How to Read a Tipster's Record: Units, ROI, and Four Padding Tricks
Every paid picks service shows a winning record. The record is the product page. No regulator sets a reporting standard for touts, so each service picks its own math, its own start date, and its own definition of a unit.
You need three numbers to judge any record: flat-unit profit, return on investment, and the number of bets behind both. This guide shows how to compute each one, the four ways records get inflated, and the sample size where a winning record stops looking like luck. Run the same checks on us. We show our numbers at the end.
Why Win Rate Tells You Almost Nothing
Win rate is the number most services lead with. Without the average price, the number is useless.
A 55% record losing money
Record: 55-45, average price -130
Break-even at -130: 130 / 230 = 56.52%
Profit at 1 unit risked per bet: 55 × (100 / 130) - 45 = -2.69 units
ROI: -2.69 / 100 risked = -2.69%
Flip the price and the read flips too. A 40% hit rate on +175 underdogs is a winner, since +175 breaks even at 36.36%. A record without the price attached is a marketing line, not data.
Number One: Flat-Unit Profit
A unit is a fixed slice of bankroll, set before the bet. Many services size their plays: 1 unit on a lean, 3 units on a "max play." Sizing by confidence is a legitimate method if the size is posted before the game. The record turns dishonest when the size gets chosen, or changed, after the result.
Here is how fast after-the-fact sizing turns a loser into a winner.
Ten picks at -110, 4 wins and 6 losses
As reported, winners at 3u, losers at 0.5u:
4 × 3 × 0.9091 - 6 × 0.5 = +7.91 units, ROI +52.7%
At flat 1u per pick:
4 × 0.9091 - 6 = -2.36 units, ROI -23.6%
Your defense is a recount. Take the pick log, give every pick 1 unit, and sum the results at the posted odds. If the flat number sits far below the advertised number, the gap came from sizing. Then ask one question: was each size posted before the game? A timestamped post settles the question. A spreadsheet the service maintains does not.
Number Two: ROI
ROI is net profit divided by total amount risked. A service risking 300 units and netting 18 units shows a 6% ROI. Unlike win rate, ROI compares fairly across favorites and underdogs.
Set the bar with the price. At -110, a bettor needs 52.38% to break even, a 0% ROI before any subscription fee. Every dollar of the monthly price comes out of whatever edge sits above the line. The vig math guide shows where 52.38% comes from.
Treat a large ROI on a short record with suspicion. The academic record on tipsters is not kind. Spann and Skiera's 2009 study in the Journal of Forecasting compared tipsters, betting odds, and prediction markets on three seasons of German Bundesliga matches. Odds and prediction markets forecast equally well. Both, in the authors' words, "strongly outperform tipsters." Amor and Griffiths extended the newspaper-tipster method of Forrest and Simmons (International Journal of Forecasting, 2000) to Australian football. Most tipsters cleared simple performance tests, under-used public information, and only two appeared to use independent information well.
Number Three: Sample Size
ROI tells you the rate. Sample size tells you whether the rate means anything. A record is a coin-flip sequence until the bet count says otherwise.
The NIST Engineering Statistics Handbook recommends the Wilson method for confidence intervals on a proportion such as a win rate. Here are Wilson 95% intervals for records at -110, plus the chance a zero-edge bettor posts the same win rate or better.
| Bets | Win rate | ROI at -110 | 95% range for true win rate | Chance with no edge |
|---|---|---|---|---|
| 50 | 60% | +14.5% | 46.2% to 72.4% | 14% |
| 100 | 58% | +10.7% | 48.2% to 67.2% | 13% |
| 200 | 56% | +6.9% | 49.1% to 62.7% | 15% |
| 500 | 55% | +5.0% | 50.6% to 59.3% | 12% |
| 1,000 | 55% | +5.0% | 51.9% to 58.1% | 4.9% |
"Chance with no edge" is the one-sided probability a bettor with a true 52.38% rate, the -110 break-even, matches or beats the record. A 58% run over 100 bets shows up 13% of the time with zero skill. A 55% bettor needs about 984 bets before the record clears the 5% bar. A 54% bettor needs about 2,575. The sample size guide walks through the formula.
Trick One: Missing Losses
The bluntest trick is leaving losers out. Picks go out in a private channel, winners get reposted publicly, and losers never reach the record.
- Check the pick count per month. A service posting daily with a blank week in the middle of the season had a reason to go quiet.
- Compare members' screenshots and recap posts against the official log. A pick members remember and the log lacks is a deleted loser.
- Look for posted-before-start timestamps on every pick. Without a timestamp, nothing stops a pick from entering the log after the game.
Trick Two: After-the-Fact Unit Sizes
Covered above with the 4-6 example: a 40% record at -110 reported as +52.7% ROI. The test is the flat recount. The fix is a unit size stamped on the original post.
Trick Three: Odds You Never Got
A service logs a winner at +175. By the time the alert reaches you, the best price left is +155. Both numbers are real. Only one was yours.
One pick, two prices, true win chance 38%
At +175: break-even 36.36%, EV 0.38 × 2.75 - 1 = +4.5%
At +155: break-even 39.22%, EV 0.38 × 2.55 - 1 = -3.1%
The record shows a profitable pick. The subscriber placed a losing bet on the same team. Ask every service which book and what time stand behind each posted price. Better still, ask for closing line value. A service whose picks beat the closing number has evidence of edge no matter which price went in the log. A service without CLV data has prices nobody audited. The CLV guide shows the calculation.
Trick Four: Survivors
Brown, Goetzmann, Ibbotson, and Ross showed in the Review of Financial Studies in 1992 how a sample cut down to survivors creates the look of skill. Dead funds leave the data. The ones left look like they predicted the market. Picks services work the same way. Services with bad runs shut down or relaunch under a new name, and the new record starts at zero.
Run the numbers on 100 services with no edge at all, each posting 200 picks at -110 and winning exactly half the time on average.
100 zero-skill services, 200 picks each
Wins needed to show a profit: 105 of 200
Share showing a profit: 26%, about 26 services
Share at +5.0% ROI or better (110+ wins): 9%
Share at +6.9% ROI or better (112+ wins): 5%
Five services with zero skill post a 6.9% ROI over 200 picks. Those five buy the ads. The other 95 close. A ranking page of "top cappers" shows you the five.
Fake history goes further than survivors. In 2001 the U.S. Attorney for the Western District of Wisconsin announced guilty pleas tied to Gold Medal Sports. The charging papers described handicapping mailers promising "guaranteed" winners and "inside information." Two of the advertised handicappers, Mike Wynn and Sam Sharp, "were not real persons." A paid celebrity front, per the filing, "was not involved in any sports handicapping whatsoever."
Calculator: Check Any Record
Enter wins, losses, and the average American price. The tool returns break-even, flat-unit ROI, the Wilson range for the true win rate, the chance a zero-edge bettor matches the record, and the bets needed before the win rate clears the 5% bar.
Tipster record check
The Six Questions to Send Before You Pay
- Will you show the full pick log with a timestamp on every pick, posted before the game?
- What is the flat-unit ROI across every pick, and how many picks?
- Where were unit sizes posted, and when?
- Which book and what time stand behind each logged price?
- What is the average closing line value over the last 200 picks?
- When did the service start, and under what name?
A service with a working log answers all six in a few minutes. Deflection is an answer too.
Run the Same Tests on Us
Our homepage shows +302.9 units on official plays since March 2025 and 16 winning months out of 19, from the Monthly Summary closed through September 30, 2026. Here is how the number holds up against this guide. Past results don't guarantee future results.
- Units are sized by the analyst at post time, not flat. The alert cards on our homepage show sizes from 0.3u to 3.5u. Judge the size calls by the timestamp on the original post.
- The Daily Recap posts every night in the Discord, wins and losses. Three of the 19 months on the record finished red, and they stay in the count.
- Our biggest wins page lists only analyst plays and two named members. We left off five-figure screenshots from brand-new accounts we had no way to verify.
Check the room before you pay. The free Discord room needs no card, and the $7 week pass covers the paid side for a week.
The verdict: judge a record on flat-unit ROI, at the price you would have gotten, across enough bets to beat luck. A 58% run over 100 bets happens 13% of the time with zero skill. Five of every 100 zero-skill services post a 6.9% ROI over 200 picks. Ask for the timestamped log, recount the log flat, and check the closing line. Everything else is the sales page.
Sources and Method
Spann and Skiera (Journal of Forecasting, 2009) supply the tipster versus odds comparison. Amor and Griffiths (University of Melbourne working paper 871) supply the extension of Forrest and Simmons's tipster method. Brown, Goetzmann, Ibbotson, and Ross (Review of Financial Studies, 1992) supply the survivorship mechanism. The Department of Justice release supplies the Gold Medal Sports case. The NIST handbook supplies the Wilson interval. Every win rate, ROI, interval, and probability above comes from the formulas shown, run at exact binomial or normal values, and you recompute each one in a spreadsheet with BINOM.DIST and NORM.S.DIST.
Limits: the "chance with no edge" column uses a normal approximation to the binomial against a 52.38% break-even. The 100-service example assumes independent picks at exactly -110. Real records mix prices, so use the calculator with the record's own average odds.