Does the Signal Travel?The BTC production signal applied to 10 other coins — plus 753 coin-specific challengers, correlation analysis, and the out-of-sample verdict.
September 2026 · Net of 10bps/order · Data: Binance daily klines per coin since listing · Companion to the BTC Strategy — Adversarial Validation Report.
01 — The production signal on 11 coins
| Coin | Since | Signal CAGR / Sharpe / MaxDD | B&H CAGR / Sharpe / MaxDD | Verdict |
|---|---|---|---|---|
| BTC | 2018 | 47.4% / 1.35 / -32% | 25.9% / 0.69 / -77% | FIRST DIVISION |
| BNB | 2018 | 35.9% / 1.03 / -35% | 67.2% / 1.03 / -76% | ties Sharpe, halves DD |
| ETH | 2018 | 31.3% / 0.90 / -51% | 20.0% / 0.64 / -86% | works, second division |
| SOL | 2021 | 29.7% / 0.86 / -46% | 25.9% / 0.73 / -96% | works, second division |
| AVAX | 2021 | 22.5% / 0.81 / -31% | -2.2% / 0.47 / -96% | clear win |
| ADA | 2018 | 22.5% / 0.75 / -60% | 24.5% / 0.70 / -95% | marginal |
| DOGE | 2019 | 24.1% / 0.73 / -60% | 79.5% / 0.82 / -92% | loses — pumps are events, not trends |
| TRX | 2018 | 25.5% / 0.66 / -38% | 43.2% / 0.83 / -78% | loses |
| XRP | 2018 | 6.8% / 0.35 / -46% | 23.0% / 0.68 / -83% | loses on both axes |
| LTC | 2017 | 5.3% / 0.32 / -61% | 1.0% / 0.45 / -89% | both bad |
| LINK | 2019 | -11.6% / -0.14 / -84% | 25.3% / 0.74 / -90% | loses money in a coin that went up |
What travels is the protection, not the alpha. Max drawdown falls from -76/-96% to -30/-60% on every single coin — the trend + volatility filter is a universal brake. But excess return only exists where trends are slow and persistent. On event-driven coins (DOGE, XRP, LINK) the signal is systematically late to vertical rallies, and the crashes outrun the MA155.
02 — Coin-specific optimization: the trap
Next, each of the four best alts got the full treatment: the 336-config RSI/MA/vol grid plus EMA crossovers, Donchian breakouts and dip-buying systems — 753 challengers in total. In-sample, the results looked spectacular:
| Coin | In-sample champion | Sharpe (full sample) |
|---|---|---|
| SOL | RSI7>58 · SMA200 · no vol cap | 1.61 |
| AVAX | RSI7>58 · EMA100 · vol<1.2 | 1.25 |
| BNB | RSI21>45 · SMA100 · no vol cap | 1.24 |
| ETH | RSI7>50 · SMA155 · vol<0.8 | 1.13 |
An equal-weight portfolio of these five "dominant" strategies backtested at 44.8% CAGR, Sharpe 1.69 — comfortably above BTC alone. Every number in that sentence is a mirage, and the next section proves it.
03 — The out-of-sample verdict
Honest protocol: re-select each coin's champion using only data through December 2023, then evaluate blind on 2024-2026. The production BTC signal — which was not re-fitted — is the control.
| Leg | Train Sharpe (≤2023) | OOS Sharpe (2024-26) | Collapse |
|---|---|---|---|
| SOL — Donchian 55/20 | 2.21 | -0.23 | -110% · loses money |
| AVAX — RSI7>58 EMA155 | 1.77 | 0.18 | -90% |
| ETH — RSI14>50 EMA155 v<0.8 | 1.33 | 0.39 | -71% |
| BNB — Donchian 20/20 | 1.60 | 0.58 | -64% |
| BNB — EMA10x20 (best survivor overall) | 1.58 | 0.78 | -51% |
| BTC — production signal (not re-fitted) | — | 1.37 | HOLDS |
Every freshly-fitted champion collapsed 51-110% out of sample. The strategy that held is the one that was not optimized for the test window — because it has survived four full market cycles instead of being curve-fit to three years of one.
04 — Correlation & portfolio construction
Strategy-return correlations with the BTC signal are genuinely moderate — BNB +0.37, SOL +0.40, ETH +0.50, AVAX +0.28 — which is exactly why the portfolio question deserved a real answer. Diversification only helps if the legs carry edge:
| Portfolio (out-of-sample 2024-26) | CAGR | Sharpe | MaxDD |
|---|---|---|---|
| BTC-only (production) | 38.3% | 1.37 | -15.0% |
| BTC 50% + 4 optimized alts 12.5% each | 25.6% | 1.14 | -22.7% |
| BTC 55% + 3 multi-family winners 15% | 23.2% | 0.99 | -20.0% |
| Equal-weight 5 | 17.8% | 0.83 | -29.3% |
Every blend is worse than BTC alone on every metric. Moderate correlation cannot rescue legs whose edge evaporates out of sample — that is dilution, not diversification.
Verdict
The signal does not travel — and nothing else found in 10 other coins deserves capital. The production signal's edge is specific to Bitcoin's slow, persistent trend regimes. Applied elsewhere it degrades to second-division results at best; coin-specific optimization produces spectacular backtests that collapse 51-110% the moment they are evaluated blind.
The book stays BTC-only. This is not a preference — it is the unanimous answer of three independent tests: one config across 11 coins, per-coin parameter sweeps, and multi-family strategy iteration, all settled out of sample. One honest corollary for BTC itself: a machine that earns 0.7-1.0 "away from home" argues for anchoring live expectations at Sharpe ≈1.0-1.3, consistent with the 3-year-block analysis in the companion report.
Methodology: signals on closed daily candles, t+1 execution, 10bps/order on every position change, no leverage. Warm-up periods excluded per coin. Train/test split at 2023-12-31 with champions selected strictly on training data. Correlations computed on daily strategy returns, common window.