Many crypto strategies fail in production for one boring reason: slippage was modeled as a constant.
In live markets, slippage is state-dependent and often nonlinear.

Why fixed slippage assumptions fail

A fixed value like "5 bps per trade" ignores:

liquidity depth at order size,
spread expansion under stress,
volatility spikes around events,
execution latency and queue dynamics.

This can turn a seemingly profitable strategy into negative net edge.

Three slippage tiers you should test

Instead of one assumption, run at least three:

Baseline: calm market estimate.
Stressed: wider spread and thinner depth.
Shock: event-like market conditions.

If your edge survives only baseline, it is likely not deployment-ready.

Slippage should scale with participation

Cost impact grows with participation rate.
Small orders can be close to top-of-book fills; larger orders move through levels.

A simple practical model:

low participation: near spread cost,
medium participation: spread + impact term,
high participation: nonlinear impact penalty.

Even a rough nonlinear model is usually better than fixed bps.

Connect slippage to regime classification

Use separate assumptions for:

trending high-volatility periods,
low-volatility range periods,
macro-event windows.

Your backtest should report which regime contributes most of the expected returns and how slippage changes there.

Validation checklist before deployment

Net edge remains positive under stressed slippage.
Drawdown remains inside operational tolerance.
Position sizing is reduced for thin-liquidity pairs.
Kill-switch rules account for execution degradation.

If these checks fail, improve execution logic before changing strategy logic.

Typical false positives

High turnover systems with optimistic fill prices.
Mean-reversion entries on illiquid pairs during spread blowouts.
Strategies calibrated only on low-volatility months.

Record slippage assumptions next to every equity curve

If the chart does not say which slippage tier produced it, the chart is not evidence.

Treat slippage tables like versioned dependencies, not informal notes.