SportSignals Track Record: Prediction Accuracy, Tips, Bet Builders and Accas
Every record SportSignals publishes, in one place: how often the model calls fixtures correctly, and what happened when those calls were staked. Refreshed as fixtures settle. Past performance does not guarantee future results.
Settled data through
Coverage right now: quoting 21 bookmakers across 33 active leagues. These counts are computed from the live feed configuration, not typed by hand.
In short
SportSignals model predictions have been correct 66.2% of the time across 29,922 settled selections, every one of them recorded before kick-off. Staked at level stakes, the published tips returned +15.2% over 740 settled tips. Bet builders returned +15.2% over 1,140. Accumulators are the loss-making product: 1 of 3 slates is in profit, and the retired long shots slate finished 0W-50L at -£500. Each slate is published separately below rather than blended into one number. Prediction accuracy and the staking records rest on different evidence, which the table states row by row. Past performance does not guarantee future results.
Every published record at a glance
| Record | Evidence | Settled | W-L | Strike rate | Profit | Return on stakes |
|---|---|---|---|---|---|---|
| Model prediction accuracy | Audited pre-kickoff snapshot | 29,922 | 19,797 correct | 66.2% | Not a staking record | |
| Football tips | Published ledger | 740 | 348W-392L | 47.0% | +£1,126 | +15.2% |
| Bet builders | Published ledger | 1,140 | 208W-932L | 18.2% | +£1,732 | +15.2% |
| Accas: Trebles | Published ledger | 57 | 12W-45L | 21.1% | -£21 | -3.6% |
| Accas: Weekend 5-folds | Published ledger | 6 | 3W-3L | 50.0% | +£168 | +280.3% |
| Accas: Long shotsretired | Published ledger | 50 | 0W-50L | 0.0% | -£500 | -100.0% |
Two kinds of evidence on this page
- Audited pre-kickoff snapshot
- Prediction accuracy is aggregated from an audited snapshot of settled selections, and only selections recorded before kick-off are counted. It is a hit rate. It says nothing about stakes or returns.
- Published ledger
- Tips, bet builders and accas are computed from the published ledger at the moment this page renders. Every slip was published before kick-off, but these figures are not part of the audited snapshot above and should not be read as though they were.
Overall
66.2%
19,797 correct of 29,922 settled predictions across all markets and competitions.
- Market scope:
- all model selections recorded before kickoff and settled
- Competition scope:
- all competitions with verified settled selections
- Numerator / denominator:
- 19,797 / 29,922
- Data through:
- Period covered:
- to
Correct means the selection settled as won on the 90-minute result. Every selection counted here was recorded before kickoff.
Accuracy by market
Each row is a separate settled market cohort across all competitions with recorded selections; samples are not directly comparable where their market outcomes differ.
| Market | Accuracy | Correct | Sample | Data through |
|---|---|---|---|---|
| Away Over/Under 0.5 | 67.3% | 1,210 | 1,799 | |
| Away Over/Under 1.5 | 31.8% | 572 | 1,799 | |
| Away Over/Under 2.5 | 87.9% | 1,582 | 1,799 | |
| Both Teams to Score | 54.7% | 984 | 1,799 | |
| Double Chance | 74.9% | 1,275 | 1,703 | |
| Draw No Bet | 63.1% | 839 | 1,330 | |
| Full Time Result | 50.3% | 905 | 1,799 | |
| Half Time Result | 50.7% | 863 | 1,703 | |
| Home Over/Under 0.5 | 24.8% | 446 | 1,799 | |
| Home Over/Under 1.5 | 62.6% | 1,126 | 1,799 | |
| Home Over/Under 2.5 | 76.5% | 1,376 | 1,799 | |
| Over/Under 1.5 | 73.3% | 1,318 | 1,799 | |
| Over/Under 2.5 | 54.4% | 978 | 1,799 | |
| Over/Under 3.5 | 72.7% | 1,308 | 1,799 | |
| Over/Under 4.5 | 86.4% | 1,554 | 1,799 | |
| Over/Under 5.5 | 94.4% | 1,698 | 1,799 | |
| Over/Under 6.5 | 98.0% | 1,763 | 1,799 |
Read accuracy alongside sample size: a small sample can swing wildly. Forecasting Both Teams to Score is not the same task as predicting an exact correct score, so naive cross-market comparisons are misleading.
Football tips: published staking record
Single selections, published before kick-off and settled in public. Every losing tip stays in the record.
+15.2%
Return on stakes
+£1,126
At £10 level stakes
348W-392L
47% strike rate
740
Settled selections
Wilson 95% confidence interval 43.5% to 50.6% strike rate. Covering to . Computed from the published ledger at render time. Every slip was published before kick-off. This is not part of the audited accuracy snapshot above. See every football tips.
Bet builders: published staking record
Same-game combinations at long combined prices. Builders win roughly one time in five, so the strike rate and the return have to be read together.
+15.2%
Return on stakes
+£1,732
At £10 level stakes
208W-932L
18% strike rate
1,140
Settled selections
Wilson 95% confidence interval 16.1% to 20.6% strike rate. Average price 6.80. Covering to . Computed from the published ledger at render time. Every slip was published before kick-off. This is not part of the audited accuracy snapshot above. See every bet builders.
Accumulators: published record by acca type
Each slate is shown at its own size. There is no combined acca figure, because one number would fold a retired experiment into the live slates and flatter both.
| Acca type | Settled | W-L | Strike rate | 95% interval | Avg odds | Profit (£10 level) | Return on stakes |
|---|---|---|---|---|---|---|---|
| Trebles | 57 | 12W-45L | 21.1% | 12% to 33% | 6.27 | -£21 | -3.6% |
| Weekend 5-folds | 6 | 3W-3L | 50.0% | 19% to 81% | 15.40 | +£168 | +280.3% |
- Trebles: Near break-even across 57 slips at 6.3 average combined odds.
- Weekend 5-folds: Small sample, so treat the strike rate as indicative rather than settled.
Retired slates
| Acca type | Settled | W-L | Strike rate | 95% interval | Avg odds | Profit (£10 level) | Return on stakes |
|---|---|---|---|---|---|---|---|
| Long shots | 50 | 0W-50L | 0.0% | 0% to 7% | 82.18 | -£500 | -100.0% |
- Long shots: Retired. A daily long shot was entertainment posing as analysis, and the record said stop. High-variance accas now publish only when the model prices them as value.
Computed from the published ledger at render time. Every slip was published before kick-off. This is not part of the audited accuracy snapshot above. See every acca.
How are these numbers produced?
Accuracy on this page is computed from settled predictions stored in the SportSignals database. Each row counts a prediction as "correct" when the published outcome matches the result on the pitch. Pending and voided fixtures are excluded. Read the full methodology for the data inputs and model architecture that generate these forecasts.
Staking records are computed from the published ledger when this page renders. Return on stakes is profit divided by the amount staked, assuming the same flat stake on every settled selection, with voided selections excluded from both terms.
Correction, 30 July 2026: the bet builder return was previously published as +16.9% across 996 settled builders. The query behind it was capped at 1,000 rows, so it was computed from part of the record rather than all of it. The corrected figure covers every settled builder and is shown above.
Track record questions
- How accurate are SportSignals predictions? +
- Model selections have settled correct 66.2% of the time across 29,922 settled selections (19,797 correct). Only selections recorded before kick-off are counted, so the figure cannot be improved after the result is known. Accuracy is a hit rate and says nothing on its own about whether staking those selections would have returned a profit.
- Do the football tips make a profit? +
- Across 740 settled tips the published record is +15.2% return on stakes at level stakes, from a 348W-392L record. That is a historical record of published selections, not a forecast. Past performance does not guarantee future results.
- What is the bet builder record? +
- 1,140 settled builders at +15.2% return on stakes, from a 208W-932L record. Builders win roughly one time in five at long combined prices, so the strike rate looks alarming beside a positive return. Both numbers are shown together for that reason.
- Do the accumulators make money? +
- Mostly no, and the record is published per acca type rather than blended. 1 of 3 slates are in profit: Trebles 12W-45L at -3.6%; Weekend 5-folds 3W-3L at +280.3%; Long shots 0W-50L at -100.0%, now retired. A single combined acca figure would fold the retired slate into the live ones and flatter both.
- Why are the long shots at 0 from 50? +
- Because publishing a long-odds slip every day was a volume habit rather than analysis, and the record said stop. The slate finished 0W-50L. It stays on this page in full rather than being deleted, because removing a losing product from a track record is exactly what a track record exists to rule out.
- What counts as settled? +
- A selection is settled once it has reached a win or a loss. Voided selections are excluded from both the record and the amount staked, because a bet that never ran should neither count as a loss nor pad the sample.
- How is return on stakes calculated? +
- Profit divided by the amount staked, assuming the same flat stake on every settled selection, expressed as a percentage. It is not compounded and it assumes no staking plan. The figure is computed from the published ledger when this page renders, so it moves as results settle.
- Is past performance a guide to future results? +
- No. Every figure on this page is a record of what has already settled. Samples of this size still carry meaningful variance, which is why strike rates are published with a 95% confidence interval beside them. Tips are informational analysis, not financial advice.
18+. Predictions are informational only and do not constitute betting advice. Past performance does not guarantee future results. Visit GambleAware for free, confidential support.