How accurate is our model?

This page is about the statistical model only: its forecasts as they were published before kick-off, scored once the match was over. Forecasts made after kick-off are never counted. The founder’s analysis and the community’s predictions are separate and are not part of these figures.

All competitions

70 matches scored.

By month of kick-off (UTC). For RPS, Brier and log loss, lower is better.
Month	Matches	Correct	RPS	RPS if guessing	Brier	Log loss	Coverage
October 2026	70	54.1%	0.188	0.235	0.567	0.941	available
All months	70	54.1%	0.188	0.235	0.567	0.941	available

Model versions: [email protected], [email protected].

Last scored on 10 October 2026.

By competition

Belgian Pro League

Only 5 matches have been scored so far, fewer than 30: an early record, not a verdict.

By month of kick-off (UTC). For RPS, Brier and log loss, lower is better.
Month	Matches	Correct	RPS	RPS if guessing	Brier	Log loss	Coverage
October 2026	5	75.0%	0.181	0.278	0.423	0.739	limited
All months	5	75.0%	0.181	0.278	0.423	0.739	limited

Model versions: [email protected], [email protected].

Bundesliga

Only 7 matches have been scored so far, fewer than 30: an early record, not a verdict.

By month of kick-off (UTC). For RPS, Brier and log loss, lower is better.
Month	Matches	Correct	RPS	RPS if guessing	Brier	Log loss	Coverage
October 2026	7	42.1%	0.314	0.234	0.860	1.349	limited
All months	7	42.1%	0.314	0.234	0.860	1.349	limited

Model versions: [email protected], [email protected].

Championship

Only 11 matches have been scored so far, fewer than 30: an early record, not a verdict.

By month of kick-off (UTC). For RPS, Brier and log loss, lower is better.
Month	Matches	Correct	RPS	RPS if guessing	Brier	Log loss	Coverage
October 2026	11	34.5%	0.216	0.238	0.627	1.037	limited
All months	11	34.5%	0.216	0.238	0.627	1.037	limited

Model versions: [email protected], [email protected].

Eredivisie

Only 5 matches have been scored so far, fewer than 30: an early record, not a verdict.

By month of kick-off (UTC). For RPS, Brier and log loss, lower is better.
Month	Matches	Correct	RPS	RPS if guessing	Brier	Log loss	Coverage
October 2026	5	33.3%	0.155	0.167	0.715	1.139	limited
All months	5	33.3%	0.155	0.167	0.715	1.139	limited

Model versions: [email protected], [email protected].

La Liga

Only 4 matches have been scored so far, fewer than 30: an early record, not a verdict.

By month of kick-off (UTC). For RPS, Brier and log loss, lower is better.
Month	Matches	Correct	RPS	RPS if guessing	Brier	Log loss	Coverage
October 2026	4	77.8%	0.103	0.241	0.333	0.611	limited
All months	4	77.8%	0.103	0.241	0.333	0.611	limited

Model versions: [email protected], [email protected].

Ligue 1

Only 6 matches have been scored so far, fewer than 30: an early record, not a verdict.

By month of kick-off (UTC). For RPS, Brier and log loss, lower is better.
Month	Matches	Correct	RPS	RPS if guessing	Brier	Log loss	Coverage
October 2026	6	50.0%	0.176	0.247	0.521	0.882	limited
All months	6	50.0%	0.176	0.247	0.521	0.882	limited

Model versions: [email protected], [email protected].

Persian Gulf Pro League

Only 9 matches have been scored so far, fewer than 30: an early record, not a verdict.

By month of kick-off (UTC). For RPS, Brier and log loss, lower is better.
Month	Matches	Correct	RPS	RPS if guessing	Brier	Log loss	Coverage
October 2026	9	80.0%	0.088	0.178	0.409	0.687	limited
All months	9	80.0%	0.088	0.178	0.409	0.687	limited

Model versions: [email protected], [email protected].

Premier League

Only 6 matches have been scored so far, fewer than 30: an early record, not a verdict.

By month of kick-off (UTC). For RPS, Brier and log loss, lower is better.
Month	Matches	Correct	RPS	RPS if guessing	Brier	Log loss	Coverage
October 2026	6	53.9%	0.191	0.226	0.608	1.003	limited
All months	6	53.9%	0.191	0.226	0.608	1.003	limited

Model versions: [email protected], [email protected].

Primeira Liga

Only 5 matches have been scored so far, fewer than 30: an early record, not a verdict.

By month of kick-off (UTC). For RPS, Brier and log loss, lower is better.
Month	Matches	Correct	RPS	RPS if guessing	Brier	Log loss	Coverage
October 2026	5	100.0%	0.136	0.278	0.334	0.630	limited
All months	5	100.0%	0.136	0.278	0.334	0.630	limited

Model versions: [email protected], [email protected].

Scottish Premiership

Only 4 matches have been scored so far, fewer than 30: an early record, not a verdict.

By month of kick-off (UTC). For RPS, Brier and log loss, lower is better.
Month	Matches	Correct	RPS	RPS if guessing	Brier	Log loss	Coverage
October 2026	4	50.0%	0.185	0.278	0.420	0.742	limited
All months	4	50.0%	0.185	0.278	0.420	0.742	limited

Model versions: [email protected], [email protected].

Serie A

Only 3 matches have been scored so far, fewer than 30: an early record, not a verdict.

By month of kick-off (UTC). For RPS, Brier and log loss, lower is better.
Month	Matches	Correct	RPS	RPS if guessing	Brier	Log loss	Coverage
October 2026	3	42.9%	0.196	0.278	0.450	0.786	limited
All months	3	42.9%	0.196	0.278	0.450	0.786	limited

Model versions: [email protected], [email protected].

Süper Lig

Only 5 matches have been scored so far, fewer than 30: an early record, not a verdict.

By month of kick-off (UTC). For RPS, Brier and log loss, lower is better.
Month	Matches	Correct	RPS	RPS if guessing	Brier	Log loss	Coverage
October 2026	5	53.9%	0.171	0.201	0.651	1.031	limited
All months	5	53.9%	0.171	0.201	0.651	1.031	limited

Model versions: [email protected], [email protected].

UEFA Champions League

No forecast made before kick-off has been scored here yet, so there is no accuracy to show.

UEFA Conference League

No forecast made before kick-off has been scored here yet, so there is no accuracy to show.

UEFA Europa League

No forecast made before kick-off has been scored here yet, so there is no accuracy to show.

What the figures mean

Matches: how many matches were scored. A match can have more than one forecast before kick-off (an early one, another once the line-ups are known); each is scored, and the averages are over all of them.
Correct: how often the outcome the model thought most likely (a home win, a draw or an away win) happened. Picking at random gets about 33% right.
RPS (ranked probability score): how far the forecast’s probabilities were from the result, with part credit when the result was next to the outcome the model favoured, since a draw is closer to a home win than an away win is. 0 is perfect. “RPS if guessing” is what a forecast of one third each would have scored on the same matches.
Brier score: the squared distance between the probabilities and what happened. Lower is better; a forecast of one third each scores 0.667.
Log loss: how surprised the model was by the result. Lower is better; a forecast of one third each scores 1.099, and being confidently wrong costs the most.
Below 30 matches a figure is marked limited: a few results can make a model look better or worse than it is.