The flat-$100 convention
Every ROI-eligible selection is treated as a $100 stake. A win at positive odds earns $100 multiplied by the odds divided by 100. A win at negative odds earns $100 multiplied by 100 divided by the absolute odds. A loss is −$100. A push returns the stake, produces $0 profit, and still counts as $100 risked.
Amount risked = $100 × ROI-eligible selections
ROI = net profit ÷ amount risked × 100
A three-selection example
Suppose a record contains a win at -120, a loss at +140, and a push at +110. The first result earns $83.33. The loss costs $100. The push adds no profit. Net profit is −$16.67 on $300 risked, so ROI is −5.56%.
Win at -120: +$83.33
Loss at +140: −$100.00
Push at +110: $0.00
Net: −$16.67
Risked: $300.00
ROI: −5.56%
Why ROI sample and win-rate sample differ
The ROI sample includes wins, losses, and pushes because each represents a $100 stake under the flat-stake convention. Win rate uses wins and losses only. In the example above, the ROI sample is three, while the win-rate sample is two and the win rate is 50%.
Voids and unresolved results are excluded from ROI because no final profit or loss can be assigned. They remain visible in the ledger so exclusion does not make the historical record disappear.
Sample size is context, not a quality badge
A short run can move sharply after one result, especially when prices vary. A larger sample does not prove that future performance will match the past, but it makes the displayed average less dependent on a single outcome. Always read the sample count, date range, included ranks, and prices beside the percentage.
Published price prevents hindsight
Using the recorded price means a result is not recalculated later with a more favorable line. If a selection cannot be matched to a valid published price or its result remains unresolved, it does not receive invented profit. The Results ledger shows the active date range, ROI, win rate, and both samples together.
What AI review can and cannot do
AI-assisted review may check formulas, arithmetic, and internal consistency. It cannot make a small sample representative, change a recorded price, or guarantee future returns. ClearInning remains responsible for the published method and corrections.