SURVIVAL OF THE FITTEST
Real markets. Simulated money. A paired strategy trial.
CONTROLLED STRATEGY TRIAL
THE FULL POPULATION
| Bot & strategy | Equity today | 7-day score | Evaluation | Drawdown today | Position | Current price | Stop loss | Take profit |
|---|
Score = mean daily return − ½ mean daily maximum drawdown − ½ daily return variability, in percentage points. Higher is better. Uses qualifying completed days within the last seven calendar days; today’s P&L does not affect the score. Protected bots have provisional scores. Open any bot for full position details.
THE DAILY RACE
Liquidation value in GBP · £1,000 starting line · sampled every 5 minutes
THE PLAYING FIELD
Completed 5-minute candles · USD · last 8 hours
EVERY MOVE COUNTS
| Time (UK) | Bot | Action | Market | Fill price | Costs | Realised P&L | Exit / entry reason |
|---|
SELECTION PRESSURE
THE FAMILY TREE
NO BOT FORGOTTEN
THE EXPERIMENT
At a healthy UK midnight on or after 7 October 2026, six existing bots become unchanged controls. Six revised copies inherit their entry settings, holding times and allocation caps. The other six bots are archived for trial setup, with their history preserved. Controls are chosen across represented strategy families in identity order, not by profitability.
Revised bots require a fresh signal crossing on a completed candle. A volatility filter skips moves too small relative to costs. Stops and targets use recent volatility, and include all trading costs. A trailing stop activates after a net gain. Planned stop loss is capped at 1% of equity by reducing allocation if needed; gaps can exceed that amount.
Breeding pauses throughout the trial. Review follows at least 14 qualifying days, with no automatic promotion of a winner. A qualifying day needs at least 90% of minute checks and timely settlement. Compare returns after costs, drawdown, trading activity and matched pairs; doing nothing is not a demonstrated trading edge.
Every bot receives £1,000 each UK day. Positions settle at midnight. Both groups use the same public prices and costs, one position per bot and no leverage. The revised package tests entries, exits and risk sizing together; the comparison cannot isolate which individual change helped.
Orders use the quoted bid/ask with 0.05% additional slippage and a fixed 0.26% fee per side. Short positions also accrue a simulated 0.02% daily borrowing charge. These are experiment assumptions, not a claim about an exchange’s current fee schedule. Equity includes estimated closing costs; shorts use cash collateral and paper liquidation floors equity at zero.
Stops are evaluated once per minute, so a fast move can overshoot them. If any required price or signal data is missing, new trades pause. A delayed daily settlement uses the last available quote and is flagged in history. Missed full days during a service outage are not invented or backfilled as trades.
Trade and chart detail is retained for 45 days. Bot identities, parentage and daily results remain stored beyond that. Historical candles warm up the strategies; the dashboard never presents invented bot performance before launch.
All money is simulated. There are no exchange credentials, deposits or real order endpoints.
Data: Kraken public prices and historical candles.