Battery record 2026-09-22; the 2026-09-25 review applied stricter criteria. Read both dated records before using these figures.
43.62%
break-even 40.9% at its realized average win/loss
re-measured (top 10 coins · 1H · 2026-11-20): 48.45%
Percentage of trades that were profitable
1.12
re-measured (top 10 coins · 1H · 2026-11-20): 1.28
Gross profit / gross loss ratio
as published (top 50 coins · 4H · 2026-08-23)
Different runs: page record 2026-08-23, 4H, top 50 coins selected for that run, PF 1.12, 917 trades; preset-card record 2026-11-20, 1H, top 10 coins selected for that run, PF 1.28, 5x. The page button uses the page timeframe and coin count. Do not compare these as the same run; use each dated record’s cost settings.
Marks next to a metric: ~ = borderline (PF between break-even and 1.5; win rate ≤5 pts above break-even, or 50–55% if none shown) · ! = below the bar, or the strategy's research status (dated on its page) is below the bar — that dated record outranks this number. No mark = clears it and its research status is not below the bar.
The dated profile on this card reports bull/bear windows and half-year walk-forward losses. Read those measurements separately; an aggregate PF does not remove losses in a smaller window.
Card re-measured: 2026-11-20 · 966 trades analyzed
2026-08-23 re-run at the card’s button settings (4h, top 50 by today’s market cap, sl 15 / tp 30, 5x, fee 0.05% + site slippage tiers): PF 1.12, win rate 43.6%, 917 trades.
미해결로 남긴 것: 아래 2026-07-14 검증 노트는 top-10 조건인데 frontmatter 는
coins: 50이다. top-10·4H 로 재면 PF 1.4351(n=229)로 또 다르다. 「버튼이 어느 유니버스를 시험해야 하는가」는 별개 결정이라 이번 변경에 안 넣었다.2026-09-26 resolution (Q-F315): the Simulate button runs
top_n=50· 4H (live request capture) — the same universe as the card metrics above. The 2026-07-14 top-10 measurements below describe that historical run. Read them alongside the separate 2026-09-25 concentration and later-window check.
2026-09-25 concentration and later-window check. The 2026-07-14 pass below was measured on the top 10 coins by market cap selected for that historical run, with SL/TP chosen on data up to 2026-07-12. Under stricter criteria added that day (point-in-time coin list, out-of-sample only after the parameter choice, coin concentration, cost stress), the evidence no longer holds: without ZEC and HYPE — two of those ten coins — the full-period PF at 0.05% fee plus 0.20% slippage per fill falls from 1.17 to 0.970, and the window after the SL/TP choice (2026-07-15 → 2026-09-24) has only 79 trades. The further-evidence criteria recorded that day require that a point-in-time top-10 re-measure clears the stated tests, the post-07-12 sample reaches n ≥ 100 with a weekly-cluster CI lower bound above 1, and the PF CI lower bound stays above 1 with any single coin removed. Record: verification index.
2026-07-14 measurement, card defaults (SL 15 / TP 30, top-10). OOS(30%) PF 1.36 (263 trades, IS PF 1.24); 3-window walk-forward PF 1.13 / 1.42 / 1.27 (220 / 228 / 319 trades). These are the recorded results of that historical run.
Overview
The Keltner Squeeze LONG strategy is the directional complement to the existing Keltner Squeeze SHORT. It uses the same squeeze detection mechanism — Bollinger Bands compressing inside Keltner Channels — but entries fire when price breaks upward rather than downward.
Found 2026-05-04 in a sweep over two bull windows (IS 2024-05..11: PF 1.63; OOS 2024-11..2025-05: PF 2.96; 4h, 50 coins). Over 2024-02 to 2026-05 with a month-by-month top-50 list, the same settings earned nothing even before costs (our 2026-09-25 test).
How It Works
- Squeeze detection — Bollinger Bands (20-period, 2σ) compress inside the Keltner Channel (20-period, 1.5×ATR). This measures a volatility contraction state.
- Breakout condition — Price closes above the upper Keltner Channel on the squeeze-release bar (previous bar in squeeze, current bar out) — upward momentum breaking through resistance
- Entry — LONG on the next bar’s open after breakout confirmation
- Exit — TP 30% / SL 15% / 48-bar (8-day on 4H) timeout
Why It Works (Thesis)
The Keltner Squeeze is a volatility compression pattern. When Bollinger Bands go inside Keltner Channels, the market has entered an unusually low-volatility state — a coiled spring. The breakout direction reveals which way the tension releases.
The rule tests a hypothesis: buy a confirmed upward breakout after volatility compression. The available results do not measure how often that breakout continues for several ATRs.
The 4H timeframe reduces false breakouts relative to 1H, at the cost of a low signal rate: about one signal per coin every 50 days in the measured bull window (367 trades / 50 coins / 1 year).
Results (Bull-regime IS/OOS validation, measured 2026-05-04)
| Metric | IS (May24–Nov24) | OOS (Nov24–May25) | Combined |
|---|---|---|---|
| Total trades | 188 | 179 | 367 |
| Win rate | ~57% | ~57% | 56.9% |
| Profit factor | 1.63 | 2.96 | 2.28 |
| Coins profitable | 32/50 | 37/50 | — |
2026-05-04 record: PF 1.63 for May–Nov 2024 and 2.96 for Nov 2024–May 2025, a ratio of about 1.82. Both windows were examined in the same sweep; one pair does not establish improvement in unseen bull conditions.
Both measured IS and OOS windows were bull markets: May–Nov 2024 (BTC +150%) and Nov 2024–May 2025 (BTC +120%). The 2026-07-14 OOS/walk-forward record above does not characterize bear-market behavior. The separate 2026-09-25 evidence check is recorded at the top.
Regime conditions
The May 2026 table describes two selected bull windows, May 2024 to May 2025. It does not establish a bull-only advantage or performance outside those windows. The separate 2026-09-25 replay covers 2024-02 to 2026-05 with a monthly point-in-time top-50 list; its gross mean was non-positive.
Caveats
- The IS and OOS windows above were bull markets; bear-market behavior has not been characterized by those measurements.
- 367 total trades across 50 coins over 1 year — thin sample; diversify across coins.
- Wide SL (15%) and TP (30%) parameters suit volatile post-breakout moves but require larger position tolerance.
- 4H signals fire about once per coin every 50 days in active bull conditions.
- Not live-tracked on OKX. Backtest only.