Slippage: The Hidden Cost That Quietly Kills Backtested Strategies

What slippage actually is

Slippage is the gap between the price your strategy assumed and the price you actually got. It is the sum of several frictions that all push your fill away from the clean number in your backtest.

• →Spread: you buy at the ask and sell at the bid, so you start every round trip down by the full bid-ask spread.
• →Market impact: your order consumes resting liquidity and walks the book, so larger size fills at progressively worse prices.
• →Latency: the time between your signal and your order reaching the matching engine; the market moves while your packet is in flight.
• →Partial fills: a limit order that only partly executes leaves you with a different position than you sized for.
• →Queue position: for maker orders, where you sit in the price-time queue decides whether you get filled at all before the level trades through.

None of these appear in a naive backtest. Every one of them is a real, recurring cost in live trading. The danger is that slippage is invisible in research and only shows up once real money is on the line.

Why backtests ignore it

Most backtesters fill orders at the mid price (or the close of the signal bar) and assume you always get exactly the size you asked for. That builds three false assumptions directly into your equity curve.

• →Mid-price fills: the backtester ignores that you actually pay the ask and receive the bid, giving away half the spread on each side for free.
• →Infinite liquidity: it assumes any size fills instantly at one price, so a $10k order and a $10M order are treated identically.
• →No queue: limit orders are assumed to fill whenever price touches the level, when in reality you may sit behind thousands of contracts and never get filled.

These assumptions are not malicious; they make the simulation tractable. But they systematically flatter your results. A strategy that looks like it earns +0.4R per trade can quietly be a losing system once realistic slippage is subtracted, and you will not find out until you have paid the tuition live.

The four components, broken down

It helps to put numbers on it. Slippage is usually quoted in basis points (bps) of notional, where 1 bp = 0.01%. Spread and impact are the two components you can estimate before you trade; latency and partial fills are situational. Here is a worked example on a single entry.

That 0.05R looks trivial. It is not, once you compound it. Add the taker fee (roughly 0.055% per side on Bybit, another ~0.11R round trip) and a strategy doing 100 trades a month is paying something like 16R per month in fees and slippage combined, before any directional edge. Latency and partial fills only make this worse on the trades where they bite.

How slippage scales with size and venue

Spread cost is roughly fixed per unit of notional, but market impact grows non-linearly with size. The first few thousand dollars fill at the top of the book; the rest walk down through progressively thinner levels. Double your size and you often more than double your impact.

Venue matters just as much. On a deep market like BTC perps on Binance, Bybit, or Hyperliquid, a $50k order barely dents the book and you pay close to the half-spread only. The same $50k in a thin altcoin perp can sweep five or ten price levels and cost you 20 to 50 bps, because the resting liquidity simply is not there.

This is why a strategy that backtests beautifully on a basket of alts can be uninvestable in practice: the edge is real, but the impact cost of getting in and out at any meaningful size eats all of it. Always size your backtest assumptions to the actual depth of the market you intend to trade.

Modeling slippage in a backtest

There is no single correct slippage model. The right one depends on how often and how large you trade. Three models cover most cases.

• Fixed bps per side

  Subtract a constant, e.g. 3 bps, from every fill. Cheap and good enough for slow, liquid strategies (daily trend on BTC/ETH) where spread dominates and impact is negligible at your size.

• Spread-proportional

  Charge half the actual bid-ask spread at order time plus a small impact multiplier. This captures the fact that spreads widen in volatility, which is exactly when many strategies trade. Best for intraday and mean-reversion systems.

• Volume / participation-based

  Scale impact with your order size as a fraction of bar volume (participation rate), often with a square-root impact term. Required for high-frequency strategies and any strategy trading meaningful size relative to liquidity.

Whichever you choose, validate it against reality: log your live fills, compare them to the mid at order time, and tune your model until backtest and live execution costs agree. A backtest whose slippage model has never been calibrated against real fills is a guess wearing a lab coat.

Which strategies are most exposed

Slippage is not evenly distributed across strategy types. Two factors decide your exposure: how often you trade, and whether you demand liquidity when it is scarce.

• →High-frequency: most exposed. Slippage is paid on every trade, and at thousands of trades the per-trade cost dwarfs the per-trade edge unless you are a consistent maker.
• →Mean-reversion at extremes: highly exposed. You buy capitulation and sell euphoria, precisely when spreads are widest and the book is thinnest, so your impact is at its worst.
• →Illiquid markets: exposed at almost any size. Even a slow strategy pays heavy impact when the book cannot absorb your order.
• →Slow trend following: least exposed. You trade rarely, often add to liquidity, and the slippage is small relative to the multi-R moves you are capturing.

The uncomfortable pattern: the strategies that look most impressive in a frictionless backtest, like tight, frequent, high-Sharpe mean-reversion, are often the ones slippage destroys most thoroughly. The boring trend follower survives contact with the real order book.

Reducing real slippage

Once you have measured it, you can attack it. Most of the reduction comes from a handful of execution choices.

• Use limit and maker orders

  Posting passively turns the spread from a cost into a rebate or a lower fee, and avoids impact entirely. The trade-off is non-fills; only use it where your strategy tolerates missing some entries.

• Slice large orders (TWAP / VWAP)

  Breaking a big order into smaller child orders over time keeps your participation rate low and stops you from walking the book in one print. Essential once your size is a meaningful fraction of volume.

• Select the deepest venue

  Route each instrument to the exchange with the tightest spread and deepest book for that pair. The same trade can cost 10× more on the wrong venue.

• Avoid the open and close

  Spreads blow out and books thin during the volatile first and last minutes of a session (and around major data releases). Delaying entry by a few minutes can halve your slippage.

• Funding-aware timing

  On perps, time entries and exits around funding settlement to avoid the crowded rebalancing flow, and to keep funding from quietly eating the edge you fought to preserve on slippage.

None of these is exotic. Together they routinely cut real slippage by half or more, which is often the difference between a strategy that survives the transition from backtest to live and one that does not.

Summary

Frequently asked questions