Equity mode (Compounding)

Position size = current account equity × position size percent. Wins grow future positions; losses shrink them. Models realistic account dynamics but contaminates signal measurement with capital-scaling effects.

In plain English

When equity mode is on, each trade uses a percentage of whatever your account balance is right now. Win a trade → next trade is a little bigger. Lose a trade → next trade is a little smaller. This is how real trading accounts work when you size by percentage of capital.

The equity curve you see reflects two things mixed together:

1. Whether the signal was actually any good (did it predict direction?)
2. Whether the wins happened to cluster when the account was large or small

You can't cleanly separate these two effects in equity mode.

Formula / mechanic

positionSizeUsd = currentEquity × (positionSizePct / 100)

currentEquity = initialEquity + net P&L from all closed trades so far

Tracked per-strategy by EquityLedger (apps/backend/src/evaluation/equity-ledger.ts). Updated on every trade close. Bankruptcy detected when equity falls below the configured floor.

UI label: Compounding. Config key: equityMode: true (default).

What question it answers

"What would happen to my actual account if I ran this strategy at this leverage?"

It simulates real account dynamics: compounding wins, shrinking losses, eventual bankruptcy if the drawdown is deep enough.

Known distortions (read before trusting verdicts)

1. Left-tail truncation (the leverage mirage) When a strategy goes bankrupt, the sim stops. The worst losing streak never fully enters the Sharpe calculation — the distribution is truncated on the left. This inflates the per-trade Sharpe and can produce a false "Real edge" verdict. See leverage mirage.

Documented in CLAUDE.md rung 6: "Forced liquidations truncate losing trades → clip left tail → inflate Sharpe → false Real edge. Re-run at 2× to verify."

2. Path-dependence Early wins → larger future positions → later wins amplified. Two strategies with identical signal quality but different win/loss sequencing will rank differently in equity mode. The leaderboard's Sharpe comparison is confounded by when in the shared window wins happen. See path dependence.

Consistency check: if a strategy shows "Real edge" in equity mode but "Inconclusive" in flat mode, the equity verdict is suspect — likely a leverage mirage or path-dependence artifact. Run flat mode as the sanity check.

Why it's the default

More intuitive for users ("my $10k account will look like this"). Shows realistic survival risk including bankruptcy. Reveals what real capital management pressure looks like during a drawdown.

Related

• flat mode — the signal-quality measurement alternative
• leverage mirage — the specific truncation-inflation failure mode
• path dependence — why equity mode Sharpes are confounded
• sharpe ratio
• drawdown
• bankruptcy

Sources

• apps/backend/src/simulation/config-manager.ts:241 (resolvePositionSize)
• apps/backend/src/evaluation/equity-ledger.ts (EquityLedger)
• apps/backend/src/evaluation/evaluator.ts:1879–1883 (equity update on close)
• wiki/qa-sessions/2026-07-01-session.md#q1