Tip of the day Predictions Bet Builder Tournaments Track record How it works

Every result, counted.

Our prediction record

How often our AI tennis predictions are right, month by month and day by day, for the tip of the day and for all matches. Every figure here comes from the same daily record, which you can download in full.

Daily record, 17 Nov 2025 to 28 Sept 2026
67%
Overall hit rate
81%
Tip of the day hit rate
37,804
Predictions graded
315
Days on the record

Daily hit rate

Last 26 weeks · one square per day

Every month, in numbers

All predictions vs tip of the day

BananaPoint.ai hit rate by month, November 2025 to September 2026.

Sep 2026 67% 80%
Aug 2026 67% 78%
Jul 2026 67% 81%
Jun 2026 67% 78%
May 2026 67% 72%
Apr 2026 67% 88%
Mar 2026 67% 87%
Feb 2026 68% 81%
Jan 2026 70% 89%
Dec 2025 67% 100%
Nov 2025 65% 86%

The last two weeks

Every graded day, in full
Day Graded Hit rate Bankers
Mon, 28 Sept 116 65% 78%
Sun, 27 Sept 95 66% 100%
Sat, 26 Sept 79 71% 100%
Fri, 25 Sept 122 57% 71%
Thu, 24 Sept 212 68% 100%
Wed, 23 Sept 256 63% 75%
Tue, 22 Sept 202 69% 67%
Mon, 21 Sept 90 67% 78%
Sun, 20 Sept 86 64% 63%
Sat, 19 Sept 66 62% 75%
Fri, 18 Sept 109 65% 100%
Thu, 17 Sept 167 67% 80%
Wed, 16 Sept 202 72% 100%
Tue, 15 Sept 171 66% 67%

Common questions

What am I looking at?

+

Every prediction BananaPoint has published, compared with the final result. You get two hit rates, one for the tip of the day and one for everything, and all the charts and files on this page come from one archive, so no result can be picked afterwards.

Why are there two hit rates?

+

Because they measure different things. The tip of the day works as a short daily list of the model's surest calls: 2,289 graded so far, and 80.6% of them won. The wider figure takes in all 37,804 predictions we have ever published, 67.2% of them correct. A small, demanding list is bound to score higher, so we never quote one number without the other.

How do you calculate the percentage?

+

It is a plain ratio: right calls over graded calls. A call is graded only when its match produces a final score, so walkovers and other matches that end without one are simply not counted, neither as wins nor as losses. Days are not all equal either: the more matches a day settles, the more it moves the overall percentages.

Why does a single day look so different from the month?

+

Because a single day rests on a handful of picks: with four of them, each result moves the day's rate by a quarter, all the way from nothing to a perfect score. Monthly and all-time figures smooth that out, and the more picks behind a number, the more it says.

Are some tournaments left out?

+

None are. If the model made a prediction, it counts, whether the match was a Grand Slam or an ITF qualifier and whatever the court. Leaving out a bad month or a weak tour would make the figures prettier and less honest, so we don't.

Since when do you keep the record?

+

Our first graded predictions date from November 2025. The charts and the CSV draw on a daily log that now spans 315 days, 17 Nov 2025 to 28 Sept 2026. Scroll the table for the latest 11 months, or download the file for every single day.

When is a new day added?

+

Once a day. A date appears only after its last match is graded, so 28 Sept 2026 is the newest complete day you will see here, and every chart, table and file is regenerated at that moment.

What is in the two downloads?

+

Open the CSV for the day-by-day totals, tip of the day and overall side by side. The Excel workbook is the full match log, split into yearly sheets, with everything the board showed for each match, from both prices to our call.

Can I check the numbers without trusting you?

+

Yes. Download both files and recount, or open any match page: the prediction stays up after the result, whether it won or lost.

That's the record. Now the picks.

Full Access opens every prediction on the board, so you can judge the next result for yourself.

See every prediction

18+ · Forecasts made by AI, for information only · BananaPoint neither takes nor places bets How our predictions work