BY 
Return on investment is the most abused number in sports betting content, and it's also the most useful one once you calculate it honestly. The formula takes ten seconds to learn. The judgment about when to trust an ROI figure takes longer, so that's where we'll spend most of our time.
The formula, and what counts as money risked
ROI equals net profit divided by total amount risked. Not your bankroll, and not your deposits. The denominator is the sum of every stake you actually put at risk, and you can compute the figure per bet or across an entire record.
Units make this cleaner than dollars, since they strip out how big your bankroll happens to be. If one unit is 1% of your roll and you make 100 one-unit bets, you risked 100 units, whatever the dollar amount was. Everything below is in units for exactly that reason.
One wrinkle worth handling now: if your stakes vary, the denominator has to reflect that. A bettor who risks 2 units on some plays and half a unit on others sums the actual stakes, not the bet count. Skipping that step quietly inflates the ROI of anyone who bets big on their winners in hindsight, which is a trick worth knowing about even if you'd never pull it yourself.
A worked example at standard juice
Say you go 54-46 across 100 flat one-unit bets, every one of them at -110. Each win returns 0.91 units of profit, so your 54 wins earn about 49.1 units. Your 46 losses cost 46 units even.
Net profit: 49.1 minus 46 comes to 3.1 units, earned on 100 units risked. That's a 3.1% ROI, and at standard juice it represents a genuinely strong run. Notice how modest the number looks next to the work behind it.
ROI versus win percentage
Win percentage means nothing without prices attached. Nothing at all.
Watch two bettors over 100 one-unit bets. The first bets underdogs at +150 and hits just 45% of them: 45 wins pay 1.5 units each for 67.5 units of profit, while 55 losses cost 55 units, leaving a net of +12.5 units and a 12.5% ROI.
The second bets heavy favorites at -200 and wins an impressive-sounding 58%: those 58 wins pay 0.5 units each for 29 units, while 42 losses cost 42 units, leaving a net of -13 units and a -13% ROI. The 58% bettor sounds sharper at a barbecue. The 45% bettor is the only one making money, because breakeven at -200 is 66.7% and the favorite bettor never got close.
This is exactly why the CAPTRACKER leaderboard ranks its 900+ tracked handicapper profiles by units won and ROI rather than by win rate, and why a gaudy record on the NFL capper pages deserves a second look at the prices underneath it.
What realistic ROI actually looks like
Efficient markets compress edges. In heavily bet markets like NFL sides and totals, long-run profitability tends to live in the single digits, and a bettor sustaining 5% over thousands of bets is doing something exceptional.
So treat triple-digit ROI claims over big samples as a red flag rather than a brag. The math is unforgiving here: at -110 juice, a 100% ROI on flat bets is literally impossible, because you'd need to win more than 100% of your wagers to get there. Claims like that survive only on tiny samples or on records nobody can audit.
Short samples are a different animal. Doubling your money across 20 bets happens to lucky people constantly, and variance alone produces runs like that all the time. Enjoy them. Don't build a self-image on them.
Sample size decides how much an ROI means
Flat one-unit bets at -110 swing about one unit per bet, which gives us a clean way to think about noise. Across 100 bets, your observed ROI can easily sit 10 percentage points away from your true skill in either direction, purely by chance. A genuinely breakeven bettor will post a +10% ROI over 100 bets more often than you'd guess.
Across 1,000 bets, that drift shrinks to roughly 3 points. Across 10,000, it's closer to 1. The number stops wobbling as the sample grows, which is the entire reason sample size matters.
- Under 100 bets: ignore the ROI entirely, positive or negative.
- 100 to 500 bets: interesting, but still very noisy.
- 500 to 5,000 bets: starting to mean something, especially alongside consistent closing line value.
- Beyond 5,000 bets: a real record, worth judging with confidence.
Verification matters as much as size. A big sample only counts if every pick was locked before the games started, which is why picks on CAPTRACKER are timestamped with a two-minute grace window, then auto-settled against ESPN data with no edits or deletes allowed. The methodology page covers exactly how the grading works.
Calculating yours this week
A confession first. In my first month as a projections analyst, I bragged about a hot win rate in a team meeting, and my manager asked one question: at what price? I never led with win percentage again.
Pull your full bet history from your book, not the flattering excerpt from one good month. Sum every stake, sum every return, subtract, and divide. If you don't have a unit system yet, my bankroll and units guide will get you set up in an afternoon.
Then hold the result loosely. A modest ROI on a big honest sample beats a flashy ROI on a tiny one every single time, and knowing that already puts you ahead of most of the people selling picks.