BetLogic Prediction Research

Research Snapshot

Updated through 10 November 2026

Publication thresholds: 20 settled to display a cohort, 50 to state a finding, 100 for strong evidence.

Market Performance

Market cohorts with at least 20 settled predictions. Sample size and evidence tier are shown for every row, and no market is ranked as a better or safer choice than another.

October 2026

Last 90 days

All time

Competition Performance

Competition cohorts with at least 20 settled predictions. Competitions below that sample are not compared, because a handful of results cannot describe a pattern.

October 2026

Competition performance — October 2026

Competition	Landed	Missed	Void	Settled sample	Actual hit rate	Market baseline	Difference	Covered / eligible	Coverage	Evidence
Copa del ReySpain	16	18	0	34	—	—	—	0/34	0.0% · incomplete	Building evidence
Major League SoccerUSA	14	8	0	22	55.6%	77.0%	-21.4 pts	9/22	40.9% · incomplete	Building evidence

Last 90 days

Competition performance — Last 90 days

Competition	Landed	Missed	Void	Settled sample	Actual hit rate	Market baseline	Difference	Covered / eligible	Coverage	Evidence
Major League SoccerUSA	35	9	0	44	73.7%	75.6%	-1.9 pts	19/44	43.2% · incomplete	Building evidence
FA CupEngland	26	14	0	40	—	—	—	0/40	0.0% · incomplete	Building evidence
Copa del ReySpain	17	22	0	39	—	—	—	0/39	0.0% · incomplete	Building evidence
Serie AItaly	22	8	0	30	68.8%	77.4%	-8.6 pts	16/30	53.3% · incomplete	Building evidence
La LigaSpain	19	1	0	20	100.0%	76.1%	+23.9 pts	8/20	40.0% · incomplete	Building evidence
Premier LeagueEngland	17	3	0	20	71.4%	79.2%	-7.8 pts	7/20	35.0% · incomplete	Building evidence
Serie ABrazil	17	3	0	20	100.0%	72.5%	+27.5 pts	7/20	35.0% · incomplete	Building evidence

All time

Competition performance — All time

Competition	Landed	Missed	Void	Settled sample	Actual hit rate	Market baseline	Difference	Covered / eligible	Coverage	Evidence
Major League SoccerUSA	35	9	0	44	73.7%	75.6%	-1.9 pts	19/44	43.2% · incomplete	Building evidence
FA CupEngland	26	14	0	40	—	—	—	0/40	0.0% · incomplete	Building evidence
Copa del ReySpain	17	22	0	39	—	—	—	0/39	0.0% · incomplete	Building evidence
Serie AItaly	22	8	0	30	68.8%	77.4%	-8.6 pts	16/30	53.3% · incomplete	Building evidence
La LigaSpain	19	1	0	20	100.0%	76.1%	+23.9 pts	8/20	40.0% · incomplete	Building evidence
Premier LeagueEngland	17	3	0	20	71.4%	79.2%	-7.8 pts	7/20	35.0% · incomplete	Building evidence
Serie ABrazil	17	3	0	20	100.0%	72.5%	+27.5 pts	7/20	35.0% · incomplete	Building evidence

Internal Score Calibration Evidence

This internal learning evidence compares raw score cohorts with their observed outcomes. It remains separate from P1.2's canonical public confidence projection and is not used as a public probability owner.

Public Calibrated Confidence

The public confidence owner uses only canonical settled Main Tips. It fits monotone broad-band rates with pool-adjacent-violators on the earlier chronological evidence, then checks the final held-out period. Exact confidence is shown only for a band with at least 100 training samples, 50 held-out samples and a held-out gap of 5 percentage points or less. Otherwise public surfaces show a band and historical landed rate, never a raw trust score.

Failure Patterns

Missed Main Tips are classified into a fixed, deterministic set of failure categories. Only aggregate counts of losses are published — no individual match, fixture or underlying evidence is exposed, and a win is never counted here even when it was, in hindsight, under-confident.

454 classified missed Main Tips from 1635 settled predictions examined, 13 August 2026 — 10 November 2026.

What the Evidence Supports

Each statement below is generated from the tables above and is only produced when its own cohort reached at least 50 settled predictions.

• ObservationBetLogic Main Tips recorded 65.9% landed outcomes across 138 settled predictions in October 2026 (91 landed, 47 missed, 0 void).
• ObservationBetLogic Main Tips recorded 72.4% landed outcomes across 373 settled predictions over the last 90 days (270 landed, 103 missed, 0 void).
• ObservationBetLogic Main Tips recorded 72.4% landed outcomes across 373 settled predictions across the published record to date (270 landed, 103 missed, 0 void).
• ObservationDouble Chance markets recorded 65.9% landed outcomes across 138 settled Main Tips in October 2026.
• ObservationDouble Chance markets recorded 72.4% landed outcomes across 373 settled Main Tips over the last 90 days.
• ObservationDouble Chance markets recorded 72.4% landed outcomes across 373 settled Main Tips across the published record to date.
• ObservationThe 70-75 internal-score cohort was overconfident across 572 settled samples: observed hit rate was 67.7%, compared with an average internal score of 71.9/100.
• ObservationThe 80-85 internal-score cohort was overconfident across 567 settled samples: observed hit rate was 75.3%, compared with an average internal score of 82.1/100.
• ObservationThe 75-80 internal-score cohort was overconfident across 419 settled samples: observed hit rate was 73.7%, compared with an average internal score of 76.9/100.
• Observation"Result went the other way" was the most frequently classified error category in this research window, appearing 383 times across 454 classified samples.

Limitations

• Sample thresholds (20 / 50 / 100 settled) control publication eligibility only. They are conservative product gates and do not claim formal statistical significance.
• This research is descriptive. It reports what settled results show and does not establish cause or skill.
• ROI is unavailable because BetLogic did not store verified decision-time decimal odds for this historical sample. It is never reconstructed from implied probability.
• BetLogic Brier remains unavailable: P1.2 publishes a qualified band-level calibration mapping, not a retrospective per-selection probability scorecard.
• Void, unresolved and unsupported predictions never enter the hit-rate denominator.
• Historical performance does not guarantee future results.

Methodology

Research evidence can inform future BetLogic methodology changes, but this report does not claim a public methodology change unless that change is separately documented in the public Prediction Methodology. Internal prediction rules, configuration and reasoning traces are not published.