Can you buy a trading algorithm? What to check before you pay

Contents

1. Can you buy a trading algorithm?
2. What is actually for sale: bots, signals, indicators or source code?
3. Which red flags tell you to walk away?
4. What should you ask the seller before you pay?
5. How do you re-test a purchased algorithm on your own data?
6. Should you build your own instead?
7. Frequently asked questions
8. References

Can you buy a trading algorithm?

Yes, you can buy a trading algorithm: it is sold as a closed bot, a signal feed, an indicator or source code. But a sales page backtest proves little until you re-run the rules on your own data with your own costs. The real question is not whether you can buy one, but whether you can check it.

A trading algorithm is a fixed set of rules that decides when to enter a market, when to exit, and usually how large the position is. Bought or built, the rules are only as good as the test behind them, and the test you did not run yourself is the one you cannot trust. The ground rules for any rule set, including why most of them lose after costs, are in algo trading: what it is and what it costs.

What is actually for sale: bots, signals, indicators or source code?

The four products differ mainly in how much you can inspect and re-test. Source code lets you read and re-run everything. An indicator shows you signals but usually has no exit rule, so you cannot backtest a complete strategy from it. A signal feed sends you trades you can only judge after the fact. A closed bot executes rules you never see, so the only evidence is the seller's report.

Check	Source code	Indicator	Signal feed	Closed bot
Can you read the rules?	Yes	Partly	No	No
Can you backtest it yourself?	Yes	Partly	No	Only the seller's report
Can you change costs and sessions?	Yes	Yes	No	No
What you pay for	Rules	Signals only	Trades you cannot audit	Execution you cannot see

Which red flags tell you to walk away?

Two signs matter more than any other: no verifiable live or forward record, and results that come from the best of many tries on one stretch of history with no separate out-of-sample test. Out-of-sample data is the part of the price history the rules were not tuned on, run once with the rules unchanged. Bailey, Borwein, López de Prado and Zhu (2014) show that when many variants are tested on one history and the best is kept, the kept backtest is inflated almost by construction. Harvey and Liu (2015) reach the same conclusion and propose discounting a result for the number of tests behind it. A seller who shows one smooth curve and does not say how many curves were tried is showing you the winner of a contest you did not see.

A quick arithmetic check helps too. Put the seller's stated win rate and average win and loss into the free win rate and expectancy calculator. If the expectancy per trade is smaller than a realistic round-trip commission plus slippage, the record cannot survive live costs, whatever the curve looks like.

What should you ask the seller before you pay?

Ask seven concrete questions in order, and treat a refusal to answer as the answer.

1. Can I read the rules or the code? Without them, every later check depends on the seller's word.
2. Which commission and slippage are in the backtest? A result with no costs is a measure of price movement, not of a tradable edge.
3. Which bar type and execution timing? Next-bar execution is filling an order on the bar after the signal bar, the earliest moment a real order could fill; same-bar fills on the close use a price you could not trade.
4. How many variants were tested before this one? The answer sets how much of the result is selection.
5. What is the out-of-sample result? It should be weaker than in-sample and still positive after costs.
6. Is there a live or forward record, and can I verify it? Broker statements beat screenshots.
7. What happens when the market changes? A seller who has never had a losing year has not been selling long, or is not showing you the losing year.

How do you re-test a purchased algorithm on your own data?

Re-run the rules on your own bar data with your own commission, slippage and next-bar execution, keep a recent block of data the seller never saw for one final run, and judge both runs against a pass or fail rule you wrote before you looked.

Our own run shows why step 5 matters. AlphaPrime (Valynox) tested a stand-in rule, not a purchased algorithm: 1,024 variants of a Rate Of Change rule as the only entry and exit, with no stop, target or filter, on NQ daily bars from 2019-01-02 to 2025-06-30, long and short, one contract, with $20 commission and $10 slippage per trade. The box shows the one candidate the search kept for its return against its drawdown.

Read it the way you would read a sales page. A net profit of $172,380 against a maximum drawdown of $114,570 came from only 77 trades, and the candidate was kept after the fact because it did best among 1,024 tries. That selection flatters the result; it does not mean the Rate Of Change rule earned that money, and it is not what a reader can expect. A seller's single curve is the same object with the other tries hidden. AlphaPrime's strategy validation grades every surviving candidate against thresholds set before the search starts, which is the discipline a buyer needs as well: the pass mark exists before the result does.

Should you build your own instead?

Building your own from simple, understandable rules costs time instead of money, and in return you control the rules, the costs and every test. The seven questions above are easy to answer when you are the seller. A first algorithm does not need to be clever: a session filter, one entry condition, one exit condition and a stop are enough to learn the full cycle of test, re-test and go live. The 30-day route to a first algo trading strategy lays out that cycle week by week, and how to build and test a day trading algorithm shows what an intraday version looks like after costs. Whichever path you take, the steps from tested rules to a live order are the same, and how to trade futures automatically walks through them in order.

Your next step: take the one algorithm you are thinking of buying, write its pass or fail rule on paper, and ask the seller the first question on the list.

Frequently asked questions

Can you buy a trading algorithm and run it on your own platform?

Yes, if it is sold as source code or as an indicator for your platform. A closed bot or a signal feed runs on the seller's side, so you cannot load it into your own backtester. Source code is the only form you can read, change the costs on and re-test in full.

Are trading bots for sale a scam?

Not by definition, but most sold bots cannot be checked. A closed bot gives you the seller's report and nothing else, so you cannot see the rules, the costs assumed or how many variants were tried before the shown one. Judge any bot by how much of it you are allowed to verify, and treat a result you cannot re-run as unproven.

How much does a trading algorithm cost?

Prices range from a one-time fee for an indicator or source code to a monthly fee for a signal feed or bot, and the price tells you nothing about quality. The cost that matters more is the trading cost inside the backtest: if commission and slippage were left out, the shown result is not tradable at any price.

Can I buy an algo trading strategy that is certain to work?

No. Markets change, and a rule set tuned to one stretch of history can stop working when conditions shift. The most a seller can honestly offer is readable rules, a backtest with realistic costs, a separate out-of-sample result and a verifiable live record. Everything else is a claim you cannot check.

References

• Bailey, D. H., Borwein, J., López de Prado, M., & Zhu, Q. J. (2014). Pseudo-Mathematics and Financial Charlatanism: The Effects of Backtest Overfitting on Out-of-Sample Performance. Notices of the American Mathematical Society, 61(5), 458–471.
• Harvey, C. R., & Liu, Y. (2015). Backtesting. The Journal of Portfolio Management, 42(1), 13–28.