ACADEMY / Strategy generation

Is algo trading profitable? The numbers that decide it for retail

Algo trading is a method, not an edge. It can be profitable for a retail futures trader only when a tested rule keeps a positive average trade after commission and slippage on data it was not tuned on, and when the account can hold through a drawdown larger than the backtest showed.

By Soren Huang · Published 2026-10-30 · updated 2026-10-31 · Editorial policy

Contents

Is algo trading profitable for a retail futures trader?

Algo trading can be profitable for a retail futures trader only when a tested rule keeps a positive average trade after commission and slippage on data it was not tuned on, and when the account can hold through a drawdown larger than the backtest showed. The software decides none of this. The research process does.

Algorithmic trading is trading in which written rules decide every entry, exit and position size, and a program executes those rules without a human choosing each trade. That definition says nothing about profit. A rule can be executed perfectly and still lose, because the program only does what the rule says. The question "is algorithmic trading profitable" therefore has the same answer as "is a recipe tasty": it depends on the recipe, and the oven does not make it good.

What decides it for a retail account, roughly in order:

Whether the rule has an edge at all, meaning a price behaviour that repeats often enough to be worth trading.
Whether that edge was measured on data the rule was not tuned on, because a result measured only on the tuning data is partly a fit to noise.
Whether the edge survives the costs of trading it, which are paid on every trade and never appear on a price chart.
Whether the trader can hold the strategy through its bad months, which needs enough capital and a plan for when to stop.

Where does the edge come from, and where does it leak?

An edge is a repeatable price behaviour that is worth more per trade than the cost of trading it, and it leaks through two holes: the costs of trading (commission, fees and slippage), and overfitting that made the backtest look better than the behaviour really is. Most retail strategies that fail live had a real-looking edge in the backtest and lost it to these two leaks, not to a change in the market.

Think of the gross edge as water in a bucket. Each leak is a hole. A bucket with a small amount of water and two holes is empty before you carry it anywhere.

Trading costs are everything a real order pays that a price chart does not show. Commission and fees are the broker's commission plus the exchange, clearing and regulatory fees, charged on entry and on exit; they are fixed per contract, so they hurt strategies with a small move per trade the most. Slippage is the difference between the price the backtest assumed and the price a real order fills at; market and stop orders usually fill at least one tick worse than the trigger price, more at the open and around news. Trading costs in backtests explains how to estimate both parts honestly.
Overfitting is the part of a backtest result that comes from fitting the noise of the test data rather than a real behaviour. It is the leak most traders forget, because it does not appear as a line item anywhere. Bailey, Borwein, López de Prado and Zhu (2014) showed that with enough parameter combinations tried on the same data, almost any in-sample result can be reached while the expected out-of-sample result falls. Backtest overfitting covers the research in plain terms.

Hypothetical example

A strategy on Micro E-mini S&P 500 futures (MES) shows a gross average trade of +$12 per contract in the backtest. Commission and fees are $2.40 per round trip. Slippage of one tick on each side, at $1.25 per tick (check the current tick value with the futures contract calculator or the exchange's contract page), is $2.50. The same rule tested on data it was not tuned on earns $8 gross per trade instead of $12, so overfitting took $4.00. What is left is $12 − $2.40 − $2.50 − $4.00 = +$3.10 per trade. The edge is still positive, but a quarter of its original size, and one extra tick of slippage per side would cut it again by $2.50.

The hypothetical example above, per trade: the gross edge shrinks at each leak, and the overfitting leak is the one that never appears on a broker statement.

The practical test is simple: the gross average trade should be several times the round-trip cost before you trust it. If the average trade and the cost are the same size, the strategy is a coin flip after fees.

What do backtest numbers mean once you trade live?

Out-of-sample result, maximum drawdown and profit factor each answer a different question, and a live account usually sees a worse drawdown and a lower profit factor than the in-sample backtest showed. Reading each number for the question it answers prevents the most common disappointment: expecting the smooth in-sample curve and getting the rough one.

Three definitions first. An in-sample result is the backtest result on the data that was used to choose the rules and parameters. An out-of-sample result is the result on data that was not used for any of those choices. Maximum drawdown is the largest fall of the equity curve from a high point to the next low point before a new high, usually given as a percentage of the account. Profit factor is the one formula in this article:

Profit factor = gross profit ÷ gross loss, where gross profit is the sum of all winning trades and gross loss is the sum of all losing trades, both after costs. A value of 1.0 means wins and losses cancel out; 1.5 means the strategy won $1.50 for every $1.00 it lost.

Number	Question it answers	What usually changes live
In-sample result	How well did the rules fit the data they were chosen on?	Almost always too good; treat it as an upper limit, not a forecast.
Out-of-sample result	Did the same rules work on data they never saw?	The closest thing to live; expect live to sit near or below it.
Maximum drawdown	How much did the account fall before recovering?	Live drawdowns are often deeper, because the worst case has not happened yet in a short history.
Profit factor	How much is won per dollar lost?	Falls when slippage is higher than modelled; compare with the out-of-sample value, not the in-sample one.

Hypothetical example

The same strategy shows an in-sample profit factor of 1.9 with a maximum drawdown of 8%. On the out-of-sample years the profit factor is 1.3 and the maximum drawdown is 14%. A trader who sized the account for the 8% drawdown would be stopped out by a normal out-of-sample month. The 14% is the number to plan around, and live trading may still go deeper.

In-sample and out-of-sample curves of the same hypothetical strategy. The rougher out-of-sample curve is the one that resembles live trading.

The reliable way to get an out-of-sample number is to split the history before you start tuning, or to run walk-forward analysis, which repeats the split many times across the history. Harvey and Liu (2015) recommend reducing a reported Sharpe ratio to account for how many strategies were tried before this one was picked; the same thinking applies to profit factor and drawdown.

Is algo trading better than manual trading?

Neither is better in general: a systematic strategy is good at discipline, testing and running many markets at once, while a manual trader is good at reading context and stepping aside when conditions change, so the honest choice depends on which strengths the trader can actually use.

A market regime is a stretch of time in which a market behaves one way, for example a long trend or a tight range. A rule tuned on one regime can struggle in the next, and a program does not notice the change by itself. A manual trader often notices it in a day, but the same trader also widens a stop under pressure, which a program never does.

Task	Systematic	Manual
Discipline under pressure	Strong: the rule does not change its mind after two losses	Weak: stops get moved, sizes get doubled
Testing an idea before trading it	Strong: the rule can be backtested on years of data	Weak: a discretionary method cannot be tested the same way
Running several markets at once	Strong: one program watches ten markets as easily as one	Weak: attention runs out at two or three charts
Reacting to unexpected news	Weak: the rule trades through the news unless told not to	Strong: a human can stand aside
Adapting to a new regime	Depends: needs revalidation on a schedule	Strong: context is read directly
Time at the screen	Low: research and review, not watching	High: present for every trade

Systematic and manual trading are strong at different jobs. The useful question is which jobs to hand to the program, not which side wins.

Many retail traders end up with a split: rules decide entries, exits and size, and the human decides on a schedule whether each strategy is still behaving as tested and whether to keep running it. That split uses the strengths of both columns.

How much account size and time does it realistically take?

The account must cover the exchange margin plus a drawdown larger than the backtest's worst one with room left to keep trading, and the time budget is weekly research and review rather than screen time, which is what makes the approach workable with a job. The drawdown calculator gives the maximum drawdown and recovery of an equity series, which is the middle layer of this sum.

Margin is the deposit the exchange requires per open contract, and the exchange sets and changes it; check the exchange's contract page for the current value. Margin is only the first layer. Build the account from three layers:

Exchange margin for the contracts the strategy holds at the same time.
The out-of-sample maximum drawdown in money, not percent, for the same contract count.
A safety layer so the account survives a drawdown deeper than the tested one. A common practice is to size for about twice the tested drawdown, because a few years of history rarely contain the worst case.

Hypothetical example

One Micro E-mini contract. The exchange margin in this example is $1,500. The out-of-sample maximum drawdown is $2,000. The safety layer adds another $2,000, so the trader can sit through a drawdown twice the tested size and still hold the position. The minimum account for one contract is $1,500 + $2,000 + $2,000 = $5,500. The weekly time budget in this example is about three hours of research on new candidates and about thirty minutes of review comparing the live trades of the week with the backtest.

Hypothetical example: the account is three stacked layers (margin $1,500, out-of-sample drawdown $2,000, safety layer $2,000, total $5,500); the backtest drawdown is only the middle one.

If the three layers add up to more than the account, trade fewer contracts or a smaller contract rather than cutting the safety layer. Cutting the safety layer is the same as deciding to stop at the first drawdown that is slightly worse than the test. To see how many contracts or how small a contract the account can carry, the position size calculator turns account risk and stop distance into a contract count.

Which three habits separate traders who keep an edge?

Traders who keep an edge put realistic costs into the search from the first run, validate on unseen data before risking money, and compare live trades with the backtest on a schedule so they notice when a strategy stops behaving as tested. The three habits form a loop, and skipping one breaks the loop.

Search with real costs included. Set commission, slippage and next-bar execution before the first candidate is evaluated. Next-bar execution means a signal computed on the close of one bar is filled on the next bar, not on the close it was computed from. A search that ranks candidates without costs fills the top of the list with strategies that trade too often, and adding costs afterwards removes most of them. AlphaPrime (Valynox) evaluates each candidate with fees, slippage and next-bar execution in its backtesting, so the ranking already reflects what could be traded.
Validate on unseen data before trading. Run walk-forward analysis, read the out-of-sample result, and set a drawdown threshold the strategy must pass. AlphaPrime's strategy validation applies custom thresholds and stops the later checks when a strategy fails one, which keeps the standard the same for every candidate.
Trade small and compare live trades with the backtest on a fixed schedule, for example a short weekly check and a fuller monthly comparison. Write down the live average trade, win rate and drawdown next to the out-of-sample values. When the live numbers drift outside the tested range for longer than a normal bad stretch in the backtest, go back to step 1 instead of adjusting the parameters by hand.

The three habits as a loop. The broken arrow shows the most common shortcut: a strategy that goes from search straight to live money without validation.

What this means if you use AlphaPrime: the costs and next-bar execution you set in the research settings apply to every candidate from the first evaluation, the validation stage gives you the out-of-sample and drawdown numbers to size the account with, and the Performance Report is where you compare the live month with the tested one.

Frequently asked questions

What is algo trading, in one sentence?

Algo trading is trading in which written rules decide every entry, exit and position size, and a program executes them without a human choosing each trade. The rules can be as simple as a moving-average cross or as complex as a multi-market portfolio. The program only supplies discipline and speed; the rules supply any edge.

Does algorithmic trading work for retail traders with a small account?

It works to the extent that the tested rule keeps a positive average trade after commission and slippage on data it was not tuned on. Small accounts are more sensitive to fixed costs per contract, so the average trade must be larger relative to the round-trip cost. The account must also cover the exchange margin plus a drawdown larger than the tested one, which is why micro contracts are common starting points.

Which backtest number should I trust most before going live?

The out-of-sample result, meaning the result on data that was not used to choose the rules or parameters. Read its profit factor and maximum drawdown rather than the in-sample ones, and plan the account around that drawdown. Live trading usually sits near or below the out-of-sample curve.

How much time does systematic futures trading take each week?

Less screen time than manual trading, because the program watches the market. The time goes into research on new candidates and a regular review comparing live trades with the backtest. A few hours of research and a short monthly comparison are enough for many part-time traders, as long as the comparison actually happens on a schedule.

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.