QuantilanCrypto strategies
How they compare, and how they behave under stress.
Q-Mix, BB, RSI and IMP against Bitcoin and S&P 500 buy-and-hold on one convention, the risk model behind them, and what they did on the 10 October 2025 flash crash. Data to 26 Oct 2026. Back to the cards.
Against buy-and-hold
Sum of the yearly returns 2022 to 2026, each year restarted from $10,000, so strategies and benchmarks stand on the same ground. Strategies are simulated and include fees and slippage; benchmarks are spot Bitcoin and the S&P 500 Total Return index (dividends reinvested).
| Strategy | Σ yearly | Avg / year | Win rate | Worst annual DD | Trades |
|---|---|---|---|---|---|
| IMP · Impulse Retest | +349.9% | +70.0% | 57.8% | −21.4% | 282 |
| BB · BearBull | +340.2% | +68.0% | 62.1% | −17.6% | 5,950 |
| RSI · Advanced RSI | +304.2% | +60.8% | 62.2% | −21.4% | 8,252 |
| Q-Mix | +299.9% | +60.0% | 57.0% | −36.3% | 1,065 |
| Bitcoin, held | +197.5% | +39.5% | — | −66.9% | — |
| S&P 500 TR, held | +63.5% | +12.7% | — | −24.5% | — |
Win rate is weighted by trades over the five years. Worst annual drawdown is the deepest fall inside one calendar year; for the held assets it is the deepest fall from a peak over the whole window. 2026 is partial and, from 1 October, a live test on virtual accounts rather than a backtest. IMP trades about 50 to 66 times a year, so its figures rest on a much smaller sample, and its parameters were chosen by the worst year (not a clean out-of-sample test). How we test.
Year by year
| Strategy | 2022 | 2023 | 2024 | 2025 | 2026 |
|---|---|---|---|---|---|
| IMP · Impulse Retest | +36.9% | +69.8% | +58.7% | +86.3% | +98.2% |
| BB · BearBull | +41.7% | +84.1% | +83.0% | +73.6% | +57.9% |
| RSI · Advanced RSI | +23.6% | +85.4% | +59.4% | +62.3% | +73.5% |
| Q-Mix | +37.1% | +29.6% | +82.5% | +65.6% | +85.1% |
| Bitcoin, held | −64.3% | +155.4% | +121.0% | −6.3% | −8.4% |
| S&P 500 TR, held | −18.1% | +26.3% | +25.0% | +17.9% | +12.4% |
Two ways to hold
The same Bitcoin and S&P 500 give different totals depending on whether the year is reset or the position is simply held. Both are shown, so the comparison is honest.
Sum of the yearly returns. The same convention as the strategies.
Compounded, no reset. The 2022 fall of −64% shrinks the base for the later gains; deepest fall −66.9%.
Sum of the yearly returns on the same convention.
Compounded, no reset; deepest fall −24.5%.
The reset sum is higher than the compounded total for Bitcoin because the sum ignores the compounding drag of the 2022 crash. It describes a book that restarts from the same size every year, which is how the strategy accounts are measured.
Risk model
Every strategy trades with fixed leverage and a fixed margin per position. The loss at the stop is small by design; the exposure grows with the number of open positions.
Fixed, on every coin.
Of the balance per position in Q-Mix, BB and RSI. IMP: 40% / 30% / 10% on BTC / SOL / ETH.
Of the account per position in BB and RSI. Q-Mix: about 2% (5% margin × 5× × the 8% stop). IMP: about 6% / 3.75% / 1.75% per coin (at most 11.5% with all three open).
Per coin: no coin draws on another coin's margin.
| Strategy | Book | Largest cash exposure | Risk at the stops, all positions open |
|---|---|---|---|
| BB | 16 coins | 80% (16 × 5%) | 8–12% at base risk |
| RSI | 18 coins | 90% (18 × 5%) | 9–13.5% at base risk |
| IMP | BTC, SOL, ETH | 80% (40 + 30 + 10%) | up to 11.5% |
| Q-Mix | one shared account | 5% × the coins in the book | about 2% per position; up to 38% with all 19 open |
Adaptive RiskController
BB, RSI and IMP (and IMP on tokenized stocks) run with an adaptive controller that scales the risk of new trades between 0.35× and 1.2× of the base. Q-Mix, Q-RWA and Q-ETF run without it, at a fixed size.
Fast, mid, slow
Tracks the profit factor over the last 30, 100 and 300 trades at once. The weight moves toward the fast window during win streaks.
Cut fast, recover slow
Risk falls quickly (α = 0.3) when the profit factor deteriorates and recovers slowly (α = 0.1). A streak of 3 wins speeds recovery up; 5 makes it as fast as the fall.
Up to −50% risk
The penalty starts at a 10% drawdown and grows linearly to a 50% cut at 30%. It applies on top of the profit-factor multiplier.
0.7× after 3 days
If equity stays below its peak for more than 3 days, a second penalty lowers the risk toward 0.7× over the next 3 days. It resets on a new equity high.
The 10 October 2025 flash crash
Bitcoin lost about 10% within one 15-minute candle and altcoins fell much further. BB and RSI were already short going in, which suited the event and is not proof of resilience under a short squeeze. IMP opened one short on the crash candle itself. No liquidations and no manual intervention in any of them.
| Strategy | Equity before | Trough | After (11 Oct) | Max dip | Note |
|---|---|---|---|---|---|
| BB | $15,250 | $14,556 | $14,734 | −4.55% | 5 positions, short book |
| RSI | $16,130 | $15,692 | $15,909 | −2.71% | about 12 positions, broad short book |
| IMP | $11,840 | $11,840 | $13,468 | 0.00% | 1 trade, +$1,628 booked on 17 Oct |
| Q-Mix | $18,377 | $18,283 | $18,961 | −0.51% | closed-trade equity, vector run |
2025 backtest on a shared $10,000 account (equity includes nine months of that year's compounding), entry slippage 0.05% and stop/take slippage 0.03%. Trough is the lowest booked equity from 10 Oct 00:00 to 11 Oct 21:30 UTC. Equity is booked at closed trades, so an open position's drawdown shows only when it closes. BB, RSI and IMP come from the sequential engine, Q-Mix from the vector engine.
- BB: ETH, SOL and SUI shorts with take-profit ladders at +3% and +6% into the drop; a SEI short and one LINK long were stopped small. By the end of 11 Oct equity had not returned above its level before the event.
- RSI: a broad short book (APT, ARB, AVAX, LINK, PEPE, SOL, XRP) with ladders filled at +3.5% to +8%; a few counter-trend longs stopped small. The RiskController kept per-trade risk at its base multiplier.
- IMP: one BTC short opened on the 21:15 UTC candle at 112,082, stop at +3% never reached; both take levels filled inside the same candle (105,949 and 103,146), +$1,628 realised (+13.8% of the account), booked on the curve on 17 Oct. The model fills inside a 10% wick, and live fills in such a candle may differ.
- Q-Mix: closed-trade equity stayed within about 0.5% of its pre-event level and ended 11 Oct above it.