I tested 144 trading strategies on BTC, ETH and SOL. Here's what happened 🧪
I built a bot on my phone (Termux) and backtested it on 1 year of hourly data from MEXC, with fees included and 10x leverage.
The setup:
• 70% of the data for training, 30% kept hidden for testing
• EMA crossovers and breakouts, with and without a trend filter
• ATR-based stops and different risk/reward ratios
The results:
• Only 22 of 144 combinations were profitable on the unseen data
• The "best" strategy in training LOST 61% on the test period
• Max drawdowns were over 100% at 10x leverage, which means a wiped account
• Simply holding $BTC, $ETH or $SOL returned about +18-20% in the same period, with no leverage
What I learned:
1. A strategy that looks great on past data often fails on new data (overfitting)
2. High leverage makes most edges disappear
3. Beating the market is much harder than it looks
I'm not trading real money with this. Next I'm testing slower timeframes and lower leverage, and I'll share the results either way.
Not financial advice. Past performance doesn't predict future results.
#BinanceSquare #crypto #Backtesting #TradingBots101 $BTC $ETH
I built a bot on my phone (Termux) and backtested it on 1 year of hourly data from MEXC, with fees included and 10x leverage.
The setup:
• 70% of the data for training, 30% kept hidden for testing
• EMA crossovers and breakouts, with and without a trend filter
• ATR-based stops and different risk/reward ratios
The results:
• Only 22 of 144 combinations were profitable on the unseen data
• The "best" strategy in training LOST 61% on the test period
• Max drawdowns were over 100% at 10x leverage, which means a wiped account
• Simply holding $BTC, $ETH or $SOL returned about +18-20% in the same period, with no leverage
What I learned:
1. A strategy that looks great on past data often fails on new data (overfitting)
2. High leverage makes most edges disappear
3. Beating the market is much harder than it looks
I'm not trading real money with this. Next I'm testing slower timeframes and lower leverage, and I'll share the results either way.
Not financial advice. Past performance doesn't predict future results.
#BinanceSquare #crypto #Backtesting #TradingBots101 $BTC $ETH