Past results

Every AI pick in these five weeks, by the day it kicked off. Open a day to see its picks and how each one settled.

1,200 AI picks

730 Won

460 Lost

61% Hit rate

2010–5 2118–12 2224–9 2319–12 2414–6 2527–18 2626–17

138–79

279–12 2819–13 2928–11 308–12 3124–13 132–33 239–16

159–110

318–9 417–12 515–11 69–11 717–14 846–26 941–23

163–106

106–7 1114–10 129–20 1322–8 1421–12 1540–15 1633–25

145–97

177–2 183–7 199–9 207–3 2120–8 2243–27 2336–12

125–68

All models and sports. A green day won more picks than it lost, a red day lost more, and the colour deepens as the day's hit rate moves away from even. Today stays blue until its picks settle. Voids and pushes count as neither. Days follow UTC kickoff time.

Frequently asked questions

How accurate have SignalOdds predictions actually been?
This page is the raw record: every day's picks with how many won and lost, and each day opens to its picks and the result each was graded against. Accuracy varies a lot by sport and market, so a single site-wide number would be misleading. Each model publishes its own verified win rate, ROI and accuracy by confidence level on its profile page at /models.
How is a prediction graded as a win or a loss?
Predictions are graded automatically once an event settles, against the final score and the market the pick was made in — moneyline, spread or totals. Grading runs from the same result feed shown on the event page, so the outcome you see next to a pick is the one it was scored on.
Are past results ever edited after the fact?
No. A prediction's odds, confidence and selection are fixed when it is published, before the event starts, and the record here shows exactly what was published. Scores are reconciled against multiple sources, so a result can be corrected if a data provider reported it wrong — but the pick itself is never changed.
Does a good past record predict future results?
Not on its own. A track record over a small number of picks tells you very little, and every model goes through losing runs. Look at sample size and how a model performs in the specific sport and market you bet, rather than at a headline percentage — and treat all of it as information, not a guarantee.