Bias & Robustness Testing
We ran our EMA 13/80 strategy through 6 independent bias tests. Here are the results — including the ones that failed.
Most platforms only show you the backtest. We show you everything that could be wrong with it.
❌Walk-Forward Validation
4/9 windows profitableTrain on 6 months, test on the next 2 months, roll forward. Tests if the edge holds across different time periods.
What this means: The strategy performed well in 2024 trending markets but struggled in late 2025 choppy conditions. This is expected for a trend-following strategy — it needs directional moves to make money.
✅Monte Carlo Simulation
100% of 500 shuffles profitableShuffle the order of all 27 trades 500 times. If most shuffles are still profitable, the edge isn't dependent on getting lucky with trade sequence.
What this means: The edge is not sequence-dependent. No matter what order the trades execute in, the strategy remains profitable. This is a strong signal that the returns are real, not lucky.
❌Slippage Sensitivity
Breaks even at ~0.15% slippageWhat happens if real slippage is worse than our 0.1% assumption? Tests from 0% to 0.5% slippage.
What this means: The strategy survives our assumed slippage but has only a 50% safety margin. If real-world slippage is consistently above 0.15%, the edge disappears. This is why we forward test — to measure actual slippage.
❌Fee Sensitivity
Only profitable on Binance (0.1%)Same strategy tested across different exchange fee structures.
What this means: This strategy only works on low-fee exchanges. Binance is viable. Kraken spot and Coinbase will destroy the edge. Kraken Futures (0.02% maker) would likely be the best option.
✅Multi-Coin Generalization
7/10 coins profitableDoes the strategy work on coins it wasn't specifically optimized for?
What this means: The strategy generalizes well across assets. 7/10 coins are profitable, including strong performance on mid-cap alts. This makes it unlikely that the edge is curve-fitted to a single asset.
❌Data Snooping Check
74% chance of finding a "winner" by luck aloneWe tested 26 strategy variants and picked the best one. What are the odds it's just lucky?
What this means: Testing 26 variants creates a high risk of data snooping — finding something that looks good by chance. However, the cross-coin generalization and Monte Carlo results suggest the core edge (trend-following with EMA) is real. The optimization (V2) adds marginal improvement that may or may not be real. This is why we forward test both V1 and V2.
The honest bottom line
EMA 13/80 has a real edge in trending markets, confirmed by Monte Carlo and multi-coin generalization. But it struggles in choppy conditions, has tight slippage margins, and only works on low-fee exchanges. The V2 optimization shows promise but hasn't been validated long enough to trust.
This is exactly why we forward test. Backtests tell you what might work. Forward tests tell you what actually works. 12 strategies are running live on 48 coins across 6 timeframes on Binance right now.
Not financial advice. Past performance does not guarantee future results. These tests use historical data and may not reflect future market conditions.