World Cup 2026 — knockout stage live.See the bracket →

> data_report · published 1 July 2026

The 2026 Football Prediction Accuracy Report

We logged 851 football picks before kickoff, across 46 competitions, and let live results settle them automatically. The headline: Bawler's free Banker pick landed 73.8% of the time (628 of 851). Just as important — picks land close to the confidence Bawler gives them, which is the real test of whether a prediction means anything.

851 settled picks · 46 competitions · 18 March 2026 – 1 July 2026

Every figure here is recomputable from the public ledger at bawler.ai/track-record.

Banker hit rate
73.8%
628/851 settled
Value hit rate
50.6%
413/816 settled
Dark Horse hit rate
27.4%
176/642 settled
Picks studied
851
46 competitions

The finding: the confidence is honest

Anyone can post one lucky prediction. The harder question is whether a tool's confidence is trustworthy — when it says a pick is 75% likely, does it actually land about three times in four? We grouped every single-outcome pick by the confidence Bawler assigned it, then compared that number to how often those picks really won. The two columns track each other closely across the whole range.

Confidence Bawler assignedAvg. predictedActually wonPicks
20–30%25.5%27.2%604
30–40%34.7%32.3%31
40–50%45.3%44.0%225
50–60%52.9%52.4%569
60–70%67.1%73.8%61
70–80%75.3%71.1%537
80–90%83.6%79.4%243
90–100%92.2%90.0%20

Bands with fewer than 20 picks are omitted to avoid small-sample noise. "Actually won" is the share of picks in that confidence band that settled as a win.

Banker hit rate by competition

The free Banker pick — the safest of the three — broken down by league, for the competitions with the most settled picks. These are the regional cuts most useful for a specific desk.

CompetitionBanker hit rateSettled picks
International Friendly75.6%96/127
MLS72.4%76/105
World Cup65.2%45/69
La Liga74.1%43/58
Serie A64.9%37/57
Championship77.8%42/54
Premier League72.5%37/51
Ligue 168.9%31/45
Eredivisie83.8%31/37

Only competitions with 20 or more settled Banker picks are shown. Single-league samples are smaller than the full study, so treat them as indicative.

How the study was run

  • Sample: every one of the 851 Bawler picks that had settled between 18 March 2026 and 1 July 2026, across 46 competitions. Nothing was cherry-picked and no losing picks were removed.
  • Logged before kickoff: each pick is written to an append-only log with a content hash when official lineups drop (~75 minutes before kickoff), so it cannot be changed after the result is known.
  • Settled automatically: results come from live match data. We do not hand-grade, delete losers, or backdate winners.
  • The three tiers: every match gets a Banker (the safest call), a Value pick, and a Dark Horse (a deliberate long shot). Each is scored on its own outcome.
  • Confidence: each pick carries a probability from Bawler's model, which is built on expected goals for both teams. The calibration table compares that stated confidence to what actually happened.
  • Reproducible: the full, un-edited pick history is public at /track-record — recompute any figure above yourself.

Cite this report

Bawler, The 2026 Football Prediction Accuracy Report, published 1 July 2026. Based on 851 settled picks across 46 competitions (18 March 20261 July 2026). Key figure: the free Banker pick landed 73.8% (628/851). Source data: bawler.ai/track-record.

Journalists and researchers are welcome to reproduce these figures. For the underlying data or a league-specific cut, the public ledger is the source of truth.