# Is the crypto profit-lock trail worth the hassle? (2026-09-28)

Avg % result per trade with no trail vs the live setting vs the best of 22 arm/trail pairs, everything else at live values. "trail decides" = share of trades the trail actually closed under the live setting. crypto-tape-v1 is absent: its entry needs the live order book, which bars don't carry.

| strategy | bars | trades | no trail | live setting | best pair | trail decides | worth it? |
| --- | --- | --- | --- | --- | --- | --- | --- |
| crypto-trend-v1 | daily | 316 | +2.82% | arm 5% / trail 5%: +0.23% | arm 25% / trail 15%: +2.01% | 51% of exits | No — a trail hurts here |

## crypto-trend-v1 — daily bars, 730 days, 18 coins, hold cap none

**Worth it? No — a trail hurts here.** no-trail beats even the best trailing pair (arm 25% / trail 15%) by +0.81% per trade; the live pair (arm 5% / trail 5%) is -2.59 vs no-trail.

**What decides the exits** with the live setting (arm 5% / trail 5%): 22% the strategy's own signal, 24% the ATR stop, 51% the trail, 0% the hold cap, 4% still open at the end.

**Whole period (730d):** no trail +2.82% per trade · live +0.23% · best trailing pair arm 25% / trail 15% +2.01%. **Last 183 days:** no trail +2.34% · live +2.08% · best arm 20% / trail 10% +2.70%.

**Robust pairs** (positive in every ~183-day window AND without their 3 best coins): none.

**Pairs that matter** — "avg / trade" is the average % result per trade; "exits by trail" the share of trades the trail closed; "windows +" how many of the 4 ~183-day windows were positive.

| pair | trades | win | avg / trade | profit factor | avg hold | exits by trail | windows + | avg / trade ex top-3 coins |
| --- | --- | --- | --- | --- | --- | --- | --- | --- |
| arm 5% / trail 5% — live now | 316 | 50% | +0.23% | 1.05 | 6.6d | 51% | 3/4 | -0.66% |
| no trail | 230 | 37% | +2.82% | 1.55 | 13.1d | 0% | 3/4 | +0.72% |
| arm 25% / trail 15% — best trailing pair, whole period | 236 | 37% | +2.01% | 1.39 | 12.2d | 11% | 3/4 | +0.07% |
| arm 20% / trail 10% — best last 183d | 249 | 37% | +1.07% | 1.20 | 10.9d | 20% | 3/4 | -0.25% |

**Every pair — avg % per trade, whole 730 days.** Down a column = arming later at one trail width; across a row = a wider trail at one arm level.

|  | trail 5% | trail 8% | trail 10% | trail 12% | trail 15% |
| --- | --- | --- | --- | --- | --- |
| arm 8% | +0.34% | +0.24% | · | · | · |
| arm 10% | +0.38% | +0.25% | +0.47% | · | · |
| arm 15% | +0.59% | +0.57% | +0.78% | +0.39% | +1.34% |
| arm 20% | +0.52% | +0.63% | +1.07% | +0.74% | +1.38% |
| arm 25% | +0.73% | +0.59% | +1.19% | +0.95% | +2.01% |

**Every pair — avg % per trade, last 183 days only.**

|  | trail 5% | trail 8% | trail 10% | trail 12% | trail 15% |
| --- | --- | --- | --- | --- | --- |
| arm 8% | +2.04% | +1.38% | · | · | · |
| arm 10% | +2.40% | +1.71% | +1.87% | · | · |
| arm 15% | +2.04% | +1.60% | +2.37% | +2.01% | +1.94% |
| arm 20% | +2.26% | +1.93% | +2.70% | +2.31% | +2.15% |
| arm 25% | +2.64% | +2.14% | +2.64% | +2.37% | +2.11% |

## Caveats

- Signal-level: every qualifying signal on every coin is taken at a fixed $1,000 — no slots, pool or cash limits — so this isolates the exit rule, not how an account would have compounded.
- Exits are checked in the live order: the resting stop fires inside the bar it is touched (filled at its price — pessimistic when a bar spans both a new high and the stop); the strategy's own exit, the hold cap and the trail update happen at each bar's close, so the trail ratchets once per bar like the real poll.
- No fees; a live commission shaves every pair equally and does not reorder them.
- Coins: BTC/USD, ETH/USD, DOGE/USD, SOL/USD, LINK/USD, XRP/USD, AAVE/USD, ADA/USD, ARB/USD, AVAX/USD, BCH/USD, DOT/USD, FIL/USD, LTC/USD, ONDO/USD, PEPE/USD, RENDER/USD, UNI/USD — the ones the crypto strategies have actually traded plus the shared watchlist. Live entries come from screeners, so a fixed list is already a simplification.
- In-sample sweep: the single best cell is the least trustworthy number here. Trust the size of the spread, the no-trail comparison, and the robustness line — not the winner.

_Re-run: `pnpm backtest:crypto-trail [--strategy=<name>] [--windows=4]`. Raw per-pair stats in the .json next to this file._