Results
Edge, after costs.
A result needs a denominator, a time window and a credible alternative. Here is what a performance ledger should expose.
Placeholder figures for a 30-day synthetic study. No client performance is reported.
Illustrative performance ledger
- Input observations
- 1.2m
- Placeholder processing volume
- Modelled turnover
- £600k
- 60,000 decisions × £10
- Net expected return
- 5.775%
- After the assumed execution cost
- Incremental expected value
- £29,250
- Against the same-volume baseline
| Measure | Baseline | Selected scenario |
|---|---|---|
| Assumed win probability | 52.0% | 54.5% |
| Net win payout / stake | 0.95× | 0.95× |
| Cost / stake | 0.5% | 0.5% |
| Expected return / stake | 0.900% | 5.775% |
| Total expected value | £5,400 | £34,650 |
The return difference is +4.875 percentage points. Expected value is a model output, not realised profit. The example assumes each decision settles, a loss loses the full stake, and the stated cost applies on every decision.
How quickly does it disappear?
| Scenario | Win probability | Net expected return | Total expected value |
|---|---|---|---|
| Adverse | 51.5% | -0.075% | -£450 |
| Central | 54.5% | 5.775% | £34,650 |
| Favourable | 57.5% | 11.625% | £69,750 |
Break-even is 51.5385% win probability. These are chosen stress scenarios, not confidence bounds. Lower probability, limited capacity, larger fees or correlated losses can erase the apparent advantage.
Three levels of evidence.
The number gets less interesting when the conditions behind it are missing.
Apparent edge
A price or pattern differs from an estimate. It has not yet survived costs, leakage checks or a fair baseline.
Tested edge
The effect survives withheld evaluation data and sensible changes in the assumptions. Report uncertainty and the cases where it fails.
Executable edge
The opportunity remains after limits, latency, settlement and actual costs. Capacity and stop conditions matter as much as average return.