Ready to go further
Next: measure what a win rate really means
The expectancy guide shows how win rate, average win and average loss combine into the number that decides whether a rule makes or loses money.
What is included
- What it is, in plain words
- The formula or mechanism, with exact arithmetic
- How it hits grid, martingale and DCA bots
- A check you can run before you switch a bot on
Who it is for
- People who run or plan to run a trading bot
- Anyone reading a backtest or a bot sales page
- Readers who want the numbers, not the pitch
- Not advice, and not a signal service
Key facts
- "AI trading software" has no fixed meaning. It can describe a fitted model, a plain rule with a new name, a chat assistant that writes code, or a signal list with a glossy cover.
- A tool is judged by what it does with money after costs, not by how it is built. Ask what it trades, on what rule, with what data, and how that was tested.
- Empty claims share a shape: a high win rate with no loss size, a backtest with no costs, no live record, and no description of the method.
- Worked example (made-up): a tool claiming 80% wins scored 60% over 25 paper trades and lost $157.50 after costs (calculated by us on 29 Oct 2026, exact).
- Test any tool on paper or a demo account first, for long enough to see more than one kind of market. This page names no product and makes no claim about any.
What the label covers
The word AI is used for very different things. Four common ones:
- A rule with a new name. A moving average crossover or an indicator filter sold as intelligent. It may be fine as a rule. It is not learning anything.
- A fitted model. A statistical or machine learning model trained on past prices and other data to give a signal. See machine learning trading. It can overfit.
- A language model. A chat assistant that writes code or explains an idea. It does not see the future and it can be wrong with confidence. See the ChatGPT trading bot page.
- A signal or scanner service. It watches many markets and shows alerts. The alert is not a trade, and what you do with it is the strategy.
Behind any of these, the tool still has to send orders at prices that cost something, and the spread, slippage and fees do not care what it is called. A tool can also be a mix: a plain rule with a model filter on top, run by a bot that handles the orders. Each layer is a separate thing to ask about.
A checklist that exposes empty claims
Put each question to the seller or the documentation. A real answer is specific and can be checked. An empty one is a slogan.
| Question | A real answer looks like | An empty answer looks like |
|---|---|---|
| What does it trade, and on what rule? | Markets, timeframe, entry and exit logic in plain words | "Proprietary AI finds opportunities" |
| What data was it built on? | Source, dates, how many markets, how it was split | "Years of market data" |
| How was it tested? | Costs included, an untouched final period, number of versions tried | A single equity curve with no costs |
| What is the loss side? | Average loss, worst drawdown, longest losing run | Only the win rate |
| Is there a live record? | Statements from a real account over many months, including bad ones | Screenshots, or "results vary" |
| What does it cost and how does it earn? | A clear fee, and whether the seller profits when you trade | Unclear, or paid only if you deposit somewhere |
| What happens when it is wrong? | A stop rule, a size limit, a switch you control | "It adapts" |
| Can I test it first? | Demo or paper mode with the same logic as live | Only live, or a demo that differs |
Treat a tool that refuses the first three as unproven, however it is described. Treat a promise of fixed returns, no losing trades or no risk as a reason to stop, since no real trading method can honestly say that. Also remember the seller's records can suffer from survivorship bias: the accounts shown are the ones that did well.
Worked example: a paper test of a claim
Everything here is made up. A tool is advertised with an 80% win rate. You run it on a demo account for 25 trades and write down every result, including costs of $1.5 per trade. 15 trades won an average of $12 and 10 lost an average of $30.
| Item | Value |
|---|---|
| Trades | 25 |
| Win rate measured | 60% |
| Total of wins | $180 |
| Total of losses | -$300 |
| Result before costs | -$120.00 |
| Costs (25 x $1.5) | -$37.50 |
| Result after costs | -$157.50 |
| Average per trade after costs | -$6.30 |
| Win rate needed to break even with these sizes and costs | 75.0% |
Calculated by us on 29 Oct 2026, exact. The measured win rate is 60%, well under the claim, and that is already a reason to ask questions. But look at the second problem: even a 60% win rate loses money here because the average loss ($30) is much bigger than the average win ($12). Break-even needed 75.0%. A high win rate with small wins and large losses is a classic shape, and a win rate alone cannot say whether a tool works. See expectancy and the win rate calculator.
Twenty-five trades is also far too few to be sure of anything. It is enough to catch a gross exaggeration, not to prove a good tool. The more trades and the more kinds of market you cover, the more the result means.
How it hits bots
- Grid and martingale tools sold as smart often keep a high win rate by holding losers. The occasional large loss is the cost of that win rate. See martingale.
- EAs with an AI label are still EAs: they have inputs, a rule and a backtest that can be overfitted. See expert advisors.
- Learning on the fly can mean the tool changes its own rule after losses. A record then belongs to many different strategies at once.
How to test any such tool on paper
- Use a demo or paper account and keep a log: time, market, size, entry, exit, costs. Do not rely on the tool's own report.
- Run it long enough to see a trend, a sideways period and a sharp move. Weeks, not days.
- Compute the win rate, the average win, the average loss, the worst drawdown and the result after costs. See paper trading and drawdown.
- Compare with a simple baseline, such as doing nothing or a plain rule.
- Remember that demo fills are usually cleaner than live ones. A paper result is a best case.
- If you go live at all, start with an amount you can afford to lose entirely, with a hard stop that does not depend on the tool.
What this page does not tell you
It does not say any AI tool works or does not work, and it names none. It gives questions and a method for finding out. The numbers are made up.
Keep Going
Related Tools and Guides
Keep exploring
FAQ
Frequently Asked Questions
What is AI trading software?
There is no fixed meaning. The term is used for fitted models, ordinary rules, chat assistants and alert services. Ask what it does, not what it is called.
How can I tell if an AI trading tool is a scam?
Warning signs are fixed returns or a promise of no losses, only a win rate shown, no costs in the backtest, no live statements and pressure to pay or deposit quickly. None of these proves a scam, but each is a reason to stop and check.
Does a high win rate mean the software works?
No. If the average loss is bigger than the average win, a high win rate can still lose money. Ask for the average win, average loss and the worst drawdown.
How long should I paper trade a tool?
Long enough to cover more than one kind of market, usually weeks to months, and enough trades for the numbers to settle. A few dozen trades only catch gross exaggeration.
Can a chat assistant trade for me?
A chat assistant can write or explain code, but it has no knowledge of the future and can be wrong. Any code it writes needs the same testing as any other bot.
Are the numbers on this page real?
Example numbers are made up to show how the maths works, and the page says so. A figure only counts as exact when the page says it was calculated on a stated date. Nothing you enter in a calculator leaves your browser.
Can I use this to decide whether to run a bot?
It can show whether a bot's numbers hold together and how much an account can lose. It cannot say what the market will do, and it is not advice. Use it to ask better questions, then read the terms of your own broker or exchange.
Is this financial advice?
No. It is general education. The example numbers are made up and nothing here is a recommendation about any product.
Know the Risk Before You Run It
Free calculators and guides, added every week. Leave your details and we will email you when new ones go up.
Trading carries a risk of loss