Partial exits and effective reward-to-risk
Follow quantities through each exit instead of crediting the entire position with the furthest target.
A practical method note with explicit evidence limits.
Price-level arithmetic and user-entered samples. A target ratio is neither a measured edge nor a guarantee of the stop fill.
Request a frozen grade definition and complete grade-by-outcome counts. A-D labels do not supply a win probability, personalised size or measured target ratio.
Keep loss, flat and open classifications identical across grades. A grade with fewer observations needs its uncertainty and selection policy disclosed.
Documentation establishes the stated mechanism, not tested trading performance. Checked 9 October 2026; a later version can differ.
Follow quantities through each exit instead of crediting the entire position with the furthest target.
Separate completed-trade R, compounded equity and the changing cash base.
Stress a payoff assumption without pretending a favourable average bounds the next loss.
An invented grade review begins with the label definition effective when each original instruction was produced. A later definition cannot retrospectively classify earlier observations without a separately labelled remapping. Keep complete counts for positive-net, negative-net, flat, open and unfilled states, using the same outcome policy across grades. Several partial exit legs are still one position when the population unit is positions. A grade with a smaller count is not automatically better because its selected winners look cleaner. Missing original definitions or excluded rows remain unresolved fields rather than being supplied from a favourable later chart.
A 3:1 price plan has a clean conditional break-even line of 25%; it does not calibrate an A-D label. A sample win fraction needs a declared population and can change with gross versus net classification, incomplete positions and revised exits. A ratio can be arithmetically correct while there is no empirical information about the grade. An official interval method is relevant only after its observation assumptions are considered. This page does not generate probability from a credential, a signal colour or a supplied model aggregate. Actual frozen grade-by-outcome records and producing versions would be required.
The fictional reviewer has a readable grade and a completed hypothetical quantity example, but no actual calibration ledger. The narrow result is that the label is identifiable while its empirical probability remains unknown. Correctness fixtures do not count as additional observations for that ledger. Neither a timestamp over entry-related bytes nor an independently published competition return fills the gap. A personal allocation rule would need another policy owner and account denominator. The criterion deliberately stops short of a universal best grade, recommended risk percentage or external audit conclusion.
A fictional label policy changes mid-period. Keep the two effective versions and their eligible populations separate before a comparison. Pooling them without a declared remapping changes the producing observation. The calculator cannot infer that mapping, and a favourable later distribution cannot define what an earlier grade originally meant.