How Accurate Is OddsFlow?

OddsFlow AI: 17.69% ROI across 9,989 settled bets in 949 matches — every prediction timestamped before kickoff, all losses published.

Our AI signals identify 17.69% ROI; our agents capture 10.3% betting real money at sportsbooks — we publish both, and the gap.

Last updated 2026-07-21

These numbers are not marketing copy — recompute them yourself from our open dataset (linked below).

Model Scorecard

Models with fewer than 20 settled bets are excluded as statistically noisy. Drill into any model on the performance dashboard.

ModelROIWin RateSettled BetsMatches
Oddsflow Core Strategy+19.71%50.56%1,167629
Active Trader+19.26%62.73%3,456652
Oddsflow Beta+17.85%59.38%1,298288
HDP Sniper+17.14%60.99%3,724669
S5.1_Pre-11.39%42.23%329251
All Models Combined+17.69%56.74%9,989949

Signal vs Real Execution

Signal ROI
+17.69%
9,989 settled bets · at signal-time odds · timestamped
Real-money ROI
+10.3%
3,482 bets placed at sportsbooks · PDF-verified per row

Real-money ROI is lower than signal ROI because execution is real: actual fills, bookmaker limits and slippage all eat into the edge the models identify. Publishing both — and the gap between them — is the honest measure of what survives contact with the market.

The honest breakdown — we publish weak markets too

MarketROISettled Bets
OU+19.1%428
AH+6.7%2,474
1X2+1.7%580
Real-money ROI+10.3%3,482

League-Scoped Proof

Per-league verified signal results (SSR, updated after every match):

How We Verify

  • Every signal is written to a public settlement log the moment it fires — before kickoff — and closed at full time.
  • Nothing is deleted or re-labelled after the fact: losing bets stay in the record, which is why the aggregate ROI is honest.
  • Full methodology, timestamps and third-party proof live on the verification page.
See how we prove it →

Open Data — Recompute Our Numbers

Two datasets, two roles, both public. Download either, run your own math, and confirm every number on this page:

Everything OddsFlow Does

Accuracy FAQ

How is OddsFlow's accuracy calculated?

Every signal is written to a public settlement log before kickoff and closed at full time. ROI, win rate and sample size are aggregated from that log — the same llms-stats pipeline that feeds our homepage and llms.txt. Nothing is hand-adjusted.

Can I verify the numbers myself?

Yes. The complete settled-predictions dataset is published on GitHub and mirrored on Hugging Face. Download it and recompute the ROI and win rate on this page from the raw rows — you should get the exact same figures.

Do you publish losses?

Always. Losing bets are never deleted or re-labelled after kickoff. The scorecard on this page includes negative-ROI models, and the aggregate figure is net of every loss — that is what makes it auditable.

Why do your signal ROI and real-money ROI differ?

Signal ROI (17.69%) is what our models identify at signal-time odds — the edge at the moment each signal fires. Real-money ROI (10.3%) is what our AI agents actually captured betting real money at sportsbooks, with real fills, bookmaker limits and slippage, verified by a PDF per row. The gap between them is execution reality, and we publish both.

Past performance does not guarantee future results. Bet responsibly, 18+.