RSI mean-reversion on ETH 4h

RSI mean-reversion on ETH 4h

We tested this simple trading idea on Quant Agent, freqedge's research loop, for that start by defining the idea in plain english:

• Buy ETH when 4-hour RSI drops into oversold territory (below 30).

• Only take the trade if volume is stronger than its recent average.

• Sell when RSI returns toward normal (above 55).

What Quant Agent did

Four iterations, exploring different configurations (of RSI entry or exit level, periods, etc). Same wallet, same costs. Hold-out locked — the run was never allowed to fit on it.

Train: 28 Jul 2024 – 24 Apr 2025 Hold-out: 24 Apr 2025 – 23 Jul 2025 Eligibility: ≥30 trades. Below that, treat the result as noise.

Why it was killed before paper

The strategy did not almost work. It never produced enough trades to be eligible. A 65% win rate on 20 trades looks comforting. It is still a small sample. We stop rather than stack OBV, extra volume rules, or tighter exits until a 20-trade curve looks pretty. That is how most retail strategies die: more complexity, same scarce signal.

What this teaches (crypto or stocks)

Count trades before you read profit. A filter that never binds is not a filter. Widening a weak signal is not research — one extra trade and a worse P&L means you stretched noise. A locked hold-out with 2 trades cannot save the idea. Next step is has to be new spec (more pairs, different timeframe, less extreme threshold on a universe), not another oscillator on the same 20 events.

What we did not claim

Not paper. Not live. Not a return. Figures are simulated on one trainning window plus a locked hold-out. The Agent’s sentence is the product: no tradeable improvement found — nothing beat the eligibility bar.

Kill the idea before it spends the account.

Any figure shown in a teardown is backtested or simulated and is not indicative of future performance