A paper-trading result is a clean number. It assumes you buy at the price on the screen and sell at the price on the screen, and that nothing is charged for either. Live trading never works that way. The gap between the two is one of the most common reasons a strategy that looked fine on paper loses money when it goes live.
Exchange fees. Every order that opens or closes a position pays a fee. A round trip (in and out) pays it twice, and market orders usually pay more than limit orders.
The spread. You buy at the ask and sell at the bid. On liquid coins the gap is small. On thin coins it can be larger than the move you were trying to capture.
Slippage. The price moves between the moment a signal fires and the moment the order fills. It tends to move against you at the worst times, such as fast breakouts and stop-outs, because many other orders are reacting to the same move.
Funding. On perpetual futures, holding a position across funding times can cost or earn a small amount. Over a long hold it adds up.
Say a strategy averages a gross +0.40% per trade in a paper test, and the all-in cost of a round trip (fees plus spread plus slippage) is 0.15%. The net edge is +0.25%, which is already about 38% smaller than the paper figure.
Now say the same strategy trades a thinner coin and the all-in cost is 0.40%. The net edge is zero. Nothing about the signal changed. Only the cost did. These figures are illustrative, not from our book, but the shape is real: the smaller the average gain per trade, the more of it costs can take.
Strategies that trade often and aim for small moves are the most sensitive, because the cost is paid on every trade and the average gain is small. Strategies that hold longer and aim for larger moves are less sensitive. A backtest with thousands of trades and a small average gain should be treated with the most suspicion until costs are applied.
1. Deduct costs in the backtest. Subtract a realistic round-trip cost from every trade, then recompute the profit factor. If the profit factor drops below 1.0, the strategy had no edge after costs.
2. Use the worst-case fill. If the signal fires on a bar, assume the fill happens at the next bar's open, or at a price slightly worse than the signal price, not at the signal price itself.
3. Compare live to paper, trade by trade. For the same signals, compare the paper entry to the actual fill. The average difference is your real slippage. It is a measurement, not a guess.
4. Read the number on its own sample. A small live sample can be noisy in either direction. Treat the gap as a range, not a point.
On our research page, paper and testnet rows are labelled as such, and they are kept separate from live ones, because paper results do not include real fills. Several strategies in the book have a live profit factor below their backtest profit factor. Costs and execution are one of the reasons we look at first when that happens, and a pre-set kill rule applies either way.
Treat any paper or backtest number as an upper bound. The question that matters is what is left after costs, and whether that remainder is large enough compared to the noise in the sample. If a result is only attractive before costs, it was never attractive.
This is why the desk publishes paper, testnet and live results separately. The gap between them is part of the research.
📘 Free: get our plain-English field guide — the 7 checks before you trust anyone with your crypto.
Join the Telegram → · see the live research · or talk to us.