RAER against RAPF: the two questions that can disagree.

The two metrics

Different denominators, different truths.

RAER — the risk-adjusted efficiency ratio — measures efficiency quality relative to risk deployment: cumulative net P&L divided by cumulative risk deployed, answering whether the capital put at risk is being converted into return. RAPF — the risk-adjusted profit factor — measures profit extraction quality relative to risk, answering whether the profit being generated is structurally sound or the product of aggression and variance. Both are risk-adjusted, both are R-based, and both read the same trades — but they interrogate different properties, which is precisely why they can diverge. A system can convert risk efficiently while extracting profit poorly, and it can extract impressive profit while wasting enormous risk doing it.

The four states

Two metrics, two directions, four meanings.

The engine's core detection resolves to four conditions, each with a distinct diagnosis. RAER up with RAPF down is under-monetization: clean execution, poor profit extraction — the system is doing the hard part right and leaving money on the table. RAPF up with RAER down is profit without efficiency: over-aggressive risk and variance-driven gains — the dangerous state, because it looks like success in every P&L-shaped report. Both down is dual weakness, straightforward system degradation. Both up is alignment, the institutional-grade condition where efficiency and monetization improve together. The asymmetry worth noting: only one of these four states feels bad while happening, and it isn't the most dangerous one.

The dangerous quadrant

Profit without efficiency is the one that hides.

Every other state announces itself somewhere an operator will look. Dual weakness shows up in P&L. Under-monetization shows up as frustration — the trades work and the account doesn't move. Alignment shows up as everything being fine. But profit without efficiency produces good returns while quietly wasting risk, and no P&L-based instrument can distinguish it from genuine edge: the equity curve rises, the win rate holds, the week reads green. What's actually happening is that returns are being purchased with risk the system didn't need to spend — the same failure the alpha definition's five-part test catches as over-risk, detected here at the structural layer before the benchmark comparison gets a chance to. It's the state most likely to be celebrated while it compounds.

• Under-monetization is a capture problem — check exits, trail behavior, and giveback before questioning the edge.
• Profit without efficiency is a sizing and aggression problem — the edge may be fine; the risk spent per unit of it isn't.
• Dual weakness routes straight to the edge-change evidence stack; it's the only state that questions the edge itself.

Why cumulative

The engine reads structural values, not weekly noise.

The comparison runs on the cumulative structural values from each pipeline's Table 1 — and the construction rules matter more here than almost anywhere else in the system. Cumulative RAER is cumulative net P&L divided by cumulative risk deployed, never an average of weekly RAER ratios, because averaging ratios distorts capital-weighted efficiency and double-weights light weeks against heavy ones. The same discipline applies to composites: composite RAER uses composite P&L over composite risk, not a blend of primitive ratios. Get this wrong and the contradiction engine detects contradictions that exist only in the arithmetic — which is why the SDE's QA checklist lists ratio-averaging as its own named failure mode.

The key idea

One metric can lie comfortably; two metrics have to agree.

The engine's premise is that structural truth is easier to detect in tension than in isolation. Efficiency alone can look fine while nothing monetizes; profit alone can look excellent while risk hemorrhages. Holding both against each other means the system has to be consistently healthy to appear healthy — and any state where the two disagree is, by construction, a state where something is happening that a single-metric read would have missed entirely.