NerdyTips review 2026: 66.9% hit rate, −4.2% return, is it worth it

How we tested

• Plan: Free, no card added
• Device: Desktop browser
• Free features used: 7 daily tips, match page, team page, Bet of the Day, Bet Builder, Hit&Win, Progress
• Data check: allMatches.xlsx from the Progress page (sheets 2024, 2025, 2026; 170,689 rows with best tip, confidence and odds); every best tip settled against the final score with our own script
• Not tested: paid plans; the locked predictions; the mobile apps
• Data file downloaded: early September 2026, 170,689 rows; we plan to re-run the check each quarter
• Screenshots: 7 from our own account and Trustpilot + 1 chart built from the site's data file.

What NerdyTips is

NerdyTips is a football prediction website from Romania, with apps for iOS and Android. It started in May 2021 and says it has predicted more than 273,000 matches. One computer model, called NT Apex, makes a tip for every match. A tip is the result the model expects. For each match there is a tip for the winner (1 for home, 2 for away, X for a draw, or a double chance like 1X, which means home win or draw), a tip for the number of goals (over or under a line, for example over 2.5 means three goals or more), a tip on whether both teams will score, and a predicted score. One of these tips is marked Best Tip and gets a confidence score from 0 to 10.

Three more words you will need. A bookmaker is a company that takes bets. Odds are what the bookmaker pays if the tip wins: at odds of 1.46 you get $1.46 back for every $1 you bet, so $0.46 profit. Odds also hide a percentage. Odds of 1.46 mean that you must win about 68.5 tips out of 100 just to get your money back. This is called the break-even rate. Keep 68.5% in mind. Every number below is compared with it.

NerdyTips does not take bets and does not compare bookmakers. It shows one set of average odds per match. There is no tool to check your own bet and no closing line value, which are the two things SportBot AI is built around. NerdyTips is a list of tips with statistics behind each one.

The 7 free tips, and what else the free plan opens

Sign-up asks for a name, an email, a username and a password, plus a box to confirm you are over 18. No card. The welcome page then shows a short list of recent predictions marked won or lost, an 83% prediction accuracy line for that short list, and the price from $8.99/mo.

The free tips are on the homepage under Today's Free Football Predictions. The site promises seven every day, and there is no trial period on the free account. On our test day the seven were: a Finnish third-division match, Malmo vs AIK, Getafe vs Celta Vigo, a Spanish women's match, Palermo vs Sampdoria, an Italian Serie C match and a Paraguayan derby. Each row shows the odds for home win, draw and away win, a tip for each market, the best tip and its confidence score. All seven were rated above 5 out of 10 that day. How the site picks these seven from the other high-confidence tips is not explained.

Every free match opens a full page. For Ilves II vs P-Iirot the best tip was over 3.5 goals at odds 1.73 with a confidence of 8.0. The winner tip was 1X at 1.19 with a confidence of only 2.0. Both teams to score was yes at 1.36. The Bet Builder tip was 1X plus over 2.5 at 1.54 with a confidence of 1.5. The model also predicted a 2–2 final score and 1–1 at half-time. Under all this is a block of predicted match statistics: possession, shots, corners, cards. A warning line said the home team did not play better in the last meeting between the two.

Everything else is locked. All Matches listed 132 games on our day. Seven were open. The rest were blurred. The Bet of the Day page showed three bankers with every tip hidden, under a banner that reads:

Predictions are locked without a subscription, but you can enjoy the selection of free tips. A team page (Celta Vigo) opened with the club's statistics, its form and a predictability rate of 71.8%, but the next match said Unlock prediction. The free plan is a sample of the day's tips, and the site chooses which ones.

What the confidence score means

The score is the model's own trust in a tip, from 0 to 10. The site publishes tips at every level. On the welcome page a tip with a score of 2.3 out of 10 was shown as a win. A low score does not stop a tip from being published. It only keeps the tip out of the free seven, which on our day were all rated above 5.

Does a high score mean a better tip? The data file answers that question, and the answer is in the table after the profit numbers below.

66.6% accuracy: we checked it on 170,689 tips

The Progress page shows 66.6% correct for all 273,644 predictions since 2021 and 73% for bankers, with a chart for the last 12 months and a table by month. Under the charts are two download buttons. The second one gives a spreadsheet with every match since January 2024: date, league, teams, final score, the odds, the best tip, its confidence and its odds. The file covers 170,689 of the 273,644 matches; the 2021–2023 tips are in the site's total but not in the download. The 2026 sheet is already larger than the whole of 2025, because the site has added leagues: a Saturday in 2026 carries 800 to 900 matches against 400 to 550 a year earlier.

We opened the file and checked every best tip against the final score with our own script. Our result: 66.9% correct since January 2024. The site says 66.6% since 2021. The two figures cover different periods, but they are three tenths of a point apart, and the site's monthly table has never left the 65–68% band. The data is real, and the accuracy claim is honest.

NerdyTips profit: the number the site does not show, −4.2%

The Progress page never shows money. We added it. We put one unit (the same amount, for example $10) on every best tip at the odds in the file:

Two tips out of three win, and the account still loses 4 cents on every dollar. The reason is the odds. At average odds of 1.46 you need 68.5% correct tips to get your money back. The model gives 66.9%. The gap is 1.6 points, and it is enough to lose money. The loss is also bigger than a simple multiplication suggests, because the tips that win have lower odds than the tips that lose, so a winner pays back less than the 1.46 average. That is why the account loses 4.2% and not the 2.3% you get from 66.9% × 1.46.

Month by month the result is the same. Not one of the 32 full months from January 2024 to August 2026 ended in profit, and the first days of September 2026 are negative too. The best full month was −1.8%. The worst was −7.5%.

In money: $10 on seven tips a day, every day for a year, is $25,550 in stakes. The free seven are tips rated above 5, and that group returns about −3.6% in the file, so the loss is about $900 a year. Someone who follows every tip at the file's −4.2% loses about $1,070 on the same stakes. Both figures assume no bookmaker limits, the average odds the site shows, and no tip skipped.

How to check the record yourself in five steps

You do not have to trust our numbers. The site gives you the data.

1. Open the Progress page and scroll to the bottom. Press Download All Matches. You get an “Excel file” with one sheet per year.
2. Open it in Excel or Google Sheets. The columns you need are Goals Home, Goals Away, Best Tip and Best Tip Odd.
3. Add a column that says whether the tip won. For 1 the home goals must be higher; for under 2.5 the goals added together must be 2 or fewer, and so on.
4. Add a column for profit: odds minus 1 if the tip won, −1 if it lost.
5. Sum the profit column and divide by the number of rows. That is the return per unit; anything below zero is a loss. To see why it is negative, also count the share of tips that won and compare it with 100 divided by the average odds, which is the win rate you need to break even. Our betting calculators do that last step for any odds.

If your return is close to −4%, you did it right.

Bankers and the 73%: the best tips lose less, not nothing

The site's strongest sales line is 73% banker win rate. Bankers are the tips with the highest confidence. The file does not mark which tips were bankers, so we used the confidence score as the closest test: the top-confidence tips in the file are the nearest thing to the bankers the site sells. The file stores confidence from 0 to 100; 80 in the file is 8.0 on the site.

Confidence works as a ranking. Tips with higher scores win more often. But their odds are lower. A tip with a score of 90 or more wins 71.4% of the time at odds of 1.38, and 1.38 needs 72.5% to break even. As in the whole file, the winners here have lower odds than the group average, which is why the return is −2.4% rather than the −1.5% a simple multiplication gives. The top-confidence tips lose less than the other tips. They do not win.

Sorted by odds, the result repeats in every group. Tips priced up to 1.30 win 76% and return −4.1%. Tips between 1.30 and 1.50 win 69% and return −4.1%. Tips between 1.50 and 1.80 win 59% and return −4.4%. This means the model's win rates fit the bookmaker's prices very well. The model has the same information as the bookmaker. To make money, a model needs more.

The big five leagues alone (4,512 tips) came out at 67.8% correct and −3.9%. The loss does not come from the small leagues. It is the same everywhere.

Which market loses least: under, over, 1X or home win

The next question a reader asks is whether one type of tip does better than the rest. The file answers that too. The most common best tip is under a goals line, one third of the file. The only type that loses less than 2% is a plain home win.

The result is the same as with confidence: safer tips have lower odds, and at lower odds a 65–70% hit rate loses more. No market is in profit.

One more comparison shows how much the model adds. In the same 170,689 matches, a bet on the favourite in every game, using the odds in the file, wins 51.4% at average odds 1.95 and returns −4.9%. NerdyTips returns −4.2%. The model beats the simplest possible strategy by 0.7 points. It does not beat the bookmaker's margin, and the margin is what costs the money.

Bet Builder: the site's own 32% number

The Bet Builder joins several tips into one slip. On the free plan you cannot set a target above total odds of 2.00; the slider stops there with the note subscribe to raise the limit. The site also prepares one free slip a day, built to that target. Ours came out at 2.05: three picks, average trust 6.2. Under it the site prints:

Total odds: 2.05 · Average trust: 6.2 · Hit probability: 32.3%

Use the site's own number. A slip at 2.05 needs to win 48.8% of the time to break even. The site says this one wins 32.3% of the time. That means you lose about a third of every stake on average. The site is honest to print the number. Why it still offers the slip is not explained.

Hit&Win is a contest, not a tool. You predict the result (1, X or 2) of ten chosen matches. If all ten are right, you win a lifetime subscription. Free users get one entry a day. Each match has three possible results, so ten in a row is very unlikely.

The NerdyTips app

The app exists on both stores under the developer's name. On the App Store, NerdyTips — AI Betting Picks has a 4.5 rating from 133 ratings and an 18+ age label. On Google Play, NerdyTips has 4.7 from about 2,200 reviews and more than 100,000 downloads. The store screenshots show the same features as the website: bankers, all matches, match pages, Bet Builder, and an AI accumulator that builds a slip to a target total odds. We did not test the app. The free limits are the same seven tips as on the website, according to the store description.

NerdyTips pricing: $8.99 or $18.99?

We saw two price lists. Before login, the homepage showed the monthly plan at $18.99, 3 months at $42.99 and 6 months at $67.99. After login, the plan page showed the monthly plan at $8.99, 3 months at $19.99 ($6.69 a month), 6 months at $31.99 ($5.29 a month) and a Club Member lifetime plan at $99.99. Same plans, less than half the price. The price seems to depend on whether you have an account. Sign up before you pay.

All plans open the same things: every prediction in every league, the bankers, the full Bet Builder. There are no feature levels. The lifetime plan adds dedicated support and early access to new features.

The monthly and multi-month plans carry a 3-day money-back guarantee, which is better than most tools in this category; the lifetime plan does not mention one. The terms say plans renew automatically. The site's own pages show 4.5 Trustpilot; the Trustpilot page itself shows 4.2 from 166 reviews, with the company inviting customers to review and replying to three quarters of the negative ones. The newest visible negative review, from July 2026, reads in part:

A tentative one month subscription on my part, then continue to renew it monthly without your consent (at the highest rate on their price list).

The company replied. The complaint matches how the renewal rule is written.

What NerdyTips does well, and where the sales pages promise more

What works. The data download is the most open thing we have seen in this category: every tip since January 2024, with odds, free, for anyone. The seven free tips are the same tips paying users get, with the same odds. The coverage is huge; if you follow a small league, NerdyTips probably has a page for it. The match and team pages are good reference material. The price is low, there is a refund window, the FAQ says plainly that the site sells no fixed matches, and the footer has a responsible gambling line.

What raises questions. Three lines from the site's own pages:

Profitable tips and reliable, long-term profits that grow steadily with you (homepage)

73% Banker Win Rate (plan page, as a reason to buy)

Data proves us right (Progress page)

The site's own file shows a loss in every full month and in every confidence group. The 83% prediction accuracy on the welcome page is a short recent list, not the record. The 4.5 Trustpilot on the same page is 4.2 on Trustpilot. The site hides none of this, but you have to open a spreadsheet to see it, and most buyers will not.

Our 3.3 out of 5 comes from this split. Openness and coverage would earn a 4 on their own. The tips themselves, judged by profit, which is the only number that matters to a bettor, earn a 2. The free plan, the price and the refund window score between the two.

66.9% right and 4% down: who NerdyTips is for

NerdyTips fits a reader who wants a second opinion and a statistics page on a match nobody else covers, and who treats a tip as information, not as a bet to copy. It fits people who build low-odds accumulators and want a quick shortlist, as long as they know that the shortlist also loses money over time. It does not fit anyone who reads 73% as profit. And it does not fit anyone who plans to follow every tip, because the site's own numbers show the result: about 4% lost.

Alternatives: SportBot AI, ScoreGPT and FootballPredictAI

SportBot AI is the closest comparison and the opposite design: far fewer picks, profit and closing line value instead of hit rate, and a free tool to check your own bet. Its soccer totals model shows a small profit; NerdyTips shows a small loss on a hundred times more tips. ScoreGPT uses five named language models and a public leaderboard, with far fewer leagues. FootballPredictAI is a predictor on demand for the top-5 leagues and UEFA cups: you pick the fixture, it generates the tip, and it keeps no archive of past predictions, so its accuracy figure cannot be checked the way NerdyTips' can. The full comparison is in the AI prediction tools guide.

SportBot AI is the closest comparison and the opposite design: far fewer picks, profit and closing line value instead of hit rate, and a free tool to check your own bet. Its soccer totals model shows a small profit; NerdyTips shows a small loss on a hundred times more tips. ScoreGPT uses five named language models and a public leaderboard, with far fewer leagues. FootballPredictAI is a predictor on demand for the top-5 leagues and UEFA cups: you pick the fixture, it generates the tip, and it keeps no archive of past predictions, so its accuracy figure cannot be checked the way NerdyTips' can. The full comparison is in the AI prediction tools guide.