EMA Cross

A trend-following entry signal that fires when a fast Exponential Moving Average crosses a slow one.

In plain English

You pick two moving averages — one short (e.g. EMA-10, averaging the last ~10 candles) and one long (e.g. EMA-50). When the short one crosses above the long one, the recent price action is faster than the longer-term price action, which suggests an emerging uptrend. You enter long. When it crosses back down, you exit (or short, depending on the strategy mode).

It is the canonical "trend-confirmation" signal in technical analysis — simple, transparent, and used by millions of retail traders.

Formula / mechanic

For each new candle:

• Compute EMA_fast and EMA_slow.
• If EMA_fast was below EMA_slow on the previous candle and is now above → bullish cross (entry for long).
• If EMA_fast was above EMA_slow previously and is now below → bearish cross (exit for long; entry for short).

EMA formula: EMA_t = α × price_t + (1 − α) × EMA_(t−1) where α = 2 / (period + 1).

Why it matters for this fleet

The entire 210-strategy EMA-cross dossier is built on this one signal, varying side (long vs short), interval (the candle duration), leverage (a position-size multiplier that scales both profit and loss), and an optional volume gate. It is the foundation of the analysis. But the signal is mostly weak: only 5 of 210 rows clear the edge-significance bar (the statistical test for "this is more than luck"), and they collapse to just 3 distinct signals — all "21/50 crossover going long".

Examples from the live fleet

• id523 (EMA 21/50 · SOL · 1h · 2× · long): win rate 31.9% (the share of trades closing in profit), profit factor 1.46 (gross wins ÷ gross losses), Sharpe 0.110 (mean per-trade return ÷ its volatility), Wilson confidence interval ±4.4pp, edge SIGNIFICANT. The only signal family in the whole dossier with a real in-sample edge — winners average ~3.1× the size of losers, so it makes money despite losing about two trades in three.
• id511 (EMA 21/50 · BTC · 1h · 2× · long) — the famous near-miss. Its win rate is tightly pinned (24.9%, Wilson interval ±3.9pp — a precise measurement), yet the edge is NOT significant: to prove an edge this thin you would need roughly N ≥ 9,213 trades and it has only 469. Same row, opposite answers to the two confidence questions — the win rate is trustworthy, the edge claim is not.

Related

• trend following
• whipsaw
• market regime
• noise to trend ratio
• profit factor
• sharpe ratio

Sources

• wiki/qa-sessions/2026-05-17-session.md#q1 (first asked here)
• /api/analytics for fleet metrics
• wiki/2026-05-17-ema-cross-symbol-breakdown.md