Research status — Not yet OOS-validated. Do not use for live trading.
58.1%
break-even 49.4% at its realized average win/loss
re-measured (top 10 coins · 1H · 2026-10-06): 49.68%
Percentage of trades that were profitable
1.42
re-measured (top 10 coins · 1H · 2026-10-06): 0.94
Gross profit / gross loss ratio
as published (577 coins · 1H · 2026-01)
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 is already retired or shelved — a closed verdict outranks its number. No mark = clears it and is still standing.
No live tracking — measured, not guaranteed
Card re-measured: 2026-10-06 · 2,641 trades analyzed
Overview
HV Squeeze uses the same Bollinger Band squeeze detection as BB Squeeze — the bands are computed and used directly. What differs is the filter set: a lower volume threshold (1.5x vs 2.0x) and a candle-color direction filter instead of the EMA trend.
The hypothesis: a lighter filter set (previous candle color instead of EMA trend, a lower volume bar) might catch expansions that BB Squeeze’s stricter filters miss.
How It Works
- Detect Squeeze — BB (20, 2σ) width drops below 0.8x its own 20-period moving average; a squeeze within the last 10 candles qualifies
- Wait for Expansion — BB width increasing (current > previous)
- Volume Filter — previous candle’s volume >= 1.5x its 20-period average
- Direction Filter — previous candle color (bearish candle + close below BB mid = SHORT)
- Enter Short — Same risk management (SL 10%, TP 8%, 48h max hold)
Test Results
| Metric | HV Squeeze | BB Squeeze | Difference |
|---|---|---|---|
| Win Rate | 58.1% | 68.6% | -10.5%p |
| Profit Factor | 1.42 | 2.22 | -36% |
| Signals fired | 12,240 | 2,110 | 5.8x |
| HV signals BB also fires | 14.8% | — | — |
| BB signals HV also fires | 85.9% | — | — |
Corrected 2026-09-21. These two rows used to read “Signal Overlap 85% / Unique Signals 15%”, which had the direction backwards. Re-measured on the top-10 market-cap coins, 1h candles, 2023-12-10 → 2026-07-12: only 14.8% of HV’s signals land on a bar where BB Squeeze also fires the same way, while 85.9% of BB Squeeze’s signals are ones HV also fires. HV is not a subset of BB Squeeze — it is a looser superset that fires 5.8x as often. No tolerance window or direction rule reaches 85% on the HV side (±24 bars, direction ignored: 43.2% at most). Method and per-coin numbers:
backend/research/20260921/hv_bb_signal_overlap.json(reproduce with the.pybeside it).The old line “the 15% of unique HV signals had a win rate of 47%” is withdrawn — its population was the inverted 15%. Now measured (2026-09-22, same scope and the card’s own config —
hv-squeezeregistry defaults, SL 10% / TP 6%, futures costs): the real HV-only set — bars where BB Squeeze does not fire the same way — wins 50.18% across 2196 trades with a profit factor of 0.91. The 218 trades that do overlap BB Squeeze win 55.96% with a profit factor of 1.35.So HV’s extra signals do not add edge, they dilute it: the 9% of trades BB Squeeze also catches carry the profit factor above 1, and the rest drag it below. That is a different reason from “redundant” — and it is the reason the card was shelved. Method, per-coin numbers and the sanity check against the card:
backend/research/20260922/hv_only_subset_performance.json(reproduce with the.pybeside it). Trades were labelled, not filtered — the engine holds one position per coin, so re-running with overlapping signals removed would create trades the card never had.
Why It Was Shelved
- A looser superset, not a subset — HV covers 85.9% of BB Squeeze’s setups and fires 5.8x as often. The redundancy is real, but it runs the other way: almost everything BB catches, HV also catches, plus a large tail of its own.
- Check the latest measurement — see the HV Squeeze live card on the strategies page for current metrics and their measurement scope. The table above is a historical measurement, not the current card’s results.
- Complexity without edge — Adding a second volatility measure doubles code maintenance but adds zero PnL
- BB Squeeze already validated — 2+ years, 535 coins, 2,898 trades. No reason to replace a working system with an equivalent one
Lesson Learned
Not every variation of a working idea is an improvement. Sometimes the first implementation is also the best one.
HV Squeeze taught us that indicator substitution is not optimization. The edge in BB Squeeze comes from the squeeze-to-expansion transition pattern, not from the specific method of measuring volatility. BB and HV measure the same underlying phenomenon — price compression — just with different math.
Potential Revisit
HV Squeeze could become relevant if:
- BB Squeeze edge degrades below profitability threshold
- Market microstructure changes (e.g., new exchange mechanisms alter BB behavior)
- Combining both as an ensemble shows uncorrelated alpha on HV’s own signals — the 85.2% that BB Squeeze does not fire (the old wording said 15%; see the 2026-09-21 correction above)
For now, simplicity wins. One validated signal is better than two redundant ones.
Status: SHELVED
Scope note (corrected 2026-09-25): the numbers on this page (WR 58.1%, PF 1.42) come from the January 2026 Binance-era test on 577 coins. An earlier version of this note said
coins: 235was the OKX universe of a 2026-04-10 re-measure — there was no such re-measure: 235 was the OKX roster count copied in on 2026-05-02. Today’s coin list is on /coins.
Not actively traded. Last tested: January 2026 on 577 coins with 2+ years of data.