The live headwind: how costs eat a paper edge, line by line

A worked example

Everything below is a constructed illustration. The numbers are chosen to be ordinary, not to describe any particular strategy or venue. The point is the shape of the arithmetic, and the shape is what happens to almost everyone.

Take a strategy on a naive backtester. One coin, perps, 300 round-trip trades a year, average position $10,000, and a paper result of +40 percent on a $10,000 account. The chart looks wonderful. Now make it pay its bills.

Exchange fees. Say it pays a 0.05 percent taker fee on each side. That is 0.1 percent per round trip, times 300 trades, on $10,000 positions. The fees are about $3,000. The paper profit was $4,000. Read those two numbers again. Fees alone took three-quarters of the edge, and the backtester never mentioned them.

Spread and slippage. Every entry crosses a spread, and fills walk the book when it is thin. Call it 0.03 percent per round trip on a liquid pair, and more in fast markets, which is when the strategy trades most. Another $900 or so. Slippage is polite in calm markets and greedy in exactly the moments your stop fires.

Funding. The strategy holds longs through a bullish stretch, and longs pay funding in bullish stretches. A few hundredths of a percent every eight hours adds up to several hundred dollars over the held hours of the year. Some strategies collect funding instead. This one pays. The backtest that ignores funding does not know which kind it has.

Partial fills. The limit entries that backtested as filled were sometimes not filled live, and the missing fills cluster on the best trades, because the best trades are the ones where price ran and never came back to the order. Knock a few percent off the gross win column. No invoice arrives for this one. It is the money that was never collected.

The liquidation event. Once, in the worst week of the year, a leveraged position gets stopped in a cascade, and the fill lands far beyond the stop level. The backtest recorded a 2 percent loss on that trade. The account recorded 6. The difference, on this position size, is several hundred dollars, and it arrived in one minute.

The total

Add it up. Roughly $3,000 in fees, $900 in spread and slippage, several hundred each in funding, missed fills, and the bad stop. The +40 percent paper year is now somewhere near +4 to +8 percent, and a slightly weaker version of the same strategy is below zero. Nothing about the signal changed. The market simply presented the bill.

This is the live headwind. It is not bad luck and it is not a broken platform. It is the cost of trading, and the only question is whether you measured it before deploying or after.

What the headwind teaches

Trade frequency is a cost. The fee line was the largest line, and it scales with trade count. Cut the trades in half and you cut the biggest cost in half. The fastest way to improve many strategies is to make them trade less. See trading styles.

Small targets cannot pay big tolls. A 0.4 percent average win cannot give 0.13 percent to fees and spread and remain a business. Higher-timeframe strategies with larger targets keep more of each win. The toll is fixed. The cargo is not.

The worst cases are part of the average. The liquidation fill and the volatile-day slippage are not outliers to be excused. They recur, and they belong in the arithmetic. A strategy evaluated without its worst fills is a different, imaginary strategy. See risk.

Test against the headwind, not the breeze. Every one of these line items can be modeled. Fees and funding are published numbers. Slippage and partial fills follow from book depth. A backtest with these mechanics on tells you the net result before you pay tuition. That is the entire argument of realistic backtesting, and this article is just that argument with receipts.

The one-line summary

Gross edge is a story. Net edge is a business. The headwind is the difference, and it is measurable in advance.

How Edgecraft handles this

Edgecraft backtests include every line item above by default: per-venue fees, spread and depth-based slippage, historical funding, partial fills, and realistic stop behavior. The result you see has already sailed into the wind. When a strategy survives that, the live account has a fair chance of resembling the backtest, which is the only promise a backtest should ever make.