Experiment 3 detail

Exit Optimization, Round 1

First bounded exits-only comparison while holding entries and regime source fixed.

Experiment 1 status
Untouched
Operating paper branch remains the default
Experiment 2 status
Paused
Detector work is not the active path here
Best Step 1 exit variant
pct_tp_7_sl_3p5
First bounded percentage winner
Return vs baseline
19.66%
Baseline 86.59%
Isolation logic

What stayed fixed

Bounded search

Search grid

VariantTake profitStop loss
pct_tp_3_sl_23.00%2.00%
pct_tp_5_sl_25.00%2.00%
pct_tp_7_sl_27.00%2.00%
pct_tp_3_sl_3p53.00%3.50%
pct_tp_5_sl_3p55.00%3.50%
pct_tp_7_sl_3p57.00%3.50%

Keep the grid small because this is Step 1, meant to answer whether exits matter at all before a larger sweep.

Use targets near the existing 24h trigger scale: 3%, 5%, and 7% take-profit levels bracket a modest, medium, and stretch mean-reversion payoff.

Use two stop widths, 2% and 3.5%, to test whether tighter protection or slightly wider breathing room works better without exploding combinations.

This yields 6 percentage variants plus the current fixed-hold baseline: enough to learn shape, not enough to overfit by brute force.

Core evidence

Comparison results

VariantTotal returnDelta vs baselineLatest equityMax drawdownWin rateProfit factorTradesAvg bars held
pct_tp_7_sl_3p5106.25%19.66%2.0625x-3.67%50.88%1.845723.63
baseline_fixed_hold_1686.59%0.00%1.8659x-4.19%58.93%2.415616.00
pct_tp_5_sl_3p556.41%-30.18%1.5641x-6.99%51.56%1.486416.09
pct_tp_7_sl_245.06%-41.53%1.4506x-4.46%32.86%1.417016.33
pct_tp_5_sl_236.95%-49.64%1.3695x-6.18%38.96%1.347710.16
pct_tp_3_sl_3p531.96%-54.63%1.3196x-6.62%62.03%1.307910.13
pct_tp_3_sl_25.33%-81.27%1.0533x-10.34%46.32%1.07956.80
Verdict

Read