SportBot AI tested: the free plan and the public record

How we tested

Plan: Free, no card added
Device: Desktop browser
Bet check: Real Madrid vs Inter Milan, Champions League, odds 1.60, stake $30
Match analysis: AEK Athens vs LASK (short card), Borussia Dortmund vs Villarreal (full analysis), a third match blocked by the daily limit
Pages read: performance, models, methodology, pricing, terms, responsible gambling
Not tested: Pro, Premium and lifetime features; the AI chat; the picks feed behind the paywall; the iOS and Android apps
Screenshots: 11, all from our own account

What SportBot AI is, in simple words

SportBot AI is a website with iOS and Android apps. It works with soccer, basketball (NBA and EuroLeague), hockey, tennis and baseball, and each of these five sports has its own AI model with a public record. The site also lists American football and esports, but at the time of checking there was no public record for them. A model here is a computer program that looks at past matches and gives every outcome a chance in percent.

A bookmaker is a company that takes bets. Odds are the number the bookmaker pays if you win: odds of 2.00 mean you get $2 back for every $1 you bet. Your stake is the money you put on one bet. Every bookmaker adds a margin, a small fee hidden inside the odds, so the prices are always a little worse than the true chances.

SportBot AI has two parts. The bet check takes a bet you have already chosen, removes the bookmaker margin, and shows the fair price. The AI picks are bets the models choose on their own. The AI picks have a public record: every settled pick is shown on the site, win or lose. The bet check gives a verdict on your own bet and keeps no public record.

Getting started takes three steps. Sign up with an email or a Google account; no card is asked. Open “Bet Check” if you already have a bet in mind, or “Analyze” if you want the model to look at a match. Then open “Performance” before you trust either: the whole record is public and free. Both features are explained step by step below.

The sign-up page shows a counter of settled picks and a feed of “verified plan purchases”, one Pro Monthly per day. That feed is advertising. You cannot check it.

How the SportBot AI bet check works, step by step

We tested the bet check on Real Madrid vs Inter Milan in the Champions League.

1. Choose the sport and the league. A list of matches for the next few days appears. If your match is missing, you can type it in by hand.
2. Choose the market and your pick. The market is the type of bet. We chose “Winner” and picked Real Madrid.
3. Type the odds your bookmaker gives you and your stake. We typed 1.60 and $30.
4. Press “Check my bet”.

The answer came back in a few seconds, in a red box: “You're overpaying vs fair value”. SportBot AI looked at 49 bookmakers and calculated that Real Madrid has a 59.5% chance to win. That chance equals fair odds of 1.68. Our bookmaker offered 1.60, which means a 62.5% chance. This percent hidden inside the odds is called the implied probability. So the bookmaker thinks Real Madrid is more likely to win than the whole market does, and pays us less. The gap is −4.8%. On a $30 stake, we would lose about $1.44 on average every time we take this price. The box also says “sharp-book anchored”. Sharp books are the bookmakers with the most accurate odds, and their prices count more in the average.

Two fields were locked: “Recommended stake” and “Best available price”. Both said “Unlock with PRO”. The stake is a Kelly stake: a formula that sets the bet size from your bankroll (the money you keep for betting) and the size of the edge. The bigger the edge, the bigger the stake, and with no edge the formula says zero. On the free plan you get the verdict and the fair odds. The size of the bet and the bookmaker with the best price are paid features. The homepage button says “Check my next bet — free”, and the FAQ says checks on your own bets are part of the paid plans. Both sentences are true at the same time.

We tried a second check on Porto vs Manchester City. The site answered “Daily check limit reached”. The free limit is one check per day.

What the free match analysis shows, step by step

The second free feature is the match analysis. It lives on the “Analyze” page, and it works differently from the bet check: here the model chooses the bet for you.

1. Open “Analyze” in the top menu. Matches are grouped by day, and every card shows the average market odds for home win, draw and away win. At the time of checking, the list showed every Champions League match of the week.
2. Press “View Insights” on a match. A short card opens for any match: the verdict, the best price among 32 bookmakers, a “Why” block and a “What could go wrong” block. This card does not count against the daily limit.
3. Scroll down to open the full analysis. A progress bar says “Building this analysis” and lists four steps: starting, gathering match data, running the models, preparing the answer. The site says it usually takes 15–30 seconds; ours was faster. This full analysis is the one-a-day free feature.
4. Read the tabs under the verdict: Price, Signals, Teams, Totals.

We opened two matches. AEK Athens vs LASK came first, as a short card. The verdict was a green banner: “Worth backing LASK — slight edge”. The line under it explains the idea in one sentence:

“We make AEK Athens the likelier winner, but the price is wrong on LASK.”

For a beginner this is the key sentence on the site. A value bet is a bet where the bookmaker pays more than the real chance deserves, even if the team is the less likely winner. LASK at 4.60 (the best of 32 bookmakers, at Betfair) was such a case, according to the model.

Then the limits of the free plan appeared. The two most important numbers, “Books' true price” and “Fair by our numbers”, were blurred. The over/under percentages in the summary block were blurred too. A purple button said “Unlock the exact numbers”. So on the free plan you get the verdict and the direction of the edge. You do not get the size of the edge.

The “Why” block was open. It listed three facts: AEK scores 2.2 goals a game and LASK 1.4, a form rating of 61 against 37, and the last five results. A “What could go wrong” block listed one missing AEK player and LASK's one win in five matches. This is plain and useful, and it is more than most prediction apps show for free.

The second match, Borussia Dortmund vs Villarreal, was our full analysis of the day. It got a purple banner:

“The price matches our numbers — nothing to play here. The gap is too small for us to call it an edge.”

The model agreed with the market, so it told us not to bet. We rarely see a prediction tool say this.

Price shows the odds of 32 bookmakers for the winner market, spread and over/under, with the best price marked: 1xBet at 1.81 for Dortmund, Betfair at 4.40 for the draw and for Villarreal. This part is fully open. The bet check uses a wider set of 49 bookmakers for its fair price; the analysis table lists the 32 with odds you can actually take. Signals shows form, tempo (about 1.8 goals a game) and squad availability (six players out, five of them at Dortmund), but the player names and the form bars are locked. Teams shows key players with most stats blurred.

The Totals tab gave us the strangest moment of the test. The predicted score was 2–1 and the expected goals total 3.4, so over 2.5 goals looked likely: 62% for over, 38% for under. Under that, a yellow note said:

“Score projection and totals model differ — our tracked totals pick is Under 2.5. The score projection is an illustration; the pick comes from the venue-adjusted totals model.”

So the page shows a score of three goals and a public pick of fewer than three goals, on the same screen. The site explains that the score is only an illustration and the real pick comes from a different model. That is honest. It is also confusing for a beginner, who will read the big numbers first and the small note last.

The next match stopped with “You've used today's free analyses. Upgrade for unlimited matches, or come back tomorrow.” One full analysis a day, exactly as the FAQ says. The short cards stay open for every match. This is enough to see the format and not enough to follow a whole matchday.

ROI: where the profit comes from

ROI means return on investment. It shows how much money the picks made for every $100 bet. A ROI of +5% means $105 back for every $100. The site counts every pick as one unit, a fixed bet size, and only counts picks where the model found an edge of 2% or more. An edge is the difference between the model's chance and the market's chance.

At the time of checking, the five model tiles on the performance page showed: soccer +8.3%, baseball +0.5%, basketball −2.5%, tennis −1.7%, hockey −3.9%.

The soccer card shows 716 bets, 53.8% accuracy and +59.6 units of profit. That is more than the whole profit of the service: the full record stands at +49.4 units, so the other four models lose part of what soccer earns. The profit chart shows it clearly: the green soccer line reaches about +85 units and ends near +60. The other four lines stay close to zero or below it.

The models page splits soccer into two parts, and this is the most useful table on the site. Totals are bets on how many goals will be scored (over or under a number). Moneyline is a bet on which team wins. Soccer totals: 363 wins, 270 losses, +12.9% ROI. Soccer moneyline: 22 wins, 61 losses, −26.9% ROI. Roughly, the totals model earned about +82 units and the winner model lost about −22 units. Together they give the +59.6 on the soccer card.

So “the soccer model is profitable” is only half true. The soccer totals model is profitable. The soccer winner model is one of the worst models on the board. A beginner who follows “soccer picks” without reading the table will follow both. One more detail for readers outside the US: at the time of checking, most of the last ten soccer picks came from MLS, the American league, with one Turkish match among them, so the soccer record is not built on the Champions League or the Premier League.

The same page marks two rows as “collecting”: baseball totals with 43 of 50 bets settled and basketball moneyline with 28 of 50. Both are winning, and both stay out of the headline until they reach 50 bets. This is a fair rule.

Closing line value: the number is negative

Under the chart there is a block called “Beating the closing line”. Closing line value, or CLV, compares the odds when the pick was made with the odds just before the match starts. If the odds get shorter after the pick, the market agreed with the model. If they get longer, the market disagreed. The site explains this well:

“Consistent +CLV is the gold standard proof of a genuine edge.”

Its own number is −0.6%. Only 43.8% of tracked picks beat the closing line. The soccer card shows −1.5%. The market moved against SportBot AI's picks more often than with them.

The CLV sample is 352 bets, about a quarter of the record, so most picks have no CLV data yet. A totals model can also make money for a while without positive CLV. Still, the site chose this test itself, and by this test it has not proved an edge. It gets credit for showing the number. Most competitors do not track CLV at all.

Five pages, five bet counts

The homepage: 206 qualified settled picks. The sign-up page and the performance header: 1430 bets, +2.2% ROI. The full public table: 1502 bets, 782 won, 720 lost, 52.1% accuracy, +3.3% ROI, +49.4 units. The CLV block: 352 tracked. The picks page: 569 settled top picks at +10.7% ROI and average odds 2.24.

Most gaps have a reason. The 72 bets between 1502 and 1430 are close to the two “collecting” models, so the headline is stricter than the table. The 569 top picks are a part of the whole. The 206 on the homepage has a filter called “model pipeline only” that the page does not explain. There is no sign of cheating in any of this, only confusion. The number to use is the full table: 1502 bets and +3.3% ROI. The profit is real. It is also small.

One outside check supports the trend. An earlier third-party review quoted 1022 tracked predictions and +7.4% combined ROI on the site. At the time of our check the record was 1502 and the ROI +3.3%. As the sample grows, the ROI shrinks. Betting records often go this way, and it is why a big sample matters more than a big percentage.

SportBot AI pricing: free, Pro, Premium and lifetime

Prices are on the pricing page. Free, permanent, no card: one bet check a day (verdict and fair odds only), one full match analysis a day with the fair price blurred, short verdict cards for every match, the odds comparison for 32 bookmakers, the performance, models and methodology pages, the picks page with the picks hidden, and the free calculators (an EV calculator, a hedge calculator and an odds converter, the same jobs our own betting calculators do).

Pro, $18.99 a month: ten match analyses a day, fifty questions a day to the AI chat, the recommended stake and the best price inside the bet check, a bankroll tracker, and the real picks on the picks page. The yearly plan is 39% cheaper.

Premium, $39.99 a month: no daily limits, unlimited AI chat, alerts when a price moves into value, and full history.

Lifetime Pro, $299 once: 250 analyses and 1,000 chats a month, a sure-bet scanner (the kind of tool we compare in the arbitrage scanners section) and a weekly “AI Coach” report. Lifetime Premium: $999 once.

The site gives no refunds on any plan and says so on the pricing page:

“Because the service is digital and delivered immediately, we do not offer refunds.”

$299 equals about sixteen months of Pro, so the lifetime plan only wins if you use the tool for two years or more and the company is still running. The methodology page says “Editorial review: June 2026”, and the footer carries Product Hunt and BetaList badges. This is a young product. $999 for a lifetime of a service this new is a bet in itself.

Payment goes through Stripe (Visa, Mastercard, Amex) and PayPal. You cancel in account settings; the site offers a discount on the way out, but there is no fee and no contract. The terms of service and the responsible gambling page repeat that nothing is guaranteed and the service is for users aged 18 and over.

What the AI models can and cannot do

On the free plan the bet check does not use the AI model at all. It uses the bookmakers. When 49 bookmakers agree that Real Madrid has a 59.5% chance, that agreement is the fair price. A single bookmaker paying less is a worse deal. Serious bettors have done this by hand for years. SportBot AI does it in a few seconds. This is the strongest part of the product.

The AI picks are a different thing. The model estimates a chance, compares it with the fair price, and calls the gap an edge. The average edge on the soccer card is 15.1%, and on the basketball card it is 26.4%. Real edges over the market are usually a few percent. Numbers this big mean the model disagrees with the market a lot. The negative CLV says the market has been right more often. The picks are one opinion with a public record; the bet check is the reliable half of the product.

What holds up at SportBot AI, and what does not

What works. The bet check answers a real question in seconds and asks for no card. The match analysis explains its verdict in plain sentences and is willing to say “nothing to play here”. The public record shows the odds, the edge, the price move and the result for every pick, losses included. Losing models are not hidden. The methodology page uses simple words. The price comparison includes 1xBet, Betfair and Matchbook, so the tool is useful outside the UK and the US (our odds comparison section covers dedicated tools for this).

What raises questions. The first screen leads with soccer +8.3% and does not say that the totals model makes it and the winner model loses. CLV is negative by the site's own rule. Five different record counts make the honest number hard to find. The free plan is one check and one full analysis a day, with the fair price blurred on both. The Totals tab can show a 2–1 score next to an under 2.5 pick. There are no refunds. The “verified purchases” feed and the “your first win in 2 minutes” heading promise more than the legal text does.

Our 3.8 out of 5 comes from four parts. The bet check earns close to full marks: it answers a real question, uses 49 bookmakers and costs nothing. The public record earns high marks for openness and loses some for the five different counts. The AI picks score low: one profitable model, four losing or flat ones, and negative CLV. The free plan and pricing land in the middle: a permanent free tier is rare in this category, but the blurred numbers, the no-refund rule and the $999 lifetime price pull the score down.

One free check a day: enough for a weekend bettor, not for a matchday

The tool fits people who already choose their own bets and want to know if the price is fair before they place them. One free check a day is enough for a weekend bettor with one or two bets. It fits soccer totals bettors who are ready to pay for Pro and follow a model with a public record, because the picks themselves are hidden on the free plan. It does not fit anyone who wants a ready list of winners, because the winner models lose money. And it does not fit anyone who expects the AI to beat the bookmakers on its own.

Alternatives: ScoreGPT, FootballPredictAI and Rithmm

ScoreGPT works the other way round: five named language models predict the match and a public leaderboard grades each one. It covers fewer leagues and has no fair-price check. FootballPredictAI is cheaper to start with and covers similar soccer leagues with one statistical system and no public record. Rithmm gives no ready picks and lets users build their own models, mostly for US sports. SportBot AI has a model builder of its own on the models page. Building and backtesting is free; saving a model, which gives it a live record and a daily feed of picks, needs Pro. Three ready-made “house” models are listed under the production ones: one for totals across all sports stands at 34 wins, 22 losses and +11.6%, and two soccer totals models have only six and seven settled picks each. A live ROI only appears after 30 settled picks, and until then a model is ranked by a simulation. The full comparison is in the AI prediction tools guide.