---
type: research
title: Stop-loss overlay for BTC funding mean reversion
date: 2026-11-14
status: completed
---

# Stop-loss overlay — TL;DR

**Question.** The funding mean-reversion alpha has no price stop: it holds 48h
and exits on the clock. Does adding a stop-loss help, and should it be scanned
every minute?

**Method.** Entries are held fixed from the baseline signal. For each trade we
compare the 48h time exit against a stop at a fixed % (3/5/8/10/15%) or an ATR
multiple (3/6/12/24 ATR). Stops trigger on Bybit-linear **1-minute** lows/highs,
fill at the stop price or the 1m open on a gap, plus 5 bps slippage. Metrics are
geometric, from a 1m PnL path aggregated to hourly.

**Answer.**

| Forward (2024-11→2026-11) | CAGR | Sharpe | MDD | Worst trade |
|---|---:|---:|---:|---:|
| No stop (current) | 46.2% | 1.48 | −26.6% | −8.8% |
| Fixed 3% | 35.6% | 1.30 | −19.1% | −3.1% |
| Fixed 5% | 49.0% | 1.58 | −23.5% | −5.1% |
| Fixed 8% | 46.6% | 1.49 | −26.5% | −8.1% |
| 3 ATR | 34.4% | 1.33 | −16.9% | −4.4% |

- **Tight stops hurt.** 3% and 3 ATR cut forward CAGR by 10–12 points; 53–63%
  of stops were premature (the trade would have recovered by the 48h exit).
- **A wide catastrophic stop is defensible.** 8% fixed is nearly neutral on
  CAGR/Sharpe/MDD but caps the worst trade at −8.1% instead of −8.8%. 5% caps
  it at −5.1% and looks better forward, but the forward window is contaminated.
- **Scanning every minute is not the main issue.** The rule and the level
  matter. A tight stop just sells the dip the strategy is trying to buy.
- **The stop does not fix the drawdown.** MDD stays −32% to −39% in the
  backtest; the drawdown is a sequence of trades, not one trade. Position
  sizing / vol targeting is the first line of defense.

**Recommendation.** Add an **8% fixed catastrophic stop** if a hard loss cap is
required (5% for a tighter tail). Do not use a tight stop. Re-run the official
strategy with the chosen stop and paper trade; do not optimize further on the
forward window.

**Implemented.** The official strategy now runs 1h entries + a 5m 8% stop. The
updated report shows back Sharpe 1.52 / forward 1.52, with the worst bar capped
at −5.1% / −3.9% (from −15.9% / −5.0% without the stop). See
`brain/strategy/2026-11-14_btc_funding_mean_reversion/summary.md`.

Full report: [report.md](report.md). Reproduce: `src/study.py`.
