PerpForge
Get started

Family dossier · EMA Cross

EMA Cross: 26 variants, every one tested.

PerpForge is a perpetual-futures strategy simulator. We ran the EMA Cross (a trend-following signal: a fast moving average crossing a slow one) across SOL, ETH, BTC, on multiple timeframes, at multiple leverage rungs (leverage = trading with borrowed money to size a position larger than your balance), long and short. Winners and busts published together. The verdict in one sentence, the evidence, then all 26, ranked by edge significance (whether the profit edge is real rather than luck).

The finding

A moving-average crossover family tested across BTC, ETH, and SOL. Ranked by statistical significance, not by the biggest dollar number.

How many actually worked

3%
of BTC variants beat buy-and-hold
2%
of ETH variants beat buy-and-hold
3%
of SOL variants beat buy-and-hold
205/210
lose to buy-and-hold

Of 186 variants with enough trades to judge (≥30 trades), only 5 beat holding. Profitable means it made money over the window. Beating buy-and-hold is a higher bar, and proven (a real edge, not luck) is higher still.



↑ The five that hold up

Top 5 by edge significance — the confidence-adjusted measure of whether the profit edge is real rather than luck. Each beat buy-and-hold.

#StrategyVerdictvs Buy & HoldWorst drop
1
EMA Cross 21/50
SOL·1h·long
Real edge-2013.4%-9.9%
2
EMA Cross 12/26
SOL·1h·long
Inconclusive-2036.6%-12.8%
3
EMA Cross 21/50
SOL·4h·long
Inconclusive-1926.1%-9.9%
4
EMA Cross 21/50
ETH·1h·long
Inconclusive-366.3%-8.6%
5
EMA Cross 21/50
BTC·4h·long
Inconclusive-502.2%-6.6%

↓ The five worst outcomes

We publish the busts next to the winners. These are the proof we are not selling you anything: bankrupt (the account wiped out) or down more than half the starting balance.

#StrategyVerdictvs Buy & HoldWorst drop
26
EMA Cross 9/21
Heavy lossBTC·1m·long
No edge-632.0%-98.0%
25
EMA Cross 9/21
Heavy lossETH·1m·long
No edge-485.7%-98.0%
24
EMA Cross 9/21
Heavy lossBTC·5m·long
No edge-632.6%-98.0%
23
EMA Cross 9/21
Heavy lossETH·5m·long
No edge-487.1%-98.0%
22
EMA Cross 9/21
Heavy lossSOL·1m·long
No edge-2242.0%-98.0%

All 26 variants · by market

How to read this table. Within each market, rows are ranked by edge significance: whether the profit edge is real rather than luck, not by the return. So a small but proven edge outranks a big-but-unproven one, and a robust loser can outrank a lucky winner. vs Buy & Hold is the return above simply holding the coin; Worst drop is the worst peak-to-trough loss. Inconclusive and busted results sit at the bottom of each market. Publishing every loser is how we keep that honest.
StrategyVerdictvs Buy & HoldWorst drop
SOL-PERP2 of 64 beat buy-and-hold
EMA Cross 21/50
1h·long
Real edge-2013.4%-9.9%
EMA Cross 12/26
30m·long
No edge-2071.0%-25.8%
EMA Cross 12/26
15m·long
No edge-2149.2%-56.6%
EMA Cross 9/21
Heavy loss15m·long
No edge-2164.8%-70.2%
⋯ 4 more SOL variants ⋯
EMA Cross 50/200
4h·long
Inconclusive-1825.0%-5.8%
ETH-PERP1 of 64 beat buy-and-hold
EMA Cross 12/26
30m·long
No edge-406.7%-28.0%
EMA Cross 12/26
Heavy loss15m·long
No edge-447.2%-58.7%
EMA Cross 9/21
Heavy loss15m·long
No edge-462.0%-73.5%
EMA Cross 9/21
Heavy loss5m·long
No edge-487.1%-98.0%
⋯ 4 more ETH variants ⋯
EMA Cross 50/200
4h·long
Inconclusive-352.0%-7.9%
BTC-PERP2 of 58 beat buy-and-hold
EMA Cross 12/26
1h·long
No edge-540.7%-19.3%
EMA Cross 12/26
30m·long
No edge-566.4%-35.4%
EMA Cross 12/26
Heavy loss15m·long
No edge-601.0%-67.6%
EMA Cross 9/21
Heavy loss15m·long
No edge-612.1%-77.6%
⋯ 3 more BTC variants ⋯
EMA Cross 21/50
1h·long
Inconclusive-527.1%-11.2%
A curated set per market — the strongest plus the notable busts. All strategy families in-sample · Binance · Jan 2021 – Jun 2026 · flat fees, no funding or slippage · as of Jun 2026

What this engine models, and what it doesn't

Trading fees — Yes. Flat per-side cost charged on entry and exit (the fee on each leg of a trade).
Liquidation — Yes. Force-close when losses eat through the margin (the collateral backing a leveraged position).
Funding rates — No. The recurring payment between longs and shorts that tethers a perp (a futures contract with no expiry) to spot. Not modeled, so real results would be slightly worse.
Slippage & maker/taker — No. Candle-price fills, one flat fee. No price drift on entry, no separate maker (resting-order) vs taker (market-order) fee.

The next experiment

The fast variants bled out in chop. Does a volatility filter save them?

We never tested these crossovers with a filter that sits out sideways markets. Pre-fill the Lab with one and find out — your pair, your timeframe, the same engine these 26 ran on.

Run your own idea Free account, no card