Paper Trading vs Live Trading: Why Your Results Drift

You run a strategy on paper for a month. It looks fine. You switch to real money, and the same rules do worse. Nothing in the code changed.

That gap between simulated results and real results is called drift. Every strategy has some. The question isn't whether your paper results will drift. It's how much, and whether the idea still holds up after it does.

This post covers where the gap comes from and how to measure it before real money is on the line. If you haven't tested the idea yet, start with how to test an AI trading strategy before you use real money. If the backtest looks too good, check the 7 signs of backtest overfitting first. Drift on top of an overfit strategy is two problems, not one.

Why paper and live give different results

A paper trade is a guess about what would have happened. A live trade is what did happen. The guess usually leaves out the friction.

1. Fees

Many paper tools skip fees or use one flat number. Real exchanges charge different rates for different order types and account tiers. A strategy that trades often can lose most of its edge to fees alone.

2. Slippage

On paper, you buy at the price on the screen. Live, a market order takes whatever is in the order book, and the price can move while you fill. The bigger the order and the thinner the market, the worse it gets.

Slippage is often worst exactly when strategies like to trade: fast moves, news, and breakouts.

3. Fills that never happen

Paper tools often assume your limit order filled the moment price touched it. Live, you're in a queue behind other orders. Price can touch your level, bounce, and leave you with nothing.

The trades you miss are often the good ones, because price moved away fast. The trades you do get are more often the ones that keep going against you. Paper counts both. Live gives you mostly the second kind.

4. Partial fills

Sometimes you only get half your order. Now your position is the wrong size, and your stop and target are set for a trade you don't fully have. Most simple paper tools never model this.

5. Latency

Your signal fires. Then the data has to arrive, your system has to decide, and the order has to reach the exchange and be accepted. On paper, that takes zero time. Live, it never does.

For slow strategies it barely matters. For anything that reacts to fast moves, a second or two can change the price you get.

6. Funding and borrowing costs

On perpetual futures, you pay or receive funding every few hours while a position is open. Margin and borrowing have costs too. Paper tools that ignore them make long holds look cheaper than they are.

7. Your own behavior

This is the one nobody models. On paper, you follow the rules. With real money, people skip the next trade after three losses, close winners early, move stops, or double size to "make it back."

The rules didn't drift. The person running them did.

How to measure the gap before going live

You can't remove drift. You can measure it, so the real number doesn't surprise you later.

Make paper as real as you can. Use the real fees for your account, real margin rules, and real timestamps. Assume market orders pay the spread plus some slippage. Assume limit orders fill only when price trades through your level, not when it just touches it.

Make costs worse on purpose. Rerun the test with double the fees and extra slippage. If the strategy only works with perfect costs, it doesn't really work. Same for backtests: see how to backtest without fooling yourself.

Log what each paper trade assumed. For every trade, write down the signal time, the price you assumed, and the fee you assumed. Later, you can compare line by line against real fills. Totals alone hide the cause.

Go small first, and compare. When you do go live, wherever you trade, start small and keep paper running on the same signals. Then compare trade by trade: fill price vs assumed price, fees paid vs fees assumed, trades missed, and time from signal to fill. The difference is your real drift. Now it's a number you measured, not a guess.

Write your behavior rules down. Decide before going live what you'll do after a losing streak. Then check whether you followed them. Behavior drift is still drift.

What drift can look like

Here's a made-up example. Say a strategy shows 14 of 20 weeks net positive on paper (simulated, example only). Add real fees, and it's 11 of 20. Assume limit orders only fill when price trades through, and it's 8 of 20.

The idea didn't change. The friction did. The last number is the one to plan around, not the first.

How Pickled Lizard handles drift

Pickled Lizard is an AI trading and research platform. It isn't an exchange and it doesn't hold your money. It connects to your own exchange accounts. Right now, testing runs on paper.

• Testing goes in steps: backtest, then shadow, then paper, then verify. Each step is closer to real conditions than the last.
• Paper trades use real fees, margin rules, and timestamps, and a strategy only sees data that existed at that moment.
• Risk is watched tick by tick. A 4-hour strategy doesn't mean waiting 4 hours to notice risk changed.
• When we found some of our own exit records couldn't be trusted, we kept them in the log and refused to count them. A result we can't prove isn't a result.
• Failed tests stay in the record next to passing ones, and results are shown one way: "X of N net positive (simulated)."

We're not building an AI that tells you what to buy. We're building the infrastructure that forces AI trading ideas to prove themselves.

A quick drift checklist

Before you trust paper results, ask:

• Are the fees in my test the real ones for my account?
• Do market orders pay the spread and some slippage?
• Do limit orders fill only when price trades through?
• Does it still hold up with double costs?
• Did I count funding on positions held for days?
• Do I log the assumed price and time for every trade?
• Do I have written rules for my own behavior after losses?

If several answers are no, your paper results are probably better than live will be.

Try it free

The beta is free. Bring a strategy, from you or your AI, and run it on paper with real fees, margin rules, and clean timestamps. You can see simulated results on our paper trading leaderboard.

Test it free at pickledlizardz.com/join

More plain guides: see all posts on Learn. Why I'm building this: read the About page.

Paper trading only. Results are simulated. Not financial advice.