Why win rate alone is misleading: prices, stakes and missing losses
Win rate counts decided results. It leaves out the payout of each win, the size of each loss and the completeness of the sample. The same percentage can describe very different financial outcomes.
Pair win rate with the price
| Record | Price on every ticket | Wins / losses | Net profit | ROI |
|---|---|---|---|---|
| A | −200 | 6 / 4 | −$10 | −10% |
| B | +100 | 6 / 4 | +$20 | +20% |
| C | +200 | 4 / 6 | +$20 | +20% |
In A, each win earns only $5. Six wins earn $30, while four losses cost $40. Total return is $90 on $100 staked. B has the same 60% win rate but earns $10 per win. C wins only 40% yet also produces +$20 in this invented sample.
For one fixed binary price, break-even win rate is the reciprocal of decimal odds. At −200 it is 66.67%, so winning more than half is not enough. Mixed prices require ticket-level accounting; a headline average of American odds is not a reliable substitute.
Stakes can reverse the impression
Consider nine +100 wins risking $1 each and one +100 loss risking $20. The win rate is 90%, but profit is $9 − $20 = −$11 on $29 staked, or about −37.93% ROI. The larger loss dominates the many smaller wins.
Report actual stakes and, if useful, a separately labeled flat-stake comparison. Do not quietly switch from actual to flat stakes when one makes the history look better. Both summaries require the complete same selection set.
Count exclusions before reading a streak
In a hypothetical set with 12 wins, eight losses and five refunded tickets, the decided win rate is 12 ÷ 20 = 60%. Calling it 12 ÷ 25 = 48% measures something else. Disclose refunds separately rather than treating them as wins or silently changing the denominator.
Now suppose a recap shows six winning picks but omits four losses from the same declared period. The visible screenshot could show 100% while the complete decided record is 60%. Correct arithmetic on a selected subset does not make the summary complete.
Do not reuse past win rate as the next forecast
A historical rate combines the opponents, prices, thresholds and decisions actually present in that sample. A new pick with different conditions does not inherit that percentage automatically. Sample uncertainty and selection bias also remain even when all entries are recorded.
Read win rate alongside net profit, settled stakes, odds, dates, sample size and original evidence. Use the linked ROI calculator to reproduce record A’s −$10 and −10%; it checks totals but cannot detect omitted tickets.
Try the numbers yourself.
Open the calculator with this guide’s example inputs. Change one assumption at a time and see what changes.
Open roi calculatorCommon questions
Is a higher win rate always better?
Not across different prices, stakes, market types or samples. Compare the full accounting and uncertainty.
Should pushes count as wins?
Not under the decided-win-rate convention used here. Report pushes and voids separately and disclose the denominator.
Sources and example scope
All records and omission scenarios are original hypothetical examples, not descriptions of any creator or evidence of a betting edge.