Capture efficiency is a ratio, and ratios have blind spots the grade never mentions.

The denominator problem

When MFE is small, the ratio stops describing management and starts describing arithmetic.

A trade that reached 0.15R at its best and closed at 0.10R records a capture efficiency of 67%, which sounds respectable and means almost nothing — the numbers are close together because they are both close to zero, and a single tick of difference would have moved the percentage by twenty points. The same 67% on a trade that reached 2.4R describes a real management outcome worth studying. The ratio is stable and informative in the upper part of the MFE range and increasingly noisy as opportunity shrinks toward the entry. This is why the lab's quadrant map bins by MFE magnitude before judging capture, and why an aggregate capture figure computed across a week that contained many low-opportunity trades is being pulled around by trades that never had anything to capture.

The numerator floor

MAX(0, Outcome R) means every losing trade scores identically, and losers are half the population.

A trade that reached 1.8R and closed at −1R scores zero capture. So does a trade that never went favourable at all and closed at −1R. So does every other loss in the sample. That is the correct behaviour for the formula — negative capture is not a meaningful quantity, and allowing it would let a bad loss drag the aggregate below zero and produce a percentage nobody could interpret. But it means capture efficiency is structurally silent about roughly half of what happened. The trade that offered 1.8R and lost is one of the most expensive events in the log, and it is precisely the event capture cannot describe.

Which formulas cover the gap

Giveback and adverse utilisation carry the losing half, which is why the family has six members.

Giveback R is MAX(0, MFE R − Outcome R), and on that 1.8R trade that closed at −1R it prints 2.8R. It is the formula that sees the disaster capture is floored past, and it is denominated in R rather than percent so it aggregates honestly across trades of different opportunity. Adverse utilisation, the absolute MAE as a share of the stop, describes how much of the risk allowance was consumed before the outcome arrived — the entry-quality half of the story. Fee R drag and net outcome R handle friction. Read as a set, the six formulas cover the whole trade path; read individually, each has a region it is blind to, and capture's region is the largest.

• Capture grades the winners' management; giveback grades every trade, winners and losers alike.
• Giveback is in R, so it aggregates across trades; capture is a percentage and does not.
• A high-giveback, high-capture week is possible and means the losses are where the leak is.

Why they are still worth having together

Two formulas built on the same two inputs disagree usefully because they normalise differently.

It is fair to ask why capture and giveback both exist when both are computed from MFE and outcome. The answer is that one is scaled and one is absolute, and the two orderings differ. A trade that offered 1.1R and kept 0.9R has excellent capture and trivial giveback. A trade that offered 4R and kept 2.8R has mediocre capture and enormous giveback. Ranked by capture the first trade wins; ranked by giveback the second is the one costing real expectancy. Neither ranking is wrong, and an operator who only watches the percentage will systematically under-prioritise the large trades where the R at stake is greatest.

The recording standard

Ratio formulas amplify input sloppiness, and the amplification is worst where it matters.

MAE and MFE are logged as R multiples at the actual extremes, unrounded. The reason is mechanical rather than pedantic. A rounding error in the denominator propagates through the ratio with a magnitude that grows as the denominator shrinks — the same region the previous sections identified as already noisy. Half an R of optimistic rounding on the favourable side, applied habitually, produces a capture aggregate that is wrong in a consistent direction and a repair engine confidently ranking fixes for a problem that exists only in the log. The formulas are honest instruments fed by a manual process, and the process is the part that can drift.

The key idea

Know what each formula is silent about, or the silence gets read as a passing grade.

The failure mode this brief is guarding against is small and common: an operator reads 71% capture, decides exits are healthy, and stops. The 71% was computed over the winning trades only, weighted toward the small ones, on a week whose real damage was three trades that offered 2R and closed red. Every one of those facts is available in the other five formulas. The verdict on a week's execution is a six-number verdict, and any single member of the family read alone will eventually flatter something it was never designed to see.