How we calculate

FIFO matching

Per pair, each sell closes against the oldest open buys. Each match is a closed lot:

pnl = qty × (exit − entry) − fees

Buy and sell fees are pro-rated by quantity. A sell with no prior buy is recorded as orphan_sell and excluded.

USD value

Stablecoins (USDT/USDC/FDUSD/BUSD) = $1. Other quote assets and fee assets (e.g. BNB) use the public Binance 1-minute kline close (cached). Offline — the amount is marked as an estimate.

Equity proxy E (approximation)

The CSV has no balances, so: for each day in the last 30 days — open positions at cost + "estimated cash" = the largest entry in the window. E = the median. You can enter your account size manually on the scan page.

Minimum data

Fewer than 20 closed lots — every check returns insufficient_data. We never invent findings.

Total Cost of Behavior (TCB)

Sum of detector costs with de-duplication: each closed lot is counted once, in the detector with the highest total cost.

Discipline Score

score = 100 − min(100, 100 × TCB / max(volume × 1%, |sum of losses|))

Rounded to an integer.

Performance metrics

win rate = wins / lots   (pnl > 0 = win)
avg win = Σ pnl⁺ / wins ;  avg loss = Σ pnl⁻ / losses
expectancy = Σ pnl / lots ;  payoff = avg win / |avg loss|
avg hold = mean(exit − entry)
max drawdown = max(peak − equity), equity = Σ pnl by exit time, peak from 0
fees = Σ fees (USD)

Zero losses: payoff = "—". No lots: every metric is "—".

Trend, pairs, compare

A lot belongs to the month / ISO week (Monday start) of its exit time in your time zone. Period score = the same score formula on that period's lots, using their de-duplicated cost. Compare periods = the same metrics over two ranges.

Calculators

position: qty = (account × risk%) / (|entry − stop| + fee% × (entry + stop))
R/R: risk = |e − s| + f(e + s) ; reward = |x − e| − f(e + x)
break-even win rate = risk / (risk + reward)
fee impact = trades × size × fee% × 2

Pure calculators: nothing is stored or sent.

What-if simulator (hindsight)

Replays your lots in time order and skips any entry a rule would have blocked: cooldown X min after a loss, max entries per day, position cap (% of equity), daily stop after losing X% of equity. Stop: a lot's loss is capped at X% + fees — an optimistic bound, since we have no intra-trade prices. Hindsight only, not a forecast.

Failrail rules block

Built only from findings actually detected with cost > 0, sorted by cost. Each suggested value comes from your data (e.g. cap = min(2 × median entry, 10% of equity)). The total = TCB, no double counting.

Share link

A separate snapshot with a new random token (192 bits): read-only, without your journal and without delete. Revoke any time; deleting the report deletes it too.