Systematic strategy · crypto
Shock Engine
Systematic post-shock continuation for Bitcoin and Ethereum: it enters after an abnormal move that keeps its direction, and manages risk with volatility.
Historical simulation after modeled transaction costs. Not live performance.
Performance
Two markets, one engine
Half of the capital in each market, each half compounding on its own, from 2018-09-01 to 2026-09-30. Commission of 0.045 % on every order.
Equity, indexed to 100
Log scale · daily closes
Dashed line: buy & hold BTC/ETH 50/50 over the same period (half in each coin on 2018-09-01, no rebalancing, no costs): 100 → 1,073, CAGR 34.1%, max drawdown −76.9%, Sharpe 0.78, Sortino 1.11.
How it works
Four steps, computed at every 15-minute close
Closed bars only, no look-ahead. The rules were set on Bitcoin and run unchanged on Ethereum.
Shock
Detects an abnormal directional move: a 15-minute return far outside its recent distribution, breaking the recent range.
Trend
Confirms that the move is aligned with the higher-timeframe structure, with a decisive candle and active volume.
Regime
New positions are only opened when daily volatility is calm, and short positions only when the daily trend regime is bearish.
Risk
One position per market, no leverage. The stop scales with volatility; an opposite shock closes the trade.
Proprietary thresholds are not published. The engine trades about 44 times a year on BTC and 44 on ETH, with a median holding time of 13 hours, and is in the market 25% of the time.
Research evidence
We tried to break it
Each test has a pass/fail criterion fixed before the result is seen. Single-market figures use 15-minute returns over each market's full test period.
Not reproduced by simple trend following
Once buy & hold and six standard trend strategies (momentum, breakouts, moving averages, an intraday breakout) are removed, 81% of the average return remains unexplained. Pre-registered test, on the historical period the rules were chosen on.
The Bitcoin parameters, applied to Ethereum without any ETH calibration, met every pre-set criterion; an independent ETH price feed gives the same verdict (Sharpe 1.82).
Share of 200 runs with identical exits and random entry times that the strategy beats. The timing of entries carries the edge, not the exits or the market's drift.
Share of the average trade gain kept when every entry is taken one 15-minute bar late. The result does not depend on a perfect fill.
Doubling commissions (equivalent to 0.045 % slippage per fill) keeps the portfolio Sharpe at 1.88; with historical perpetual funding it is 1.97.
All parameters perturbed together by up to ±20 %, 300 neighbors per level: median Sharpe 1.35 on BTC and 1.94 on ETH. On ETH the parameters sit mid-neighborhood, a plateau rather than a peak.
On Bitcoin, settings re-selected every quarter on the previous 36 months only, then traded on the next 3: 28 windows since 2020, CAGR 14.3 %. Positive, below the fixed rules. With longer windows (12 designs fixed in advance): median 0.97, range 0.83–1.36.
Risk
Drawdowns, stress and stability
The same weight as the returns: how deep, how long, and how stable the results have been.
Drawdown from the previous peak
Daily closes
Rolling 12-month Sharpe ratio
365-day windows of daily returns
Named market stress episodes
| Episode | Dates | Portfolio | BTC spot | ETH spot |
|---|---|---|---|---|
| 2018 Q4 capitulation | 2018-11-08 → 2018-12-15 | 41.7% | −51.1% | −61.8% |
| COVID-19 crash | 2020-03-01 → 2020-03-31 | −0.3% | −24.7% | −38.9% |
| May–July 2021 sell-off | 2021-05-10 → 2021-07-20 | 9.0% | −48.9% | −54.5% |
| Terra/LUNA and 3AC | 2022-05-05 → 2022-06-30 | 30.2% | −49.8% | −63.6% |
| FTX collapse | 2022-11-06 → 2022-11-30 | 0.5% | −19.4% | −20.4% |
| August 2024 carry unwind | 2024-08-01 → 2024-08-10 | 1.2% | −5.7% | −19.3% |
| Oct–Nov 2025 deleveraging | 2025-10-06 → 2025-11-30 | 7.6% | −26.8% | −33.7% |
Episode windows were fixed from known market events, never from strategy results.
Portfolio
Why BTC and ETH together
The two coins move together. The two strategies much less: each is out of the market most of the time, and their trades rarely overlap.
Daily correlation, 2018-09-01 → 2026-09-30.
Research status
What is established, and what is not yet
Stated plainly, so it can be weighed.
- Historical simulation2018-09-01 → 2026-09-30, reproduced from versioned code (40/40 checks)
- Modeled transaction costscommission 0.045 % per order; cost stress published
- BTC / ETH evidenceevent study, placebo entries, delayed entries, two ETH data sources
- Robustness testingparameter neighborhoods, cost stress, resampled histories
- Forward validationshort-entry condition pre-registered, observation from 2026-10-12; calm-regime restriction not yet
- Testnet botplay money, since 2026-10-09; does not yet apply the calm-regime restriction · live page
- Live track recordnot available
- Capacity / market impactexecution costs measured on Binance for all 1,150 executions · dossier; Hyperliquid and own impact need live trading
Slippage and perpetual funding are not in the headline figures.
Research
Explore the research
Historical simulation after modeled transaction costs. Not live performance.
Important information. All performance figures on this site are hypothetical: they come from a historical simulation of fixed rules, after modeled commissions, and do not represent actual trading. Simulated results have inherent limitations. They benefit from hindsight in the design of the rules, they do not reflect slippage, funding, market impact or the liquidity of a real account, and past simulated performance does not predict future results.
Nothing on this site is investment advice, a recommendation, or an offer or solicitation to buy or sell any financial instrument or to provide investment management services.