One rule changed. Both versions ran on the same 2,976 completed Binance BTCUSDT perpetual candles from January 2025. Change the threshold to inspect a precomputed replay.
Change one rule
15-minute candles. Enter on the next bar open, with a stop two ATR (14) from entry and a 2R target. Risk 1% of simulated equity per trade, with one open position at a time.
Result for RSI 30
Historical simulation. Monetary figures are in USDT.
RSI 30 replay returned 3.70% after costs across 20 closed trades.
Measure
Original · RSI 30
Your choice · RSI 30
Closed trades
20
20
P&L before modelled costs
5,885.72
5,885.72
Trading costs
2,186.41
2,186.41
Net P&L
3,699.30
3,699.30
Net return
3.70%
3.70%
Max drawdown
4.23%
4.23%
Inspect the costs and one trade
Total fees 1,462.14 · slippage 533.33 · funding 190.95 USDT.
Signal 2025-01-20 17:45 UTC. Entered at 102,964.09 on 2025-01-20 18:00 UTC and exited at 99,312.52 on 2025-01-27 06:30 UTC (stop).
Trade P&L before costs -973.12 · fees 27.25 · slippage 10.90 · funding 53.38 · net -1,064.66 USDT.
Pre-cost P&L is reconstructed from this run’s trade ledger by adding modelled costs back to net P&L.
Keep your version
Create a free account to reopen this exact RSI setting and its result after email verification and profile setup.
At RSI 35, the bot closed 35 trades against 20 at RSI 30. Modelled costs absorbed 93.58% of the RSI 35 version’s pre-cost profit. Its net return was 0.27%, compared with 3.70% at RSI 30. Inspect the trade and cost breakdown above to see where the difference came from.
All settings were inspected on one month and one market. This replay does not establish how the rule would perform on unseen data, and its fills and funding are simulated.
Reproduce the test
Use the Binance archive from 2025-01-01 00:00 UTC up to 2025-02-01 00:00 UTC (exclusive). Start flat with 100,000.00 USDT. The Edge Theory v4 engine uses 0.05% taker fees per side, 0.02% for maker targets, 2 bps of adverse slippage for market fills and conservative funding of 0.01% per eight hours when a position spans a settlement. Stops take priority if the stop and target share a candle. Open positions close at the end of the sample.
We’re inviting 5–10 builders to try the workflow with their own strategy and show us where it slows them down. A high-level description is enough to start.
