- 461 resolved picks with a 52.2% hit rate.
- Yield −1.0% on the units staked.
- One month is a small sample: check the all-time figures before drawing conclusions.
October 2026 summary: the numbers
In October 2026 the AI published 461 picks.
461 were resolved: 238 wins, 218 losses and 5 voids.
The hit rate was 52.2%.
The yield was −1.0%, calculated on the units staked at the suggested stake.
| Metric | October 2026 | Previous month | Cumulative at month end |
|---|---|---|---|
| Resolved picks | 461 | 294 | 1644 |
| Hit rate | 52.2% | 50.7% | 55.7% |
| Yield | −1.0% | +2.7% | +0.6% |
| Average odds | 1.87 | 2.05 | 1.84 |
| Average implied probability | 54.0% | 50.9% | 56.2% |
The average implied probability of the odds was 54.0%.
The hit rate was 52.2%, a gap of −1.8 points.
The margin of chance with 456 picks is ±4.6 points.
The hit rate stayed within the margin of chance of what the odds implied.
Results by sport in October 2026
| Sport | Picks | Wins | Hit rate | Yield |
|---|---|---|---|---|
| Football | 210 | 105 | 50.0% | −6.6% |
| Basketball | 113 | 60 | 53.1% | +1.4% |
| Hockey | 59 | 34 | 57.6% | +13.2% |
| American football | 40 | 20 | 50.0% | −6.5% |
| Tennis | 23 | 14 | 60.9% | +23.2% |
| Baseball | 11 | 5 | 45.5% | −14.6% |
The sport with the best yield was Tennis, with +23.2% over 23 picks.
The one with the worst yield was Football, with −6.6% over 210 picks.
Only sports with at least 15 resolved picks are compared.
Results by confidence level
| Confidence | Picks | Hit rate | Yield |
|---|---|---|---|
| Maximum | 2 | 50.0% | −23.0% |
| High | 31 | 58.1% | +10.7% |
| Medium-High | 358 | 53.1% | −1.6% |
| Medium | 65 | 44.6% | −17.2% |
Most picks had Medium-High confidence: 358 of 456.
Each pick's confidence is set when it is born and never changed afterwards.
Do not read the table as a promise: it measures what happened, not what will happen.
How to read these numbers
There are three numbers worth reading together.
The first is the hit rate, the percentage of picks won.
The second is the implied probability of the odds, which sets the bar.
The third is the yield, which says whether the result made or lost money.
- High hit rate with negative yield: it happens when low-odds picks win and the big ones lose.
- Low hit rate with positive yield: it happens when high odds hit, and it is normal in value markets.
- Hit rate against implied probability: if the hit rate beats the odds, the model beat the price.
- Margin of chance: the fewer the picks, the wider the band where the result may be luck.
Why hit rate alone is not enough: the break-even point
The hit rate you need to avoid losing depends on the odds.
It is calculated by dividing 1 by the decimal odds.
The lower the odds, the higher the hit rate you need.
The table shows some examples.
| Odds | Hit rate needed | How to read it |
|---|---|---|
| 1.40 | 71.4% | You must win a lot |
| 1.60 | 62.5% | You must win a lot |
| 1.90 | 52.6% | Close to a coin flip |
| 2.40 | 41.7% | You can lose more often |
| 3.00 | 33.3% | You can lose more often |
In October 2026 the average odds were 1.87.
That equals a break-even point of 53.3%, a simple reference, not an exact one.
The month's hit rate was 52.2%.
Yield is the measure that joins both in a single number.
Why one month says little
With 456 decided picks, the margin of chance is ±4.6 points.
A bad month does not prove the AI fails.
Nor does it prove the opposite if next month changes.
That is why we show the cumulative figures next to it: a large sample weighs more than any single month.
The result also depends on the price.
A pick at odds of 1.40 needs to win about 71% to break even.
One at odds of 2.40 needs to win a little over 42%.
That is why hit rate alone is never enough.
How to verify it yourself
Every pick stays in the record with its odds and result.
There is also an open data file with all the picks.
With that file you can recalculate the hit rate and yield of any month.
Each pick counts in the month of its match date, the fecha column of the file.
The cumulative figure in this report closes on the last day of the month, while the live record keeps adding picks.
See the public track record, the open data to download and our guide to AI for sports betting.
To judge any tool, read the 7 questions to judge a betting AI.
These numbers measure what happened, not what will happen.
Bet only what you can afford to lose.
If betting stops being fun, take a break.
Frequently asked questions
How many picks did the Gambeta AI resolve in October 2026?
It resolved 461 picks.
That was 238 wins, 218 losses and 5 voids.
0 were still unresolved when this report was written.
What was the hit rate and yield in October 2026?
The hit rate was 52.2% and the yield −1.0%.
Yield is calculated on the units staked at the suggested stake.
A single month is a small sample.
How is yield calculated?
It is the total profit divided by the amount staked, as a percentage.
A won pick adds its profit and a lost one subtracts the stake.
Voided picks count the stake but add no profit.
Why do you publish the bad months too?
Because a record that shows only the good months is useless for deciding.
Every pick is published before the match and resolved with outside data.
That is what makes the numbers verifiable.
Is one month enough to know if an AI works?
No.
With a few hundred picks, chance moves the result a lot.
You need several months and a large sample to see a trend.
Where can I verify these numbers?
In Gambeta's public track record, with every pick and its result.
There is also an open data file you can download.
You can recalculate the hit rate and yield yourself.