Mistake-Adjusted Expectancy: Separate Execution Errors From Strategy Edge
Calculate mistake-adjusted expectancy by comparing aligned-trade expectancy against the classifiable executed sample, then read the gap beside deviated-trade rate as a diagnostic signal, not a cause.
Mistake-adjusted expectancy is the diagnostic label this article uses for a specific comparison: the standard win-rate/average-win/average-loss expectancy formula, calculated separately for the classifiable executed-trade sample (aligned + deviated trades) and for the subset of trades classified execution-aligned. Read the gap between the two figures beside the deviated-trade rate (the share of that same aligned+deviated sample containing a deviated trade) to see whether execution deviations are associated with a lower calculated edge in this sample — not to prove what caused it.
This is not a new metric layered on top of existing ones, and it is not an established industry-standard term; it is a specific way of combining two frameworks that already exist on their own: the expectancy calculation from trading performance review and the aligned/deviated/unclassified execution-quality classification from measuring execution quality. Neither framework alone answers the question this article addresses: when expectancy and the deviated-trade rate move together in the same sample, what does that association actually support?
What mistake-adjusted expectancy adds that the existing frameworks don’t
Trading performance treats execution as one of three review layers but does not fold an execution classification into the expectancy formula itself. Execution-quality measurement produces decision-level alignment rates by rule, but deliberately excludes P&L from the classification and warns against mixing decision-level and trade-level units. The cost-of-rule-breaking framework sums the raw dollar P&L of deviated trades, which answers “how much P&L is associated with deviation in this sample” — a different question from “how does the calculated expectancy statistic change when deviated trades are excluded.”
This article owns the narrower question in between: recomputing the formal expectancy statistic — win rate, average win, average loss — separately for the aligned subset and the classifiable sample (aligned + deviated trades), and reading the gap between them alongside the deviated-trade rate. It does not replace the decision-level scorecard, and it does not replace the dollar-cost framework; it uses both as inputs to one observational comparison. If the result metric is profit factor rather than expectancy, profit factor and mistake frequency owns the two-axis outcome/process comparison; this article still owns the expectancy split.
Define the terms before calculating anything
classifiable-sample expectancy
= expectancy calculated on aligned + deviated executed trades
aligned expectancy
= expectancy calculated on aligned executed trades
expectancy gap
= aligned expectancy − classifiable-sample expectancy
deviated-trade rate
= deviated trades / (aligned trades + deviated trades)Aligned expectancy, classifiable-sample expectancy, the expectancy gap, and the deviated-trade rate all share one universe: aligned + deviated executed trades. Unclassified executed trades are reported separately and never enter any of these four figures. An optional all-executed-trade expectancy — every executed trade, including unclassified — can still be reported for general performance review, but it uses a different denominator and must not be substituted into the expectancy-gap comparison above.
An expectancy gap is a difference between two observed subsets of the same sample. It is not a measured effect of execution error, because the aligned and deviated trades are not the same market opportunities under different execution — see “Read the expectancy gap as an association, not a cause” below.
The standard expectancy formula, and where execution error hides inside it
Expectancy per trade is normally calculated as:
expectancy
= (win rate × average win)
− (loss rate × average loss)This formula treats every trade in the sample as equally representative of the strategy. It has no term for whether the trade was executed the way the strategy specified. A trade taken without the predefined entry gate, at the wrong size, or after an unplanned re-entry is counted exactly like a trade that matched the plan. If deviated trades are frequent enough, or skewed enough in outcome, the full-sample expectancy figure blends strategy performance and execution behavior together, with no way to separate the two from the number alone.
Calculate mistake-adjusted expectancy in four steps
1. Classify every trade before calculating anything
Use the same aligned/deviated/unclassified classification defined in the execution-quality framework: an active rule, an eligible decision, and sufficient planned-versus-actual evidence. Do not classify from the outcome. A trade is not “deviated” because it lost, and it is not “aligned” because it won.
2. Define how decisions roll up to a trade
Execution-quality measurement classifies individual eligible decisions — entry, size, management, exit, re-entry, cutoff — not trades, and explicitly warns against mixing decision-level and trade-level rates. Expectancy is a per-trade P&L calculation, so it needs exactly one classification per trade. The execution-quality framework does not define that rollup on its own, so this article uses an explicit aggregation convention rather than assuming one:
- A trade is deviated if it contains at least one deviated eligible decision.
- A trade is aligned only if every eligible decision inside it is classified aligned, with none deviated or unclassified.
- A trade is unclassified if it contains no deviated decision but at least one unclassified decision.
This is a deliberately conservative rollup: one deviated decision anywhere in a trade removes that trade from the aligned subset, even if every other decision inside it was aligned. State the convention next to any mistake-adjusted expectancy figure — a looser or stricter rollup rule changes which trades land in each subset, and therefore changes both expectancy figures.
Because the unit changes, keep these two rates separate:
decision-level execution error rate
= deviated eligible decisions
/ (aligned eligible decisions + deviated eligible decisions)
deviated-trade rate
= deviated trades (per the rollup convention above)
/ (aligned trades + deviated trades)The execution-quality scorecard reports the first rate, per rule. Mistake-adjusted expectancy is paired with the second. Do not quote a rule-specific decision-level error rate next to a trade-level expectancy split without naming the unit change; they answer different questions and will not numerically agree.
3. Calculate expectancy for each subset separately, on a consistent basis
| Subset | What it estimates | Denominator |
|---|---|---|
| Aligned only | Expectancy observed on the aligned trades that occurred | Trades classified aligned by the rollup convention |
| Deviated only | Result profile of trades containing a deviated decision | Trades classified deviated by the rollup convention |
| Classifiable sample | Baseline for the expectancy gap and the diagnostic comparison | Aligned + deviated trades (excludes unclassified) |
| Unclassified | Excluded from aligned, deviated, and classifiable-sample expectancy | Reported as a separate count, never folded into any expectancy figure |
| All-executed (optional) | General performance review only — not the mistake-adjusted comparison | Every executed trade, including unclassified |
Calculate win rate, average win, and average loss independently within each subset, on the same P&L basis throughout: net realized P&L after commissions, fees, and slippage — see measuring trading slippage and execution costs for how to size that slippage component from a fixed reference price rather than estimating it after the fact. Do not mix a gross-P&L subset with a net-P&L subset. Do not reuse the classifiable-sample win rate with a subset’s average win, and do not pool subsets that used different rule versions.
A trade with exactly $0 net P&L is a breakeven trade. It stays in the subset’s total trade count but contributes $0 to both the win sum and the loss sum. Compute win rate, loss rate, and breakeven rate against total trades in the subset — winners/total, losers/total, breakevens/total — so the three sum to 100%. The standard formula expectancy = (win rate × avg win) − (loss rate × avg loss) still holds with breakeven trades present, because the missing + (breakeven rate × $0) term contributes nothing; do not restrict win rate or loss rate to a non-breakeven denominator. Net realized P&L ÷ total trades is an equivalent, and often the cleanest, reconciliation check. The worked example below has no breakeven trades, so all three approaches agree.
4. Report the deviated-trade rate beside both expectancy figures
Report the count (n) behind every expectancy figure and the deviated-trade rate together. The deviated-trade rate denominator — aligned trades + deviated trades — is the same classifiable-sample universe used for aligned and classifiable-sample expectancy above, so all three figures stay comparable. Unclassified trades are excluded from that denominator for the same reason they are excluded from an alignment rate: they were never resolved into a classifiable outcome, so including them would misstate how often execution actually broke down versus how often the record was simply incomplete. Report the unclassified count and unclassified rate as a separate line.
Read the expectancy gap as an association, not a cause
The comparison is informative through its pattern, not through either number in isolation — and the pattern describes an association within one sample, not a causal decomposition of “strategy edge” versus “execution noise.” Filtering out deviated trades changes the composition of the sample: execution deviations can correlate with setup type, volatility regime, session, direction, discretionary context, position size, market conditions, or rule version. An aligned-only expectancy figure is an estimate for the observed aligned subset, not proof of what the classifiable sample would have earned under different execution.
| Pattern | What it supports investigating | What it does not establish |
|---|---|---|
| Aligned expectancy notably higher than classifiable-sample expectancy | Whether deviated trades share a common setup, trigger, or context worth reviewing | That removing deviations would have produced the aligned figure in real time |
| Aligned and classifiable-sample expectancy are close | The deviated-trade rate is not currently associated with a large expectancy difference in this sample | That the strategy has durable edge |
| Aligned expectancy is negative or flat | A strategy-level question that needs review independent of execution behavior | That better execution alone would fix results |
| Deviated-subset expectancy is higher than aligned | A deviation pattern coincided with favorable outcomes in this sample | That the deviation is a better rule, or that it will repeat |
| Deviated-trade rate is high but the expectancy gap is small | Deviations may be close to outcome-neutral in this sample, though still a process problem | That deviations are safe to continue |
The last row matters because it is the case teams most often get wrong: a high deviated-trade rate does not automatically mean expectancy is being damaged, and a small expectancy gap does not mean execution is fine. Report both numbers. Do not let one stand in for the other.
Worked example: same trade sample, three expectancy figures
Consider a 40-trade sample, already classified using the rollup convention above, with 0 unclassified trades — so the classifiable sample equals all 40 executed trades here. That will not always be true; when unclassified trades exist, the classifiable sample is smaller than the all-executed count.
- Aligned: 31 trades — 16 wins, 15 losses
- Deviated: 9 trades — 2 wins, 7 losses
- Classifiable sample (aligned + deviated): 40 trades — 18 wins, 22 losses (the aligned and deviated win/loss counts sum directly to the classifiable-sample counts)
| Subset | n | Win rate | Avg win | Avg loss | Expectancy per trade |
|---|---|---|---|---|---|
| Classifiable sample | 40 | 45.0% | $304.44 | $192.73 | $31.00 |
| Aligned only | 31 | 51.6% | $300.00 | $180.00 | $67.74 |
| Deviated only | 9 | 22.2% | $340.00 | $220.00 | −$95.56 |
Deviated-trade rate = 9 / (31 + 9) = 22.5% — the same aligned+deviated denominator as the classifiable-sample expectancy above (no unclassified trades to exclude in this example).
All figures reconcile from the same underlying integer counts, not independently rounded inputs:
- Aligned win/loss P&L: (16 × $300) − (15 × $180) = $4,800 − $2,700 = $2,100 net → $2,100 / 31 = $67.741935… → displayed as $67.74
- Deviated win/loss P&L: (2 × $340) − (7 × $220) = $680 − $1,540 = −$860 net → −$860 / 9 = −$95.555556… → displayed as −$95.56
- Classifiable-sample net P&L: $2,100 + (−$860) = $1,240 → $1,240 / 40 = $31.00
- Classifiable-sample average win: ($4,800 + $680) / 18 winners = $304.444… → $304.44; average loss: ($2,700 + $1,540) / 22 losers = $192.727… → $192.73
The underlying subset P&L totals reconcile exactly: $2,100 + (−$860) = $1,240, matching the classifiable-sample net P&L. The unrounded expectancy values reconcile the same way when weighted by subset size — 31 × (2100/31) + 9 × (−860/9) = 2,100 + (−860) = $1,240 — because each term is just that subset’s net P&L restated. The rounded, two-decimal figures shown in the table do not reconcile this cleanly: 31 × $67.74 + 9 × (−$95.56) = $2,099.94 − $860.04 = $1,239.90, about ten cents off $1,240 from display rounding alone. Treat the underlying integer totals as the authoritative check, not the rounded per-trade figures.
The aligned-only figure is roughly double the classifiable-sample figure, and the deviated subset is sharply negative. That pattern supports investigating what the nine deviated trades have in common — rule type, trigger, session context — before concluding anything about the strategy itself. It does not establish that the trader would have earned $67.74 per trade with perfect execution: the deviated trades were real market opportunities, not a controlled experiment, and a nine-trade subset is a small basis for the deviated-only figure specifically. Treat the deviated-subset number as a flag for review, not a stable estimate.
What this comparison cannot reconstruct
The method only has realized P&L for trades that actually occurred. An omission error — a missed valid entry, a skipped setup that met the predefined criteria, a trade the process called for but the trader never took — is an execution deviation with no executed trade attached to it, so it has no realized P&L to place in any subset. Mistake-adjusted expectancy can only compare trades that were taken; it cannot assign a hypothetical outcome to a trade that wasn’t. Do not fabricate a counterfactual return for a missed trade to “complete” the aligned subset. Omission errors still belong in execution-quality review — they simply cannot enter a realized-trade expectancy calculation, and their absence means this comparison is not a full reconstruction of “the strategy as designed,” only of the trades that were actually executed.
Position size can distort the comparison
Position sizing is itself a decision that can be classified aligned or deviated. If deviated trades in the sample also carry different average size than aligned trades, part of the dollar expectancy gap reflects size, not setup or timing. Two related figures serve different purposes:
- Dollar expectancy (used throughout this article) is the account-level realized effect. It is useful for account-level review and is sensitive to trade size.
- R-normalized expectancy — win rate × average win in R minus loss rate × average loss in R, where R is the trade’s predefined planned risk, not a risk figure recalculated after the fact — is more comparable across subsets when position size varies, because it holds the risk unit constant.
Use the planned-risk field already captured for the execution-quality scorecard’s size classification to compute R. This article does not require R-normalization by default, but a size difference between the aligned and deviated subsets is a specific, checkable reason to add it before trusting a dollar-expectancy gap.
Control for composition before reading the gap as execution-related
Because filtering on execution classification can also filter on everything correlated with it, compare aligned and deviated subsets only after checking whether they are otherwise like-for-like on the fields that matter to the strategy:
- strategy or setup
- rule version
- session
- direction
- market or volatility regime
- planned-risk basis (see R-normalization above)
This is descriptive stratification, not statistical control — checking a handful of fields for gross imbalance, not isolating a causal effect. If the deviated subset is concentrated in one setup, session, or regime that the aligned subset barely touches, the expectancy gap may be describing that composition difference more than execution quality itself. Report which fields were checked. When a specific setup keeps surfacing this way, trading setup performance covers evaluating that setup on its own terms — its own sample size, session/regime segmentation, and decision thresholds — separately from this article’s execution-alignment comparison.
Failure modes specific to mistake-adjusted expectancy
Recomputing expectancy on a subset too small to support it
A deviated or aligned subset with a handful of trades can swing sharply from one outlier. Report n next to every expectancy figure. There is no universal minimum trade count; uncertainty increases rapidly as a subset shrinks, so treat a small-subset expectancy as a direction to investigate rather than a number to act on. As an optional, non-mandatory robustness check on a subset large enough to support it, a confidence interval or bootstrap resample around the subset expectancy can show how much the figure might move on a different sample of the same size. Trading data statistical reliability covers why that uncertainty shrinks with sample size in more depth.
Letting the aligned figure imply a guaranteed counterfactual
The aligned-only expectancy is an estimate of the aligned trades that actually occurred, not a forecast of what every deviated trade would have returned had it been executed correctly. The market opportunity a deviated trade encountered cannot be re-run under aligned conditions.
Reclassifying trades to make the two figures converge
Classification happens on the active rule and available evidence, decided before the expectancy split is calculated. Re-labeling a borderline trade after seeing that it would change the gap is outcome-driven reclassification, and it defeats the purpose of the comparison.
Mixing rule versions or unlike compositions inside one expectancy figure
If a rule changed mid-window, a trade evaluated as “deviated” against the old rule and a trade evaluated as “deviated” against the new rule do not belong in the same subset. Split the window at the rule-version boundary before calculating either expectancy figure, and check the composition fields above before attributing the remaining gap to execution.
Treating a favorable deviated-subset expectancy as license to keep deviating
A profitable deviated subset is evidence to review for a possible rule update through a structured feedback-loop test — not evidence to continue deviating informally. The cost-of-rule-breaking framework covers why a profitable rule violation still needs review rather than automatic promotion to a new rule.
When the gap is unclear: check for a mixed cause
If aligned and classifiable-sample expectancy stay close while the deviated-trade rate itself is high and rising, the deviations may be close to outcome-neutral so far, but the process problem persists independent of the P&L comparison. If the aligned subset itself declines while the deviated-trade rate stays flat, that is a strategy-review question, not an execution-review question. When the cause of a specific gap is not obvious from the classification and composition checks alone, hand the individual decisions to the trading-mistakes framework to distinguish a strategy problem, a risk problem, an execution problem, and a behavioral problem before changing anything.
How often to recalculate
Recalculate both expectancy figures and the deviated-trade rate on the same review cadence you already use for the execution-quality scorecard — after enough classified, non-unclassified trades have accumulated in the current rule version to make a subset comparison meaningful, not after every session. A single session rarely contains enough deviated trades to support a deviated-only expectancy figure on its own; treat single-session numbers as inputs to the next scheduled review, not standalone conclusions.
Where Costante fits
Costante’s session planning, self-defined behavioral guardrails, and low-friction trade and behavioral logging make the planned-versus-actual record this comparison depends on, and structured review surfaces discipline trends and repeated drift alongside the resulting numbers. Costante does not calculate a strategy’s statistical edge, determine whether a strategy has proven itself, prove expectancy, predict performance, automatically identify causal execution errors, or generate, backtest, or validate strategies. The classification, the rollup convention, the expectancy split, and its interpretation remain the trader’s responsibility.
Frequently asked questions
Is mistake-adjusted expectancy the same as the cost of rule-breaking?
No. The cost-of-rule-breaking framework sums the raw dollar P&L attached to deviated trades in a sample. Mistake-adjusted expectancy recalculates the formal win-rate/average-win/average-loss expectancy statistic separately for the aligned subset and the classifiable sample (aligned + deviated trades), so it answers a different question: not “how much P&L is associated with deviation,” but “how does the calculated expectancy estimate differ between the aligned subset and the classifiable sample.”
Can the deviated-trade rate be high while the expectancy gap stays small?
Yes. If deviated trades are numerous but close to outcome-neutral in this sample, the deviated-trade rate can be high while the gap between classifiable-sample and aligned expectancy stays small. That still describes a process problem worth reviewing; a small expectancy gap does not mean deviations are safe to continue, only that they have not visibly moved this particular estimate.
How many trades are needed before this comparison is meaningful?
There is no universal trade count. The aligned and deviated subsets each need enough classified, non-unclassified trades that a single trade cannot swing the win rate or average win/loss dramatically. Treat a small-subset expectancy figure, especially the deviated-only figure, as a flag for review rather than a stable number, and widen the window — or run a bootstrap check — before drawing a conclusion from it.
Does removing deviated trades show what I would have earned with perfect execution?
No. The aligned-only expectancy describes the aligned trades that actually occurred; it is not a counterfactual reconstruction of the deviated trades under different execution, and it cannot recover P&L for omitted trades that were never taken. The market opportunities behind the deviated and omitted trades cannot be replayed under aligned conditions, so treat the gap as a diagnostic signal for further review, not a recoverable amount.
Costante provides educational workflow tools, not financial advice. Trading involves risk.