Slippage and trading costs: the silent tax that kills good strategies

There is a category of trading strategy that is profitable, robust, validated — and still loses money. The reason is almost always the same: it was profitable before costs and unprofitable after them. Slippage and trading costs are a silent tax, and they are the single most common reason a strategy that looked good in a backtest bleeds out live.

The four costs

• Fees — the explicit commission the exchange charges per trade. Maker and taker fees usually differ; a strategy that always crosses the spread pays the higher taker fee every time.
• Spread — the gap between the best bid and best ask. A market order to buy fills at the ask and a market order to sell fills at the bid, so every round trip pays the spread whether you notice it or not.
• Slippage — the difference between the price you expected and the price you got. Caused by the market moving between decision and fill, and by your order being larger than the liquidity at the best price.
• Funding — on perpetual futures, the periodic payment between longs and shorts. Hold a position across funding intervals and this accrues quietly, for or against you.

Why it compounds — and who it hurts most

Costs are paid per trade, so their damage scales with trade frequency. A strategy that holds for weeks pays the toll a handful of times a year — costs are a rounding error. A strategy that trades several times a day pays it hundreds of times, and a per-trade cost that looks trivial in isolation becomes the dominant term in the P&L. This is the cruel asymmetry of high-frequency retail strategies: the more often you trade, the larger the edge you need just to break even, and small edges are exactly what frequent trading tends to find.

Modelling costs in a backtest — properly

A backtest with costs switched off is not a backtest, it is a sales brochure. But switching them on naively is also wrong. Fees are easy — a known percentage. Spread should reflect the real instrument, not a guess. Slippage is the hard one: it is not a constant. It depends on order size relative to liquidity and on how volatile the market is at the moment of the fill. A fixed slippage assumption flatters fast-moving conditions and punishes calm ones. The honest approach is to model slippage as a distribution and stress it — re-run the backtest with adverse slippage and see whether the edge survives the bad draws, not just the average one.

This is also why Monte Carlo matters here: by jittering slippage across thousands of resampled runs, you see the range of cost outcomes instead of betting everything on one optimistic constant.

What to actually do

1. Model all four costs from the very first backtest — never evaluate a gross-of-cost curve, even early.
2. Use realistic, instrument-specific fee and spread numbers, not placeholders.
3. Treat slippage as a stress test: re-run with adverse slippage and keep only strategies that still work.
4. Prefer fewer, higher-conviction trades — every trade you remove is cost you keep.
5. Watch funding explicitly if you trade perpetuals; over a long hold it can quietly dominate.

Noon Barbari's backtest engine models fees, spread, slippage and costs by default — frictionless backtests are not an option — and the Monte Carlo page jitters slippage across 1,000+ runs so you see the cost distribution, not a single lucky constant.

An edge is not what your strategy makes. It is what your strategy makes after the market has taken its cut. Measure that number from the first backtest, and you stop being surprised by it live.