Successful BTC Trading Bots & Algorithms

Overview

This research analyzes 10 algorithmic BTC strategies based on academic papers, institutional studies and open-source projects. Goal: identify robust performance data, assess evidence quality, and derive concrete improvements for Botty.

Important: Most published performance figures are based on backtests. Live-trading data with independent verification is rare. We therefore explicitly rate the evidence quality (1-5 stars) of each strategy.

Methodological assessment

Evidence hierarchy

Common error sources in BTC backtests

1. Lookahead bias: parameters optimized on the full dataset (no walk-forward)
2. Survivorship bias: only currently working strategies get published
3. Bull-run inflation: 2017, 2020-2021 and 2024 make almost any long strategy look good
4. Fee ignorance: taker fees (0.04-0.07%) halve performance at high frequency

Why trend-following works structurally for BTC

BTC behaves like a momentum asset - more pronounced than traditional equities or commodities. Several independent academic papers (2015-2025) confirm this consistently. Three consequences:

1. Trend strategies (EMA crossover, Donchian breakout, MACD) have a structurally positive expected value
2. Mean reversion (classic RSI oversold/overbought) works worse - BTC consistently overshoots too far for reversal plays
3. The biggest source of return is bear-market avoidance, not entry-timing optimization

This insight is not trivial: many retail traders optimize entry signals even though the actual leverage lies in risk management and in recognizing trend regimes.

The single strongest improvement signal

The QuantPedia study on the multi-timeframe MACD (7-year backtest, Dec 2018 - Nov 2025) shows:

The daily MACD filter triples the Sharpe ratio, halves the max drawdown and reduces the trade count by ~56%. This is a consistent, methodically solid result over 7 years and several market regimes.

For Botty this means: MACD_CROSSOVER + daily trend filter is the improvement with the strongest evidence grade and the lowest implementation effort.

What does not work - important negative findings

Funding rate arbitrage (carry trade)

Until 2022 it was a nearly risk-free carry trade with Sharpe > 6. Since 2024 the strategy has been negative - the window is closed. Do not implement.

Multi-pair community bots (NFI, Freqtrade community)

NostalgiaForInfinity has 2,900+ GitHub stars and is one of the best-known open-source bots. The independent aggregator strat.ninja, however, shows Stalled - Negative over the full period. Popularity is not a proxy for profitability. The design principle (high win rate via a very tight stop -0.99%) produces rare but large losses.

Bollinger Band mean reversion (solo)

Without a trend filter and without volume confirmation, pure BB mean-reversion strategies are not profitable on BTC. Only in combination with trend and volume filters (the ETH Zurich approach) do robust results emerge.

Direct Botty action plan (by priority)

1. Daily MACD filter for MACD_CROSSOVER ★★★★

# In strategies/macd_crossover.py
# Only trade when daily MACD > MACD signal line
if daily_macd > daily_macd_signal:
    # check 1h entry logic

Expected effect: Sharpe x3, drawdown /2, trade count -56%

2. Check the ADX threshold for HOLY_GRAIL ★★★

# Canonical value from Raschke Street Smarts (1995)
ADX_THRESHOLD = 30  # not 25

Make sure: strategies/holy_grail.py actually uses >30.

3. Volume SMA filter for EMA_CROSSOVER ★★★

# In indicators/compute.py
df['vol_sma_fast'] = df['volume'].rolling(10).mean()
df['vol_sma_slow'] = df['volume'].rolling(50).mean()
# In strategies/ema_crossover.py
volume_ok = df['vol_sma_fast'] > df['vol_sma_slow']

Basis: ETH Zurich thesis, Sharpe 3.2 with the combined filter

4. Cash-out regime for EMA_CROSSOVER ★★

When daily_close < daily_ema50: close the position, no new entry until the condition is met. Basis: Grayscale study, 11.5-year backtest - Sharpe 1.9 through bear-market avoidance.