How to Evaluate a Trading Algorithm Before You Buy
Del
Updated 10 November 2026
To evaluate a trading algorithm before you buy, check five things in the seller's own data: a live track record, results after fees, the largest drawdown, the most recent 12 months and the result of every algo, including the losers. Then check that any backtest is labelled as simulated, and test the algo yourself before paying.
Key facts
- Our live account has traded since December 2023, 35 months, and 21 of those months were positive (live data per 31 October 2026).
- The largest decline at month end was −23.5 %, from February to April 2024. Per 31 October 2026 the account was −18.2 % below its May 2025 peak.
- The last 12 months, November 2025 to October 2026, gave −2.8 % or −6,485 kr after fees, with 5 positive months out of 12.
- Of the 12 algos that traded in the account before version 2.0 started on 21 October 2026, 8 were positive and 4 negative since start.
What does it mean to evaluate a trading algorithm?
Evaluating a trading algorithm means checking whether the evidence behind it is real, recent and complete, before you put money on it. It is not about finding the highest return figure, but about understanding what the algo has actually done and how bad it has been.
EU rules for investment firms give a useful checklist. Under Commission Delegated Regulation (EU) 2017/565, Article 44, past performance must not be the most prominent part of the information, should cover complete 12 month periods, must state the period and source, and must carry a clear warning that past performance is not a reliable indicator of future results. Simulated performance needs its own clear warning.
The table sums up the seven questions and how our own live account answers them.
| Question | What to look for | Our answer (per 31 October 2026) |
|---|---|---|
| 1. Is there a live track record? | A real account, start date, monthly results | Live since December 2023, 35 months |
| 2. Are results after fees? | Net figures, fees shown | Net after spreads, commissions and fees |
| 3. What was the largest drawdown? | A stated figure with dates | −23.5 %, February to April 2024 |
| 4. How were the last 12 months? | Recent results, not only since start | −2.8 %, November 2025 to October 2026 |
| 5. Are losing algos shown? | Results per algo, also negative ones | 8 positive, 4 negative |
| 6. Is the backtest labelled? | Simulated, with period and settings | Simulated, April 2018 to September 2026 |
| 7. Can you test it first? | A trial on your own platform | 14 day free trial |
Is there a live track record, and is it after fees?
A live track record is the result of real orders in a real broker account, and it should be shown after fees. It is the only evidence that includes slippage, spreads, outages and the seller's own discipline.
Our ProRealTime account at IG has published a result every month since December 2023 on the live results page, with screenshots from the platform. In October 2026 the algos made 3,791 kr before costs, fees and currency took 222 kr, and the net result was 3,569 kr. Ask any seller for the same split. A figure without fees looks better than what ends up in the account.
What was the largest drawdown, and when did it happen?
The largest drawdown is the deepest fall from a previous high, and it tells you what you would have had to sit through. A seller who does not state it, with dates, is leaving out the most useful number.
Our largest decline at month end was −23.5 %, from February to April 2024, and the worst single month was −13.9 % in April 2024. The current decline matters too. Per 31 October 2026 the account stood at 222,316 kr, −18.2 % below its peak of 271,930 kr in May 2025, and 7 of the 17 months since that peak were positive. How we size positions around numbers like these is covered in risk management in algo trading.
How did the last 12 months look, not just since start?
A result since start can hide a weak recent period, so always ask for complete 12 month periods. That is also how the EU rules above want past performance shown.
| Period | Result after fees | Positive months |
|---|---|---|
| Dec 2023 to Oct 2024 (11 months) | +315.5 %, +110,408 kr | 7 of 11 |
| Nov 2024 to Oct 2025 | +57.4 %, +83,393 kr | 9 of 12 |
| Nov 2025 to Oct 2026 | −2.8 %, −6,485 kr | 5 of 12 |
Our first year started from 35,000 kr, so a large percentage came from a small base. The most recent 12 months were slightly negative. Both are true, and a fair evaluation looks at all three rows. Past results, live or backtested, do not predict future returns.
Are losing algos and months shown too?
A seller should show every algo and every month, including the negative ones, because a portfolio is judged by what is left after the losers. If only the best algo is shown, you cannot tell what the rest did.
Of the 12 algos that traded in our account before version 2.0 started on 21 October 2026, 8 were positive and 4 negative since start. The largest gain was Nasdaq M3 pt.1, +182,807 kr from December 2023 to 21 October 2026, now running as Nasdaq M3 Vega. The largest loss was Gold M2, −28,358 kr. The full list per algo and year is on A+ Algos in numbers.
Also ask which version the live data belongs to. Our new portfolio started in the account on 21 October 2026, so its own live record is still short, and most of the history comes from the earlier algos.
Is the backtest labelled as simulated, and does live match it?
A backtest is a simulation of how the rules would have traded on past prices, and it should always be labelled as simulated, with period, costs and settings. It is useful for testing rules, not as proof of future results.
US rules for commodity trading advisors require a fixed warning on hypothetical results. 17 CFR 4.41 says such results "have certain inherent limitations" and "do not represent actual trading". Our backtest page shows simulated results from April 2018 to September 2026 and is labelled that way. Compare it with the live account for the same months. If live is much weaker than the backtest, ask why. Common myths about backtests are covered in algo trading myths checked against live data.
How does it trade, and can you test it before you buy?
You should know how often the algo trades and how much it risks per trade, and you should be able to watch it on your own platform before paying. Without that, you are buying a description, not a system you have seen.
Our ProRealTime algos average 7.7 orders a day, 69 % of trades were winners and risk per trade is under 2 % of the account, according to the live results page (1 October 2026). A high win rate says little on its own, since a few large losses can outweigh many small gains. With our 14 day free trial you can run an algo on your own ProRealTime or MetaTrader 5 account before you buy.
FAQ
How long should a live track record be?
Longer is better, and it should include at least one bad period. The EU rules for investment firms ask for the last five years, or the whole period if it is shorter, in complete 12 month periods. Our live account has 35 months, from December 2023 to October 2026, including a −23.5 % decline in 2024 and a weak 12 months to October 2026.
Is a high win rate a good sign?
Not on its own. Win rate tells you how often trades end positive, not how large the wins and losses are. 69 % of our trades were winners per 1 October 2026, yet the account still had 14 negative months out of 35. Look at the result per month, the largest drawdown and the last 12 months together.
Why are backtest results usually better than live results?
A backtest knows the whole price history and is often tuned on it, while live trading meets real spreads, slippage, outages and new market conditions. That is why a simulation must be labelled as simulated. In September 2025 IG had IT problems and our algos largely stopped trading after 10 September, something no backtest would show.
What is a red flag when buying a trading algorithm?
Red flags are results without a period, no live account, figures before fees, only one algo shown, no drawdown and promises of certain profit. Under Swedish marketing law such claims must be possible to back up. A seller who cannot show monthly live results with dates gives you nothing to evaluate.
Summary
To evaluate a trading algorithm, ask for a live record after fees, the largest drawdown with dates, the last 12 months, every algo including the losers and a labelled backtest, and then test it yourself. Our own answers per 31 October 2026 include both a strong first year and a slightly negative last 12 months.
See the results live. Every month since December 2023 we publish the result of our own algos after fees on the live results page. Our main product is A+ Algo Portfolio V2.0, nine algos for ProRealTime or ten for MetaTrader 5.
Trading CFDs involves high risk and you can lose money. Past results, live or backtested, do not predict future returns.