Research / GW-S01
A specimen of the work.
Fictional inputs. Inspectable arithmetic. A redacted publication format.
GW-S01 / Research noteSample — fictional
The price of a small edge.
- Counterparty
- Venue / source
- Study window
- 30 days · synthetic inputs
A 2.5 percentage point change in assumed win probability produces £29,250 of additional expected value on £600,000 of modelled turnover. The result depends on the probability estimate surviving fees and execution.
- Net expected return
- 5.775%
- Incremental expected value
- £29,250
These are invented assumptions and calculated expectations. No client, transaction history or realised return is represented. A positive expectation can still produce a loss.
01 / Question
Does a selected set of decisions have a higher expected return than a baseline when both use the same stake, payout and execution cost?
02 / Inputs
The example assigns 60,000 hypothetical decisions a £10 stake each, across a 30-day study window. A win returns the stake plus 0.95 times the stake; a loss loses the stake. Every decision costs a further 0.5% of the stake. The baseline win probability is 52%; the selected scenario is 54.5%.
The 1.2 million input-observation count is an illustrative processing workload. It is not a real dataset or evidence that the selected probability can be achieved.
03 / Calculation
EV per £1 staked = p × 0.95 − (1 − p) − 0.005
- Baseline
- 0.900% × £600,000 = £5,400
- Selected
- 5.775% × £600,000 = £34,650
- Difference
- £34,650 − £5,400 = £29,250
Break-even win probability is 1.005 ÷ 1.95 = 51.5385%. At 51.5%, the same turnover produces −£450 of expected value.
04 / Limits
No outcomes were observed and no statistical fit was performed. The probability uplift is an assumption, not a measured effect. The scenario omits capacity constraints, changing prices, rejected decisions and uncertain costs.
Independence is not required to add expectations. It would be required by a simple independent-trials variance estimate; correlated losses can materially change the risk of the same expected return.
05 / Source-specific appendix
Redaction specimen. No source identities or confidential text are embedded.