Are trading bots profitable?

The honest framing is that a trading bot is a way of executing a strategy consistently, not a source of edge in itself. If the underlying rules have positive expectancy after costs, automating them removes the largest source of retail underperformance, which is inconsistent execution of a plan. If the rules have no edge, automation applies that faithfully and tirelessly.

Most bots sold to retail traders are not profitable over a full cycle, and the reasons are predictable. Many are overfitted to a backtest — tuned until the historical result looked good, which guarantees the result describes the past rather than the future. Many others hide risk in martingale or grid mechanics, producing a beautiful equity curve until a single trend destroys the account.

The numbers worth demanding from any vendor are the ones that are uncomfortable to publish: maximum drawdown, the length of the test period, whether an out-of-sample period was used, and how many trades the sample contains. A win rate quoted without an average win and average loss beside it is the flattering half of the picture.

What our own figures actually say

Run as one unbroken six-month stretch on a $1,000 account with live percentage sizing — the closest measurement to an installed bot — ours returned +3.5% in total, not per month, across 42 trades. Maximum drawdown was 7.2%, the win rate 38%, and four of seven months were green. The profit was made between February and May; June, July and August lost money.

The rules were separately validated across seven independent monthly windows, each restarting the learning from a blank slate at a fixed trade size. Those runs were positive on January to July 2026 at a 21.2% drawdown, and positive again on August to December 2025 — data the configuration had never seen — at a 20.5% drawdown with two of five months negative. That method shows the edge holds across separate periods. It is not what a continuously running bot reproduces, and the difference is position sizing.

Forty-two trades is a small enough sample that no percentage here should be treated as precise, and the honest drawdown to plan against is the worst of the three measurements — around 21% — rather than the kindest. Anyone quoting a tighter number over a shorter period is quoting noise.

What would make a bot not worth running

If the account is too small for the strategy's minimum position, the broker minimum lot forces real risk above the configured percentage and the arithmetic stops working regardless of the edge. If the account cannot tolerate a 20% drawdown without the owner intervening, the system will be switched off during the drawdown, which converts a temporary loss into a permanent one.

And if the expected return is compared against the cost of the subscription rather than against the capital at risk, the picture is misleading in both directions. Read the risk disclosure before deciding.