The forensic framework

We don't just measure the backtest. We reconstruct what happened inside it.

A conventional report tells you net profit, win rate, drawdown, Maximum Adverse Excursion (MAE) and Maximum Favorable Excursion (MFE). BackTest Forensics asks what those numbers were doing through time, in what sequence, under which market conditions, and whether the explanation survives a counter-test.

That is the difference between measuring a backtest and investigating one.
Trade-path reconstruction

One Maximum Adverse Excursion (MAE) number can hide an entire story.

Maximum Adverse Excursion (MAE) is the largest move against a position while the trade is open. It is only the starting point. The useful information is in the path: when the heat arrived, how long it lasted, whether the trade recovered, what happened next, and whether that behavior repeated under the same conditions.

Maximum Adverse Excursion (MAE)

How did the trade behave while it was wrong?

Measure maximum heat, time underwater, time to Maximum Adverse Excursion (MAE), repeated adverse waves, recovery time, and whether the worst adverse move occurred before or after favorable excursion. Then compare that path across winners, losers, direction, session, volatility and market state.

Maximum Favorable Excursion (MFE)

How much opportunity did the strategy create versus monetize?

Maximum Favorable Excursion (MFE) is the largest move in the trade's favor while it is open. Measure time to MFE, realized-P&L-to-MFE capture, post-MFE giveback, and whether the exit occurred during a normal retracement or after genuine deterioration. Two +$80 winners can be radically different if one reached only +$110 and the other reached +$500 first.

SEQUENCE

What happened first?

Maximum Adverse Excursion (MAE) → Maximum Favorable Excursion (MFE) is different from MFE → MAE. A trade that suffers early, recovers and trends is not the same setup as one that moves immediately in favor and later gives everything back. Sequence can expose entry quality, stop pressure and exit inefficiency.

A real forensic question

“What stop would improve the curve?” is usually the wrong first question.

The optimizer asks which threshold improves historical results. Forensics first asks what kind of trades the threshold would remove—and what valuable trades it would accidentally destroy.

Example from SATS research

Weak long trades absorbed roughly twice as much adverse movement as favorable opportunity.

In one SATS study, losing long trades averaged about 107 points of Maximum Adverse Excursion (MAE) against only about 54 points of Maximum Favorable Excursion (MFE). In plain English, those losing longs moved almost twice as far against the position as they ever moved in its favor. That was more useful than simply knowing the longs lost money. It suggested a recurring trade path: substantial heat with limited opportunity.

Did maximum heat arrive immediately after entry or later?
How long did those trades remain underwater?
Which losing longs recovered after deep Maximum Adverse Excursion (MAE)?
Would a tighter stop also eliminate the rare large long winners?
Did the weak path concentrate in one market state?
Did aligned longs behave differently from conflict longs?
The next test is not “optimize the stop.” It is: what was different about the trades that immediately went underwater, and can that condition be identified without destroying the long trades that later became major winners?
Maximum Favorable Excursion (MFE) as exit forensics

Profit alone cannot tell you whether the exit was good.

A backtester can tell you a trade made $80. We ask how much opportunity existed before that $80 was realized, how long it existed, and how much was surrendered.

Capture efficiencyWhat percentage of available Maximum Favorable Excursion (MFE) became realized P&L?
Time to opportunityDid Maximum Favorable Excursion (MFE) arrive quickly, slowly, or only after enduring large Maximum Adverse Excursion (MAE) first?
Post-MFE givebackHow much open profit disappeared before the strategy exited?
Exit-state dependencyDoes capture efficiency change under trend, chop, volatility, session or other market-state conditions?
Example: two trades can both finish +$80. If one reached +$110 of Maximum Favorable Excursion (MFE) and another reached +$500, a conventional report treats them as identical winners. Forensics sees roughly $30 of unrealized opportunity in one and about $420 in the other—and asks why.
Loss clustering

Are the losses random—or are they repeating the same failure path?

We group trade paths, not just losing outcomes. The goal is to find whether the same adverse behavior keeps appearing under related conditions.

DirectionDo longs and shorts fail differently? Does one side spend more time underwater or recover less often?
Session & timingDoes early-session Maximum Adverse Excursion (MAE) behave differently from midday MAE? Are late failures mostly giveback rather than bad entries?
Volatility & regimeDoes the same nominal setup produce different excursion paths under expansion, compression, trend or chop?
Sequence & drawdownDo losses cluster after prior losses, during drawdown phases, or after a measurable deterioration in trade-path behavior?
Counter-test: before removing a losing condition, determine whether the same condition also produces a meaningful share of the strategy's largest winners. The right response may be risk scaling, different exits, or no change at all—not a hard filter.
Market-state attribution

Did the strategy change, did the market change, or did both change together?

Trade-path evidence becomes more useful when attached to the environment surrounding the trade. Where data supports it, EdgeQuery IQ compares excursion behavior with trend, volatility, session, macro and structural market-state context.

Market behavior changed

The same entry logic may encounter more adverse excursion, less follow-through or more giveback because the surrounding environment changed.

Trade behavior changed

If Maximum Adverse Excursion (MAE), Maximum Favorable Excursion (MFE), recovery time or capture efficiency deteriorates even inside comparable market states, the evidence points back toward strategy-side behavior.

Both changed together

A strategy can keep firing the same nominal setup while its trade path changes because the environment changed. We test the interaction rather than forcing a one-cause story.

From metric to investigation

Measure → reconstruct → compare → explain → challenge → validate.

The output is not “here is the parameter that maximized the old equity curve.” It is a chain of evidence that produces a testable explanation and then tries to disprove it.

01 · RECONSTRUCT

Rebuild the trade path.

Maximum Adverse Excursion (MAE), Maximum Favorable Excursion (MFE), time underwater, recovery, sequencing, giveback, duration and realized outcome.

02 · COMPARE

Find what changes with the path.

Direction, time, volatility, regime, market structure, macro context and other available states.

03 · CHALLENGE

Try to break the explanation.

Use matched controls, chronology, holdouts and out-of-sample checks where the dataset permits them.

Optimizer: “Filtering this condition improves the historical curve.”
Forensics: “This condition appears to amplify a specific failure path, but it does not fully explain it. What survives after controlling for the obvious confounders?”