Cost drag trading fees slippage and trading fees impact on returns are foundational SEO topics because they decide whether a "great" backtest is economically real. Cost drag is the performance lost to fees, spread, and slippage versus a frictionless simulation.

The failure mode

You optimize on mid prices, then pay spread and fees in live trading. High-frequency edges die first; slower edges may survive with smaller size.

What to include in a serious model

Maker vs taker fees and whether your logic can actually rest liquidity
Funding for perpetuals when relevant
Slippage that scales with size and volatility, not only a constant

How cost drag interacts with metrics

Backtest transaction costs impact changes:

Profit factor
Expectancy per trade
Sharpe and Sortino (Sharpe vs Sortino)

Freqtrade note

Freqtrade backtest slippage fees settings must match your venue. If drag is large relative to per-trade expectancy, prioritize robustness under costs over raw in-sample return (How to read Freqtrade results).

A toy example that explains why drag dominates turnover

Suppose a strategy expects +6 bps per trade before costs, but pays 4 bps in fees plus 5 bps in realistic slippage on average.

Net is negative even if the signal is directionally right. High turnover makes small per-trade losses compound quickly.

This is why drag is not a footnote for scalping and short-hold systems.

Scenario ladder (baseline, stressed, shock)

Run three cost tiers and compare rankings:

baseline: your best estimate of live costs
stressed: +50% slippage, wider spread proxy
shock: +150% slippage on volatile weeks only

If your chosen strategy only wins in baseline, you learned something material.

Live monitoring: drag is not static

After deployment, track realized:

average slippage vs model
fee tier drift (maker/taker mix changes)
funding drift for perps

When realized drag diverges, revisit size and kill-switch thresholds before you rewrite the strategy.