Whipsaw

A false signal: you enter on what looks like a breakout, price immediately reverses, and you exit at a small loss. Repeated whipsaws are how trend-following strategies bleed in choppy markets.

In plain English

Imagine the fast EMA crosses above the slow EMA. You buy. Within the next few candles, price wobbles back down, the lines cross again in the opposite direction, and you sell — losing a small amount on slippage and fees. That round trip is one whipsaw. In a sideways market, this can happen ten times in a week.

Whipsaws are the natural enemy of ema cross and other trend following systems. Each one is a small loss, but they accumulate fast.

Why it matters for this fleet

Whipsaw damage in this fleet is symbol- and regime-dependent, not rule-dependent. The exact same crossover rule chops one symbol to pieces while it runs cleanly on another. The variable is the symbol's character over the window — how trendy versus how sideways it traded — not the strategy's logic. Whipsaw also gets dramatically worse on faster candles, where most signals are noise.

How to detect whipsaw damage in metrics

• Profit factor near 1.0 (e.g. 0.95–1.05) with many trades is a whipsaw signature: the strategy is finding signals, but they cancel out.
• High max loss streak (e.g. maxLossStreak > 15) on a short interval suggests the market regime was hostile to the signal.
• Low Sharpe with non-trivial PnL means returns are noisy — likely whipsaw-driven gains and losses cancelling unevenly.

Examples from the live fleet

Take one rule — EMA 21/50 · 1h · 2× · long — and run it across the three symbols. Profit factor (gross profit divided by gross loss; above 1 means the wins outweigh the losses) rises as the symbol gets cleaner to trade:

• BTC (id511): PF 1.06 — the most chopped of the three. The wins barely outpace the losses; a profit factor this close to 1 with hundreds of trades is the classic whipsaw signature.
• ETH (id517): PF 1.20 — cleaner.
• SOL (id523): PF 1.46 — the cleanest of the three, and the only one of these three to clear the edge bar.

Same signal, same window, three different fates. The variable is the symbol's character, not the rule.

Death by whipsaw on fast candles

Now change only the EMA pair and the speed. The 9/21 scalp pair (fast, tight EMAs on quick candles) is 0% profitable across the fleet — not a single configuration of it makes money — with a median profit factor of 0.49. A PF of 0.49 means the strategy loses about two dollars for every one it makes. On fast candles almost every cross is noise: price flickers across the EMAs, fires a signal, then immediately reverses. That is whipsaw at its purest — the strategy is paying fees to trade noise.

Caveat: this dataset has no take-profit/stop-loss and no trend-filter variants, so it cannot show whipsaw being reduced by those defenses — only the raw, unfiltered damage.

How strategies fight whipsaws

• Higher timeframes (1h → 4h → 1d) — slower EMAs filter out short-term noise.
• Trend filters (e.g. ADX > 25) — refuse to trade when no clear trend exists. (The dossier #1 baseline does not include trend-filtered variants, so this defense is described here but not yet measured on this data.)
• Wider EMAs (e.g. EMA-20 vs EMA-200 instead of EMA-10 vs EMA-50) — fewer crosses, fewer false signals, but later entries.

Related

• ema cross
• trend following
• market regime
• noise to trend ratio
• profit factor

Sources

• wiki/qa-sessions/2026-05-17-session.md#q1 (first asked here)
• /api/analytics