and Value Analysis of the month, graded against the final result. Hits and misses count alike; void matches are left out.

Final figures for September 2026

Hit rate in September 2026

Best Picks

The single highest-conviction selection per fixture, across all markets.

72.6%
Hit rate · 1,358 of 1,871 correct

+2.6 pts vs August 2026

Value Analysis

Selections where the model's probability diverges most from consensus market pricing.

58.1%
Hit rate · 843 of 1,451 correct

+1.0 pts vs August 2026

Compared to guessing

An accuracy figure means nothing on its own. The only useful question is what a naive rule would have scored on exactly the same matches.

50.6%
Model, match result
43.7%
Always pick the home team
+6.9
Difference

Measured across every graded fixture, not a selected subset. Home advantage is why the naive rule scores as well as it does; the gap above it is what the model contributes. n = 7,668.

Day by day in September 2026

Each line is a seven-day rolling average of settled selections, with a node marking the days behind it. Days with nothing settled break the line rather than being bridged across.

Best PicksValue Analysis
Day by day in September 2026
Day by day in September 2026
DateBest PicksValue Analysis
01.09.202673.1%52.3%
02.09.202673.1%52.7%
03.09.202673.4%53.5%
04.09.202672.4%53.5%
05.09.202668.5%55.6%
06.09.202671.0%55.1%
07.09.202670.3%56.9%
08.09.202671.7%57.4%
09.09.202670.9%58.6%
10.09.202671.4%60.0%
11.09.202671.8%61.4%
12.09.202672.4%59.2%
13.09.202670.1%60.5%
14.09.202670.5%60.8%
15.09.202669.9%60.3%
16.09.202669.7%59.3%
17.09.202669.3%57.5%
18.09.202668.5%57.5%
19.09.202670.8%58.9%
20.09.202672.0%56.8%
21.09.202672.6%54.9%
22.09.202673.0%55.4%
23.09.202673.2%55.5%
24.09.202673.0%57.5%
25.09.202672.7%54.8%
26.09.202673.7%53.6%
27.09.202675.0%58.4%
28.09.202675.4%60.8%
29.09.202675.1%59.8%
30.09.202675.4%60.2%

Accuracy by competition in September 2026

The competitions with the most graded Best Picks this month, each with at least 20.

CompetitionBest PicksValue Analysis
USAUSA: Major League Soccer68.9%42 of 61 correct70.6%48 of 68 correct
ArgentinaArgentina: Primera Nacional83.3%40 of 48 correct64.3%9 of 14 correct
WorldWorld: UEFA Nations League66.7%24 of 36 correct68.8%22 of 32 correct
ArgentinaArgentina: Primera C82.9%29 of 35 correct100.0%5 of 5 correct
ArgentinaArgentina: Primera B Metropolitana69.2%18 of 26 correct83.3%10 of 12 correct
ArgentinaArgentina: Torneo Federal A69.2%18 of 26 correct73.3%11 of 15 correct
EnglandEngland: National League80.0%20 of 25 correct77.8%7 of 9 correct
USAUSA: MLS Next Pro79.2%19 of 24 correct58.3%7 of 12 correct
MexicoMexico: Liga MX62.5%15 of 24 correct37.9%11 of 29 correct
BrazilBrazil: Serie B77.3%17 of 22 correct46.9%15 of 32 correct

Are the probabilities honest?

This is the test that matters. We group every match-result forecast by the probability we published, then plot what actually happened. A calibrated model sits on the diagonal.

Perfect calibrationOur model95% interval
How often it actually happened
Are the probabilities honest?
Probability we published

What this chart coversMatch result across every graded fixture, not just selected picks - a selected subset could not test calibration honestly. Buckets holding too few fixtures to plot are merged into the end points, and point size scales with sample. Vertical bars are 95% intervals: a thin bucket is an uncertain reading, not a precise miss.

Brier score: 0.2001Mean squared error between published probability and outcome. Lower is better; 0.25 is what you would score by predicting 50% on everything.

Calibration by probability bin

Probability bandProbability we publishedHow often it actually happened95% intervalFixtures
0.05-0.2014.5%14.4%13.2% – 15.7%2969
0.20-0.2522.9%23.4%22.1% – 24.7%4060
0.25-0.3027.2%27.1%25.9% – 28.4%4893
0.30-0.3532.3%31.3%29.5% – 33.1%2576
0.35-0.4037.5%35.8%33.9% – 37.8%2322
0.40-0.4542.3%44.0%41.6% – 46.3%1670
0.45-0.5047.5%46.2%43.6% – 48.8%1387
0.50-0.5552.2%53.4%50.5% – 56.4%1097
0.55-0.6057.2%57.4%53.6% – 61.1%671
0.60-0.6562.3%67.1%62.8% – 71.2%477
0.65-0.7067.4%70.5%65.4% – 75.2%326
0.70-0.7572.5%75.1%69.3% – 80.1%241
0.75-0.8077.3%79.5%72.2% – 85.2%146
0.80-0.9585.6%79.9%73.2% – 85.2%169

Interval is the 95% range the observed figure could take at that sample size. Narrower means more matches in the bin.

Which markets the picks came from

A hit rate across a mixed basket is not the same claim as a hit rate on match result alone. These are the six most-used markets in each basket, with the remainder grouped, so you can judge both headline figures for yourself.

Best Picks

  • Home Win or Draw56.9% of picksn = 1,065
  • Home Win18.4% of picksn = 344
  • Draw or Away Win8.9% of picksn = 167
  • Away Win5.0% of picksn = 93
  • Over 2.5 Goals4.2% of picksn = 79
  • Over 1.5 Goals3.9% of picksn = 73
  • Other markets (3)2.7% of picksn = 50

Value Analysis

  • Home Win11.5% of picksn = 167
  • Home Win or Draw11.0% of picksn = 160
  • Over 2.5 Goals10.8% of picksn = 156
  • Under 2.5 Goals6.5% of picksn = 95
  • Cards Under 4.56.4% of picksn = 93
  • Corners Under 10.56.1% of picksn = 89
  • Other markets (24)48.0% of picksn = 691

77 void matches were excluded from these figures.

Monthly reports

How the predictions did in September 2026

1358 of 1871 graded Best Picks were correct in September 2026, a hit rate of 72.6%. Value Analysis was right in 843 of 1451 matches (58.1%). That is 2.6 points more than in August 2026.

On match result, the model was right in 50.6% of 7668 matches, against 43.7% for always picking the home team. Among competitions with at least 20 graded Best Picks, Primera Nacional had the highest hit rate: 83.3% of 48.

Every prediction is graded against the result after 90 minutes, hits and misses alike, and void matches are left out. The all-time record and the method are on the Accuracy page.