Tipster rating methodology

We count public posts on X in which a tipster picks a specific match. A pick counts only if the post went up before kick-off and no earlier than 1-4 days ahead (shorter for tennis and baseball, where teams play daily). Retweets are ignored; replies count only when the author continues their own thread.

Results posts (“yesterday ✅❌”), bookmaker “boost” promos and quoted market lines (“DraftKings: −5.5”) are filtered out. Once a pick is recorded, it stays in the tipster history even if the post is deleted later.

We look for team and player names from the fixture calendar: full names, nicknames (Spurs, Barça, Jags), national team names in 10+ languages and Russian case forms. A one-word name must be capitalised, so “you” never becomes a tennis player.

Markets that can be graded from the score:

• ✓Match result (1/2, moneyline, “to win”)
• ✓Draw and double chance (1X, X2)
• ✓Handicaps and spreads, incl. Asian quarter lines
• ✓Over/under totals, incl. 2.75 and “1.5, 2.0”
• ✓Team totals (TT)
• ✓Both teams to score: yes/no
• ✓Correct score
• ✓Parlays and same-game combos

Corners, cards, halves, player props, live bets and picks posted as images are not graded. Odds are read from the text: decimal, American (−110, +150) or fractional (5/4). Lists with one combined price, a booking code or “+” lines count as a single parlay.

Final scores are checked against public match results: LiveScore for football, basketball, hockey, tennis, cricket and American football, and ESPN for MLB. Football markets are graded on regular time (90 minutes plus stoppage time). Hockey, basketball and NFL include overtime, as sportsbooks settle them. Cancelled and postponed games are void.

A parlay loses if any leg loses. A parlay with an ungradable leg is left out of the stats unless another leg has already lost.

Raw hit rate is misleading: 90% on “over 0.5 goals” takes less skill than 45% at 2.50 odds. So every pick gets an expected probability p₀:

• with odds: p₀ = 1 / (odds × 1.05), the price without the bookmaker margin;
• without odds, match result, draw, double chance and draw-no-bet use an Elo rating of teams and players built from every result in the database, using only games played before the event;
• football totals and BTTS use frequencies over 19,100 played matches (both teams score in 55%, draws 23%), and hockey and baseball totals use the frequencies of their played games too, so the line matters: over 4.5 in hockey hits far more often than over 6.5;
• handicaps in football, hockey and baseball use a Poisson model of the goal margin fitted to Elo team strength, so +1.5 on an underdog and −1.5 on a favourite get different expectations. Tennis total games come from the distribution of real matches of the same format (best of three or five sets), and set totals and set handicaps from a per-set win probability implied by the rating. NFL and basketball spreads and totals stay at 50% because the main line is set to split the action;
• finally, expectations for picks without odds are shifted per market so the average tipster (every author weighted equally) hits exactly as expected. Tipsters choose their spots; without this, anyone backing favourites would look like a genius.

Actual wins W are compared with expected wins E = Σp₀, with small samples shrunk:

50 means as good as the market. Above 60 is a steady edge, above 70 is rare. The 30 in the denominator keeps a five-pick heater from topping the table. Tipsters need 10+ graded picks to be ranked; bookmaker accounts are listed separately.

Asian half-wins and half-losses count as half a stake. Pushes are excluded from the hit rate.

Most tipsters on X do not beat the market. Picks with odds win 49.9% of the time, almost exactly what the no-vig odds predict (50.6%), so their ROI is negative: −6.0%.

139 of 281 ranked tipsters have an index above 50, and only 44 reach 60 or more. 73 of the 193 tipsters with at least 5 priced picks are in profit. Picks without odds win more often (58.2%), but they are mostly favourites: 57.1% was expected.

An edge only partly persists. We split the records of 144 tipsters with 60 or more picks into two halves in time: the top quarter by the first half (edge +10.4 pp) kept +1.2 pp in the second half, while the bottom quarter (−10.0 pp) moved back to almost zero (−0.3 pp).

When three or more tipsters pick the same match, the majority pick wins: Football 69.9% against 66.9% expected (976 matches); Tennis 59.2% against 56.9% expected (861 matches); Baseball 57.0% against 57.3% expected (728 matches); NFL & NCAAF 64.0% against 57.0% expected (293 matches); Basketball 50.0% against 52.3% expected (80 matches); Hockey 59.0% against 59.2% expected (61 matches).

By sport

By market

BetBuddyRank calculation over 48,784 graded picks from 1 August 2026 to 7 October 2026. ROI is shown for 30 or more priced picks.

Automatic parsing makes mistakes: unknown names, ambiguous wording, image-only picks and deleted posts. Losses deleted before a pick was recorded are not counted. Every row can be checked: each pick links to the original post and the final score.

The leaderboard describes past public picks and promises nothing about the future. Betting involves the risk of losing money. 18+.

Everything on the site is plain static JSON you can fetch directly:

1. Open the capper leaderboard and pick a period and a sport.
2. Sort by the index, not the hit rate: the index accounts for the odds.
3. Open a tipster profile and look at the number of picks, the markets and the form.
4. Check a few picks via the links to the posts, or look a capper up with the tipster check by handle.