Skip to content

Latest commit

 

History

History
44 lines (32 loc) · 2.86 KB

edge-score.md

File metadata and controls

44 lines (32 loc) · 2.86 KB

The Edge Score (v2)

An open, versioned composite of trading performance from 0 to 100. Closed-source journals sell an opaque score; this one is a pinned formula anyone can read, verify, and recompute from their own data. Source of truth: packages/core/src/edge-score.ts.

Requirements

  • At least 5 closed trades, otherwise the score is withheld (components still computed). Tiny samples produce impressive-looking nonsense; we refuse to print it.

Components

Each component is normalized to a 0 to 100 scale with an explicit "full marks" threshold, then combined with fixed weights:

Component Full marks at Weight Why it's in
Win rate 60% 15 Consistency of outcomes, capped low because win rate alone is gameable
Profit factor 3.0 25 The heart of edge: gross profit per dollar of gross loss
Avg win / avg loss 2.5 : 1 20 Asymmetry: are winners structurally bigger than losers
Drawdown 0% of (initial balance + peak); zero score at ≥25% 15 Risk of ruin. Without an initial balance the % is unknowable → neutral 50
Recovery factor net P&L = 3× max drawdown 10 Does the equity curve earn back what it gives up
Consistency largest winning day ≤ 15% of total day profits 15 One lucky day shouldn't carry the score

Score = Σ(componentᵢ × weightᵢ) / Σ weights, rounded to 2 decimals.

Versioning

The formula is pinned as EDGE_SCORE_VERSION = 2 and every score carries its version. Changing any threshold, weight, or input definition is a breaking change and bumps the version; historical scores stay comparable within a version.

  • v2: gross profit and gross loss (the profit factor inputs) now sum the positive and negative net P&L of every closed trade, regardless of whether a trade is labeled "breakeven" by a tolerance band. Identical to v1 when the breakeven tolerance is zero.
  • v1: initial release.

Honest limitations

  • Thresholds are opinionated calibration points, not statistical truths. They're chosen so that a solidly profitable retail track record scores in the 60 to 80 band.
  • The score measures the record, not the future. A 90 on 6 trades means less than a 65 on 600. That's why the trade count ships alongside the score everywhere it's shown.