Frequently asked questions

You open Telegram (@DeepTennisAI_bot), choose "Generate bet" and can check Premium Picks, pick a specific match or review Deep Signals. Every pick shows market, reference odds, stake in units and the analysis explanation.

Yes. FREE is permanent, requires no card and includes three daily accesses: one real Premium Pick with a full analysis, one extra open analysis in summarized format, and one full Deep Signal.

FREE (€0): one full normal Premium Pick a day, one summarized open analysis, one full Deep Signal a day and public results. PREMIUM passes are one-time payments of €1 + VAT for 24 hours or €6 + VAT for 7 days. The monthly subscription grants 31 days for €15.70 + VAT. All three paid plans unlock normal Premium, Premium Select — the AI's highest-confidence selections — and every Deep Signal for matches whose start falls within the window. We trust our model so much that if the joint official balance of normal Premium and Premium Select ends strictly negative, we extend access for free by another 24 hours, 7 days or 31 days; extensions are covered again too. Deep Signals aren't part of that calculation.

By 12:00 Madrid time at the latest, but it moves up automatically to 90 minutes before the match when it starts early. If the analysis finishes after that time, it's published as soon as it's ready. It's never published once the match has started, and we never fabricate a replacement pick.

No. FREE already lets you check the quality, punctuality and transparency of the service every day with no card. Premium is available from €1 + VAT for 24 hours, with no automatic renewal.

Everything happens inside the official Telegram bot: @DeepTennisAI_bot. The website is the public showcase for results and methodology; live picks are served in the bot.

The pipeline runs automatically every day before play starts: it detects ATP and Challenger matches with enough data and odds, analyses them and leaves the picks ready in the bot. No picks are generated by hand or added mid-day.

No. If you've already seen a pick, opening it again doesn't use up another one of your daily uses.

Every pick comes from the same analysis, but Premium requires two extra filters: clearing the Q quality score threshold (a deterministic backend formula the AI cannot inflate) and belonging to markets that have historically proven their performance. A pick can look good and still be left out of Premium because its market is vetoed by the data.

It's the internal quality score for each pick. It combines the detected value, the edge against the market price, the quality of the available data for both players, and market stability. It's calculated in the backend after the AI's analysis, precisely so the model can't "score itself high."

It's the recommended exposure, from 1 to 5 units, assigned by a ladder based on pick quality, with per-market limits and cuts for risk signals. A unit is a fixed percentage of your bankroll that you decide (usually 1%). There's also stake 0: when quality control rejects the AI's proposal, that pick doesn't count as a bet.

Eight markets per match: match winner, set 1 and set 2 winner, total games, set 1 total games, games handicap, set handicap and correct score. Odds are compared across several bookmakers to set a realistic reference.

Because it didn't meet the minimums: without enough odds in the key markets or reliable data on one of the players, the match isn't analysed. We'd rather stay silent than give an opinion with no basis.

No. Best-of-5 draws follow a specific format route: the math for comebacks, set handicaps and totals changes completely compared with BO3. It has its own quality filters, vetoed markets and stake limits. You can see the tournament-by-tournament track record on Grand Slams.

Three public systems built on the match-winner market: Winner Elite (winners with a profile especially favourable to the model), Tapado IA (less obvious winners where the odds leave margin) and Combi Elite (the daily double that combines the 2 safest favourites of the day). Their full track record, losses included, is on Deep Signals.

Yes, always. Every pick closes automatically with the match's real result — won, lost or void — and the public performance includes everything: units, ROI, good streaks and bad ones. On Grand Slams you can even see a whole tournament finish negative, with its context.

Picks are frozen once published, and the system settles them using the real score. When we change an internal rule (a threshold, a filter), it applies from an effective date going forward: the already-published track record is never rewritten. We also log odds movement up to close to kick-off (CLV) — consistently beating the closing price is the hardest signal to fake.

ATP (250, 500, Masters 1000, Grand Slams, Finals) and Challengers, whenever the match has enough data and markets. Coverage follows the calendar: clay, grass and hard, with special pages for the Grand Slams.

No, and be wary of anyone who does. This is a statistical system with real variance: there are bad weeks and we publish them too. The goal is a sustained edge over time with risk management, not guaranteed wins.

Use fixed units (1 unit = ~1% of your bankroll), respect the stake shown on each pick, don't chase losses by raising amounts, and never bet money you can't afford to lose. This content is for over-18s only and is for informational purposes. Gamble responsibly.