#backtest
backtest
3,348 views
18 Discussing
Hot
Latest
💡 Backtesting data doesn't lie:
8 years, 53,392 trades verified👇
🥇 EMA7/25 + RSI → PF=1.67
🥈 Adding 4+ conditions → Win rate 57.1%
🥉 Adding 7+ conditions → Win rate 57.6%
Key points:
❌ Win rate 60% but PF < 1 → Long-term losses
✅ Win rate 46% but PF = 1.67 → Long-term gains
Risk-reward ratio > Win rate
This is where most newbies get it wrong
#量化交易 #Backtest #Crypto
8 years, 53,392 trades verified👇
🥇 EMA7/25 + RSI → PF=1.67
🥈 Adding 4+ conditions → Win rate 57.1%
🥉 Adding 7+ conditions → Win rate 57.6%
Key points:
❌ Win rate 60% but PF < 1 → Long-term losses
✅ Win rate 46% but PF = 1.67 → Long-term gains
Risk-reward ratio > Win rate
This is where most newbies get it wrong
#量化交易 #Backtest #Crypto
I tried testing the optimization of an EMA Crossover for 14 crypto assets, and here are the results.
Systematic dual-EMA crossover backtest (buy when the fast EMA crosses above the slow EMA, sell when it crosses below), fee 0.15% per side, for spot, daily timeframe. Tested on 14 assets: BTC, ETH, BNB, SOL, XMR, DOGE, ADA, XRP, LINK, TRX, XLM, ZEC, BCH, PAXG. Most cover complete history since listing (2000–3200+ candles per asset).
Methodology: grid search EMA fast(5-40) x slow(20-200), 90 combinations per asset, take the highest return, compare versus Buy & Hold.
Key findings:
1. 13 out of 14 assets, EMA crossover outperforms Buy & Hold
2. Optimal parameters are NOT universal. Each asset has different combinations (BTC 10/30, ETH 20/26, SOL 9/26, XMR 12/20, etc.). There is no “one setting for all”
3. Win rate is low (25–50%) but still profitable in a typical trend-following way: a few big trades carry the overall return
4. BNB actually LOSSES versus Buy & Hold. This shows the strategy doesn’t always work; it depends on the character of price movements of each asset
Each crypto asset has a different EMA Crossover optimization. This optimization may change as more price data accumulates and continues to grow for testing materials. This method optimizes the EMA Crossover using historical data that has already happened.
Methodological warning (important, don’t skip):
This is purely in-sample optimization; it hasn’t been walk-forwarded to out-of-sample. Large returns in historical backtests DO NOT guarantee the same performance going forward. Overfitting is a real risk. As a comparison, RSI and MACD were also tested using the identical methodology and consistently lost to EMA on the same data.
This is an independent research for learning, not an invitation or trading signal. Screenshots of the results are attached below.
#EMA
#Backtest
#TradingStrategy
#CryptoResearch
#DYOR
Systematic dual-EMA crossover backtest (buy when the fast EMA crosses above the slow EMA, sell when it crosses below), fee 0.15% per side, for spot, daily timeframe. Tested on 14 assets: BTC, ETH, BNB, SOL, XMR, DOGE, ADA, XRP, LINK, TRX, XLM, ZEC, BCH, PAXG. Most cover complete history since listing (2000–3200+ candles per asset).
Methodology: grid search EMA fast(5-40) x slow(20-200), 90 combinations per asset, take the highest return, compare versus Buy & Hold.
Key findings:
1. 13 out of 14 assets, EMA crossover outperforms Buy & Hold
2. Optimal parameters are NOT universal. Each asset has different combinations (BTC 10/30, ETH 20/26, SOL 9/26, XMR 12/20, etc.). There is no “one setting for all”
3. Win rate is low (25–50%) but still profitable in a typical trend-following way: a few big trades carry the overall return
4. BNB actually LOSSES versus Buy & Hold. This shows the strategy doesn’t always work; it depends on the character of price movements of each asset
Each crypto asset has a different EMA Crossover optimization. This optimization may change as more price data accumulates and continues to grow for testing materials. This method optimizes the EMA Crossover using historical data that has already happened.
Methodological warning (important, don’t skip):
This is purely in-sample optimization; it hasn’t been walk-forwarded to out-of-sample. Large returns in historical backtests DO NOT guarantee the same performance going forward. Overfitting is a real risk. As a comparison, RSI and MACD were also tested using the identical methodology and consistently lost to EMA on the same data.
This is an independent research for learning, not an invitation or trading signal. Screenshots of the results are attached below.
#EMA
#Backtest
#TradingStrategy
#CryptoResearch
#DYOR
Log in to explore more content