ORBIT
Video tutorial · BOTLAB · 6 min read

How to backtest a trading strategy: ask your AI, test it, then improve it one idea at a time

Episode 3 of the BOTLAB series turns one plain-words idea into a tested strategy on MNQ futures, with costs counted and fresh days kept aside. Watch it, then follow the same loop step by step.

By Amlan Mitra, founder of ORBIT · November 15, 2026

Why backtest at all

A backtest runs a set of rules over past market data and records every trade they would have taken. It cannot tell you what happens next, but it is the fastest way to find out that an idea does not hold up before you risk anything on it. Knowing how to backtest a trading strategy properly, with costs and with days kept aside, is what separates a test from wishful thinking. In ORBIT this is done in BOTLAB, the backtesting software built into the desk.

Step 1: describe the strategy in plain words

The idea in the video is MACD crosses taken only in the direction of the hourly trend. Typed just as you would say it:

  • MNQ, 5-minute candles: buy when the MACD line (12, 26, 9) crosses above its 9-period average line while the hourly 20 EMA is above the hourly 50 EMA; sell short on the opposite cross while the 20 is below the 50.
  • New York hours, from 15 minutes after the open to 6 hours after it, at most 3 trades a day, 1 contract.
  • A stop at 2 × ATR(14), two times the average bar size, a target of 2R, flat at the end of the session.

Press send. Your own AI builds it from BOTLAB's building blocks and ORBIT tests it straight away. Every rule is written back in plain words, so you can check it did what you asked.

Step 2: make the stop yours

The first result uses the stop you asked for only as a suggestion. The stop is always yours to choose: type your own, have ORBIT measure three stop settings on your history, or press Keep this stop. Then press Test on history.

Step 3: count the costs

Every result in BOTLAB is after costs. In the video that is $1.24 of commission for the round trip plus one tick of slippage on each side, $2.24 per MNQ round trip in all. A backtest without costs flatters any strategy that trades often.

Step 4: practice days and fresh days

The test ran on all the saved history, 109 trading days from June 11 to November 11, 2026, split in two: the first 70% as practice days and the last 30% as fresh days, kept aside the whole time. The video reads out the trades, the net after costs and the expectancy per trade for both halves.

The fresh days are the honest check. If a strategy only looks good on the days you built it on, the fresh days are where that shows.

Step 5: try one idea at a time

Under the results, BOTLAB lists Ideas to test, found on the practice days only so the fresh days stay a fair check. In the video the three ideas were short only, a breakeven stop, and no entries after 1:00 PM New York time.

Press Try it on one. "No entries after 13:00" is an exact change ORBIT makes itself, with no AI call, and the new version is tested on the same history. In the video it measured better than the original on the practice days, stayed positive on the fresh days, and had a smaller maximum drawdown. Every version is kept, so any step can be undone.

The loopAsk for a strategy, make the stop yours, test it, then try one idea at a time. Keep a version only when it does better on the practice days and still holds up on the fresh days it never saw.

Common backtesting mistakes

  1. Leaving out costs. Commission and slippage are real on every trade.
  2. Testing on the days you tuned on. Keep fresh days aside and look at them last.
  3. Changing many things at once. One change per test tells you what actually made the difference.
  4. Reading a backtest as a promise. It is a measurement of the past, not a forecast.

Join the ORBIT Discord

Come trade notes with usTell us which idea you want to test next, or bring a strategy you are stuck on. Traders on ORBIT compare BOTLAB notes in the ORBIT Discord, and new episodes are shared there first.

Launch price: sign up before November 30 and keep ORBIT at $79 a month for life, for as long as you stay subscribed. After November 30 the price is $149 a month. Get ORBIT

Love ORBIT? Share it and grow a recurring monthly income with our affiliate program. To express interest, email our founder directly: amlan@tradeonorbit.com

More tutorials like this one are on the ORBIT YouTube channel.

Get ORBIT

BOTLAB comes with every ORBIT subscription. Get ORBIT here. New to BOTLAB? Start with the three-minute BOTLAB tour, then build your own indicator. For how AI fits the rest of the desk, read AI futures trading.

Every figure in the video is a backtest on saved historical 1-minute data with modeled costs and fills; it is not a forecast and does not reflect real trading. No order was placed in the video. ORBIT supplies no strategy: every strategy is the customer's own. Futures trading involves substantial risk of loss.

Frequently asked questions

How do I backtest a trading strategy in ORBIT?

Describe it in plain words in BOTLAB's chat. Your own AI builds it, ORBIT tests it on the history saved on your PC after costs, and you can then make the stop yours and test again.

What are fresh days?

The last part of your saved history, kept aside while you build and improve the strategy, so you can check it on days it has never seen.

Is a backtest a prediction?

No. It measures how a set of rules would have done on past data, with modeled costs. It helps you discard weak ideas; it does not forecast results.

Test the idea before you trade it.

BOTLAB builds your strategy from your own words and backtests it on your saved history, costs counted, fresh days kept aside.