Bivar Capital
Bivar Capital — Quantitative Research

Ethereum Trend StrategySystematic trend-following for Ethereum — beta

A rules-based trend-following system for Ethereum, derived from a parameter grid search on ETH's own history. Invested 42% of the time. 9.2 years of Binance data (Aug 2017–Oct 2026). No look-ahead bias — but see the caveats below before treating this as validated.

Beta status. Unlike the BTC signal, these parameters (RSI>52, MA30, no volatility filter) come from a single in-sample/out-of-sample grid search on ETH alone — 294 configurations tested, only 4 survived both the fit window (2020–2023) and the holdout window (2024–2026) at a comparable bar. This one did. That is weak-to-moderate evidence, not a validated edge. No adversarial audit has been run on this strategy yet.

Performance Summary

+51.7%
CAGR
1.02
Sharpe
-60.0%
Max Drawdown
42%
Time Invested
1.02
Sortino
56.0%
Volatility
52%
Win Rate
16
Trades / Year

Equity Curve

Growth of $1 (Log Scale) — Strategy vs Buy & Hold
$455,273

Strategy: $10,000 invested → invested only 42% of the time

$80,185

Buy & Hold: $10,000 invested → MaxDD −94.0%

Drawdown Comparison

Underwater Equity — Strategy vs Buy & Hold

Strategy max drawdown: −60.0%. Buy & hold: −94.0%. The strategy cuts the drawdown roughly in half — but −60% is still severe, a direct consequence of dropping the volatility filter that the BTC signal uses.

Strategy vs Buy & Hold

MetricStrategyBuy & HoldAdvantage
CAGR+51.7%+25.5%+26.2pp
Max Drawdown-60.0%-94.0%+34.0pp better
Volatility56.0%83.5%−27.5pp
Sharpe1.020.70+0.32
Time Invested42%100%In cash 58% of the time
$10,000 →$455,273$80,1855.7x more

The strategy beats buy & hold on every risk-adjusted metric, and even on raw CAGR — not just by avoiding crashes but by sidestepping the 2018 and 2022 bear markets entirely while still catching most of the 2020 and 2024 rallies.

Annual Returns

YearStrategyBuy & HoldAlpha
2017*+117.7%+965.5%-847.8pp
2018-6.2%-82.1%+75.9pp
2019+20.0%-1.7%+21.7pp
2020+257.2%+467.5%-210.3pp
2021+137.0%+399.2%-262.2pp
2022+4.3%-67.5%+71.8pp
2023+17.5%+90.8%-73.3pp
2024+52.1%+46.1%+6.0pp
2025+52.7%-11.0%+63.7pp
2026**+32.3%-22.9%+55.2pp

* 2017 partial (Aug–Dec, MA30 still warming up). ** 2026 year-to-date through Oct 14. Strategy underperforms buy & hold in the biggest bull years (2017, 2020, 2021) — that is the cost of being out 58% of the time — but wins decisively in every bear year (2018, 2022, 2025).

How These Parameters Were Found

The BTC signal's rule (RSI>54, MA155, vol<100%) applied as-is to ETH underperforms: Sharpe 0.89 vs 1.02 for this version. A 294-configuration grid (RSI threshold 48–60, MA length 30–200, vol filter 0.6–unfiltered) was split into a fit window (2020–2023) and a holdout window (2024–2026). Configurations that ranked in the top 30% of both windows simultaneously — not just one — were kept. Only 4 of 294 qualified; RSI>52 / MA30 / no vol filter had the best holdout Sharpe (1.13) among them on the fit split, and re-estimated on the full history gives the numbers on this page.

Why no volatility filter: every version with the BTC-style vol<100% filter scored lower in both windows. ETH's high-volatility periods, it turns out, are disproportionately where this signal's edge lives — cutting them out removes the best trades along with the worst ones.

Caveats

Single-asset optimization. The grid search was run on ETH only. The same "MA30, no vol filter" shape has not been confirmed on other assets outside this BTC/ETH/SOL set — treat it as an ETH-specific finding until shown otherwise.

Short history for this asset class. 9.2 years, one full market cycle and change. Crypto regimes shift fast.

Deep drawdown. -60% is a real number investors need to be able to sit through, not just read.

Costs. ~16 round trips/yr × 0.12% (6bps each way) = modest drag, included in the backtest, but real-world slippage on exchange execution can run higher.

Cash drag. In cash 58% of the time at 0% yield in this backtest.

No adversarial audit yet. Every other live/beta strategy on this site went through a hostile second-pass review before publishing. This one has not, yet — it is published here for transparency, not as a recommendation.