Different costs on winning and losing branches
Solve break-even only after keeping each branch cost attached to its payoff.
Why the common-cost lens is narrower
The retained cost lens assumes the same additional per-unit cost on the specified winner and loser. That makes its two-branch denominator especially simple, but real or hypothetical policies may assign different costs. Let gross winning amount be G, gross loss magnitude L, winner cost Cw and loser cost Cl. Net winner is G less Cw and net loss magnitude is L plus Cl. A conditional break-even fraction, when the net winner is positive, is net loss divided by net winner plus net loss. The input costs must use the same cash or R unit as their branch amounts.
A filled unequal-cost calculation
Use invented gross winner three R and gross loser one R. Winner cost is 0.1R and loser cost 0.3R, giving a 2.9R net winner and a 1.3R net loss magnitude. Conditional break-even is 1.3 divided by 4.2, about 30.95%. At an assumed 40% winner frequency, net outcome averages 0.4 times 2.9 less 0.6 times 1.3, or 0.38R. Equal 0.1R cost instead gives a different break-even denominator and result. Keep both policies labelled; do not alter only the losing rows after inspecting an observed sample to flatter a conclusion.
Feasibility is part of the finding
If the winner cost consumes the entire gross winning amount, there is no positive winner payoff in that branch definition. If it exceeds the amount, both branches may be negative and a usual profitable-frequency interpretation fails. A mathematical quotient outside the feasible range is a warning about the assumptions, not an estimated success probability. Do not clamp the rate to 100% and call it a calibrated forecast. A third outcome, variable cost or path-dependent financing changes the model again. The branch amounts are assumptions unless linked to a defensible actual transaction population.
Retain the source and policy version
For actual records, costs belong to transactions, quantities and applicable fee schedules. A desired average can be a scenario parameter but must not overwrite those original charges. Spread already included in transaction prices is not automatically another extra fee. A software correctness check can establish this formula's arithmetic while giving no evidence about the frequency of a winner or the future average cost. Preserve the unchanged actual record alongside each counterfactual. The article supplies a completed hypothetical comparison without choosing an exit strategy, assuming a broker tariff or declaring that a provider has a profitable measured edge.
Completed invented payoff case
The invented branch record specifies winner 3R, loser magnitude 1R, winner extra cost 0.1R, loser extra cost 0.3R and assumed winner frequency 40%. Net branch amounts are 2.9R and negative 1.3R. Break-even is about 30.95% and assumed average is 0.38R. No actual trading population or cost schedule is supplied. The original equal-cost tool remains unchanged in meaning; this case is a separate declared policy, not a new empirical result.
| Object | Record | Bounded finding |
|---|---|---|
| Winner | 3 - 0.1 | 2.9R |
| Loser magnitude | 1 + 0.3 | 1.3R |
| Conditional break-even | 1.3 / 4.2 | 30.95238% |
| Assumed average | 0.4 x 2.9 - 0.6 x 1.3 | 0.38R |

Primary references and scope
Expanded source review: 13 October 2026. Documentation supports the stated mechanism, not a tested account, hypothetical fill or provider outcome.