League insights · Top Prediction

AI performance snapshot

Flat-stake closing-line recap on finished Top selections. Transparent rolling window.
2026-05-012026-05-07 UTC

Only finished Top Prediction rows whose stored snapshot falls within Admin → Homepage Recommendations defaults: max_probability 60%–100% and confidence_score 0.6–2 (same columns as the admin SQL filters).

For research only — not betting advice. Small samples may look noisy; scores prefer leagues with adequate Top rows.

Window 7d 14d 30d
62.5%
1×2 hit rate
7d · finished Top
+51.25%
O/U 2.5 ROI
best closing line
+28.25%
BTTS ROI
best closing line
8
Tracked picks
vs reference closing pool
Full market breakdown table
Market Hit rate Best closing-line ROI
1×2 62.5% · (5/8) 14.63%
O/U 2.5 75.0% · (6✓/2✗) 51.25%
BTTS 75.0% · (6✓/2✗) 28.25%

Tracked Top picks: 8

AI market notes

One-line takeaways for this rolling window.

  • 7d recap on finished Top selections; ROI uses pooled closing prints where fills exist.

Cumulative equity curve

Cumulative flat-stake units after each settled pick in window order. Zero line marks breakeven on running P/L.

OddsGPT Decision Engine is continuously recalibrated using rolling match outcomes, pricing signals and market-closing data. League-level edge rankings and market diagnostics may adjust over time as new fixtures settle.

Methodology (short)

Flat staking (1 unit per pick); ROI uses the best available closing line for our tracked side where present, compared with a reference bookmaker for diagnostics.

Rolling windows (7 / 14 / 30 days) keep the view close to “current edge”. League cards are grouped from finished Top Prediction rows only.

ROI is hidden when there are fewer than 5 settled price observations for that market slice, to avoid noisy single-bet percentages.

OddsGPT Decision Engine is continuously recalibrated using rolling match outcomes, pricing signals and market-closing data. League-level edge rankings and market diagnostics may adjust over time as new fixtures settle.

Methodology (short)

Flat staking (1 unit per pick); ROI uses the best available closing line for our tracked side where present, compared with a reference bookmaker for diagnostics.

Rolling windows (7 / 14 / 30 days) keep the view close to “current edge”. League cards are grouped from finished Top Prediction rows only.

ROI is hidden when there are fewer than 5 settled price observations for that market slice, to avoid noisy single-bet percentages.