VALYNOX
VALYNOXSTRATEGY RESEARCH TECHNOLOGY
AlphaPrimeAlphaPrime
Sign inFree 7-day trial
ACADEMY / Strategy generation

Systematic trading explained: how a rules-based strategy is built

Systematic trading means every decision (market, entry, exit, size) follows rules written before the first trade, so the rules can be tested on past prices and repeated unchanged. Judgment moves to designing, accepting and stopping the rules; the build runs idea, rules, backtest with costs, validation on unseen data, live monitoring.

By · Published · updated · Editorial policy

Contents
  1. What is systematic trading?
  2. Systematic vs discretionary trading: where does judgment still enter?
  3. How is systematic trading different from algo trading?
  4. How do you build a systematic strategy, step by step?
  5. Is systematic trading suitable for retail traders?
  6. Where do systematic strategies fail?
  7. What maintenance do rules need, and what can a backtest promise?
  8. Frequently asked questions
  9. References

What is systematic trading?

Systematic trading is trading in which every decision, which market to trade, when to enter, when to exit and how much to trade, follows rules written down before the first order. Because the rules are fixed, you can test them on past prices, repeat them unchanged, and judge them by their record instead of by memory of a few trades.

A systematic strategy is a complete set of these rules: one answer for each of the four decisions, with no blank left for the trader to fill in during the session. Think of a recipe. A cook who follows a written recipe can make the same dish on Monday and on Friday, change one ingredient, and taste the difference. A cook who seasons by feel cannot say which change made the dish better. A backtest is the test kitchen: you run the written rules over past prices, trade by trade, with commission and slippage charged, and see how they would have behaved. That test is only possible because the rules are written down.

A complete rule card written before the first trade: market NQ daily bars, entry when price closes above the adaptive average, exit when it closes below it, size one contract, each field checked as filled A rule for every decision Written before the first trade Market NQ daily bars Entry Close above average Exit Close below average Size 1 contract No blank left to fill in live. So: write all four before trading.
A strategy is systematic when all four fields are filled in before the first trade. An empty field is a decision the trader will make live.

The claim this article defends is simple: the value of systematic trading is not that rules are smarter than people. It is that rules can be tested before money is at risk and checked against that test afterwards. Keep that in mind when you read the NQ numbers further down.

Systematic vs discretionary trading: where does judgment still enter?

Judgment does not leave systematic trading; it moves from each trade to three moments: designing the rules, accepting them after testing, and stopping them. Discretionary trading is trading in which the trader decides each entry, exit and size at the moment of the trade, using the chart, the news and experience. The skill is real, but it lives in the trader's head. It cannot be backtested, and two traders looking at the same screen will take different trades.

In systematic trading the same skill is spent earlier. The trader decides which pattern is worth a rule, what numbers the rule uses, what test result is good enough to trade it, and what live result is bad enough to switch it off. Between those moments, every trade is taken by the rule, including the trades that look wrong on the screen. Skipping one because it looks wrong turns the record back into discretion, and the backtest no longer describes what is being traded.

DecisionDiscretionary tradingSystematic trading
Which setup to takeDecided live, trade by tradeDecided once, when the rules are written
Position sizeDecided live, often by convictionFixed rule, for example one contract or a set risk per trade
How the method is judgedMemory of recent tradesWritten test: average trade after costs, drawdown, trade count
When to stopWhen losses feel too largeWhen a stop rule written in advance is breached
Two timelines: discretionary trading makes a judgment call on every trade, while systematic trading uses judgment only to design, accept and stop the rules, and every trade in between follows the rules only Where judgment enters Diamonds: judgment. Dots: rules only. Discretionary every trade Systematic rules only Design Accept Stop Judgment moves; it does not leave. So: use it at three points only.
Both styles use judgment. Discretionary trading spends it on every trade; systematic trading spends it at three points and lets the rules take the trades in between.

How is systematic trading different from algo trading?

Systematic describes how decisions are made; algo describes who sends the orders, so they are two separate choices. Algorithmic trading is trading in which a program places and manages the orders. A trader can follow written rules and place every order by hand: the Turtle traders of the 1980s worked from a written rule set and sent their futures orders through brokers, long before electronic order entry (Covel, 2007). A program can also execute a discretionary trader's clicks, routing and managing an order that the trader chose by eye.

Decisions made byOrders sent by handOrders sent by a program
Fixed rulesTrader follows a checklist each morningAutomated strategy running on a platform
JudgmentClassic discretionary traderTrader clicks, program routes and manages the order
A two by two grid: rules sent by hand is a trader with a morning checklist, rules sent by a program is an automated strategy on a platform, judgment sent by hand is the classic discretionary trader, judgment sent by a program is a trader clicking while a program routes the order; only the automated strategy is both systematic and algo Two separate choices Only the gold cell is both Sent by hand Sent by program Rules Judgment Checklist each morning Automated on a platform Discretionary classic trader Trader clicks program routes
Systematic and algo are independent choices, which gives four combinations. Only the top-right cell is both at once.

Most systematic futures strategies today run as algos, for a practical reason: a rule that fires at 02:30 or on a 5-minute bar is hard to follow by hand without missing signals or hesitating. The program adds speed and discipline; it does not add an edge. Automating a rule that was never tested only makes it lose faster and more consistently. What the program side costs, and why most rules lose even when executed perfectly, is covered in algo trading: what it is, what it costs and why most rules lose.

How do you build a systematic strategy, step by step?

Build a systematic strategy in five steps that run in one direction: idea, written rules, backtest with costs, validation on data the rules never saw, then live monitoring against the backtest. A strategy that fails validation goes back to the idea stage, not to the rules stage.

  1. Idea. One reason the market might move, stated in a sentence. "After a strong trend day, the next day tends to continue" is an idea. "Trade NQ" is not. If you need starting points, three automated trading strategies with futures examples shows what a testable idea looks like.
  2. Written rules. Turn the idea into the four decisions with exact numbers: bar size, indicator length, the condition that opens a trade, the condition that closes it, the size. The test of a written rule is that a program could read it without asking a question.
  3. Backtest with costs. Run the rules on history with commission, slippage and next-bar execution from the first run, not added at the end. Read three numbers: average trade after costs, maximum drawdown, and the number of trades.
  4. Validation on unseen data. Validation is the set of checks run on data the rules were not tuned on: hold back part of the history, run the rules unchanged, and compare. The split and the trade counts it needs are explained in how to split in-sample and out-of-sample data. Changing a lookback or a stop after seeing the held-back result turns that data into tuning data, which is why a failure returns to the idea, not to the rules. AlphaPrime (Valynox) runs these checks against thresholds you set before the search in strategy validation, and stops the later checks for a candidate as soon as it fails one, so the accept-or-reject decision is made by numbers written in advance rather than by how the curve looks.
  5. Live monitoring. Compare live trades with the backtest on a schedule, against a stop rule written before the first live trade. The last section covers this.
Five steps from left to right: idea, written rules, backtest with costs, validation on unseen data, live monitoring; a pass leads to live monitoring, a fail goes back to the idea, and going back to the rules is crossed out because that is re-tuning Five steps, one direction Validation decides pass or fail Idea Rules Backtest Validate Live pass Not here: re-tuning fail A failed test goes back to the idea. So: never re-tune on the test data.
The five steps run forward. A failed validation returns to the idea; returning to the rules would re-tune them on the data that was meant to judge them.

The box below shows what a step-three result looks like. The run tested NQ daily bars from 2019-01-02 to 2025-06-30, long and short, one contract, with $20 commission and $10 slippage charged on every trade. The rule is a stand-in for this article, not systematic trading itself: an efficiency-adaptive moving average, a line that speeds up in trends and slows down in sideways markets (the KAMA indicator page describes that family of averages), is its only entry and exit, with no stop, target or filter. The search tried 1,024 variants and kept this one for its return against its drawdown. The figures are therefore those of the single best survivor, and picking the winner after seeing all the results flatters it.

Source: AlphaPrime product engine (CPU), run 2026-10-31. NQ daily bars, 2019-01-02 to 2025-06-30, long and short, 1 contract, $20 commission and $10 slippage per trade.

NumberResult
Net profit (closed trades)$193,250
Max drawdown$116,210
Trades59
Win rate64.41%
Profit factor1.751
Average trade$3,275.42
Time in market92.15%
A funnel narrows over a thousand candidates tried in one NQ search down to the one candidate kept; being the best of many, the kept result includes luck and must still pass out-of-sample checks The best of many is flattered One real NQ search, 2019–2025 Many tried 1 kept Best of many includes luck The kept result is flattered. So: validate it on unseen data.
Look at the equity curve and the drawdown together: the deepest dip against the final height is the pair of numbers that selected this candidate out of the search. NQ · daily · 2019-01-02 to 2025-06-30 · CPU engine backtest

Read the box the way step three asks. Fifty-nine trades is a small sample, so an average trade of $3,275.42 can move a long way if a handful of trades are changed. Time in market of 92.15% means the rule was holding a position, long or short, almost all the time, so much of its result is NQ's direction over those years rather than timing. A max drawdown of $116,210 against a net profit of $193,250 is the shape that selected it. None of this says the rule earned that money or that it would again; it says these rules, run on those prices with those costs, produced that record, and the candidate now goes to step four on data it has never seen.

Is systematic trading suitable for retail traders?

Systematic trading suits a retail trader when three numbers fit together: trade frequency, round-trip cost and account size, and the one number that decides it is the average trade after costs compared with the round-trip cost. The round-trip cost is the commission, exchange fees and slippage paid to open and close one position.

Frequency is what makes the same cost small or large. A rule on daily bars captures a large move per trade and pays the round trip a few times a month. A rule on 5-minute bars captures a small move per trade and pays the same round trip many times a day. The cost per trade is identical; the share of the profit it eats is not.

Hypothetical example

Three rules pay the same $30 round trip. A daily-bar trend rule has a gross average trade of $900, so it keeps $870 and costs take about 3 percent of the edge. A 60-minute rule has a gross average trade of $150, keeps $120, and costs take 20 percent. A 5-minute rule has a gross average trade of $40, keeps $10, and costs take 75 percent; one extra tick of slippage on each side turns it negative. All three looked profitable before costs.

Hypothetical example with the same 30 dollar round trip: a daily trade keeps almost all of its edge and costs take 1 percent, an hourly trade of 300 dollars keeps 270 and costs take 10 percent, a 5-minute trade of 40 dollars keeps 10 and costs take 75 percent Costs eat short trades Green: kept. Red: the $30 cost. Illustration Daily 99% kept 1% cost Hourly 90% kept 10% cost 5-min 25% 75% cost Same $30: 1% or 75% of the edge. So: start on daily or hourly bars.
The same $30 round trip is a rounding error against a daily-bar trade and most of the profit against a 5-minute trade. Frequency decides whether a retail account can carry the costs.

Account size is the third number, and the drawdown, not the margin, sets it. An account must hold the strategy's backtested drawdown, with room for a worse one, without a forced exit. To check your own figures, put the win rate and the average win and loss into the free win rate expectancy calculator, which returns the average trade before you subtract the round trip. If the rule trades inside the day, how to test a day trading strategy after costs walks through the same arithmetic for intraday bars, where it matters most. Many retail systematic traders begin on daily or hourly bars for exactly this reason: fewer trades, a larger edge per trade, and a cost estimate that does not have to be precise.

Where do systematic strategies fail?

Three failure points account for most dead systematic strategies: overfitting, costs and regime change, and each has a check you can run before going live. Overfitting is tuning rules so closely to one stretch of history that they describe its noise rather than a repeatable pattern. Bailey, Borwein, López de Prado and Zhu (2014) show that the more variants a search tries, the more the best one's backtest is inflated, which is why the NQ box above, the best of 1,024 variants, has to be read as flattered until unseen data says otherwise.

The cost failure is the one from the previous section: an edge per trade that is real but smaller than the round trip. It hides well in a backtest run without costs or with a single tick of slippage. The third failure is regime change. A regime is a stretch of market behaviour with its own character: trending or sideways, calm or volatile. A rule built on one regime stops working when the market moves to another, and a jump in expected volatility, the kind of move that VIX options track, often marks the border between two regimes.

Failure pointSymptom in the backtestCheck before live trading
OverfittingSteep in-sample curve, flat out-of-sample curveRun the rules unchanged on data they never saw
CostsProfit disappears when slippage is doubledAverage trade after costs divided by round-trip cost
Regime changeOne stretch of history made all the profitProfit by year, and separately for trending and sideways periods
Three failure points with their sign and check: overfitting shows a steep in-sample and flat out-of-sample curve, checked by running on unseen data; costs show profit gone at doubled slippage, checked by average trade divided by round trip; regime change shows one stretch making all the profit, checked by profit by year and regime Three failures, three checks Each one shows in the report Failure Sign Check Overfitting Steep IS, flat OOS Run on unseen data Costs Gone at 2× slippage Avg trade ÷ round trip Regime change One stretch made it all Profit by year, regime
Each failure point has a symptom you can see in the report and one check that exposes it before the first live trade.

Run the checks in this order: unseen data first, because it is the cheapest and catches the most; doubled slippage second; profit by year third. A trend rule that made all of its profit in one strong trending year can pass the first two checks and still fail the third, and a strategy that fails any one of the three goes back to the idea stage.

What maintenance do rules need, and what can a backtest promise?

Rules need two things after launch: a scheduled review of live results against the backtest, and a written stop rule decided before the first live trade. Review once a month, and compare three figures with the backtest: the number of trades taken against the number the backtest averaged per month, the average trade after costs, and the current drawdown against the backtested maximum. A live average trade well below the backtest, with the trade count on target, usually points to slippage worse than assumed; a trade count far off target usually points to a data or session setting that differs from the test.

The stop rule is a sentence written before launch, for example "switch off when the live drawdown passes a stated multiple of the backtested maximum drawdown", with the multiple filled in. When it is breached, the strategy is switched off and the idea goes back to step one; it is not repaired live. Pardo (2008) frames strategy development as this kind of cycle, design, test and re-evaluation, rather than a one-time optimization followed by trading forever.

Illustration of a twelve-month timeline reviewed every month: the live drawdown deepens until it crosses the stop limit written before launch, and at that point the strategy is switched off and returns to the idea stage Stop at the written limit Twelve months, reviewed monthly Review every month Illustration Live drawdown Stop limit Switch off The limit is set before launch. So: switch off, back to Idea.
A live strategy is reviewed on a schedule against its own backtest, and a stop rule written before launch decides when it is switched off.

What a backtest can promise is narrow and useful: it describes how the written rules behaved on past prices under the stated costs and execution assumptions. It does not describe next year. The market can enter a regime the history did not contain, costs can change, and the best candidate of a large search carries a selection bonus that only unseen data can measure. The backtest's job is to let you reject bad rules cheaply and to give a live strategy something to be measured against.

A first thing to do today: take one setup you already trade by feel and try to fill in its four fields, market and bar size, entry, exit, size, with exact numbers. If you cannot fill a field without writing "usually" or "depending on", that field is the first rule to write, and the backtest can only start once it is written.

Frequently asked questions

What does systematic trading mean in simple terms?

Systematic trading means trading from rules written down before the first trade: which market, when to enter, when to exit and how much. Because the rules are fixed, they can be tested on past prices and repeated without change. The trader's judgment goes into designing the rules, accepting them after testing and switching them off, not into each individual trade.

How do I start systematic trading?

Start with one idea stated in a sentence and write it as four exact rules: market and bar size, entry condition, exit condition, position size. Backtest it with commission and slippage included from the first run, then test it unchanged on a part of the history you held back. Trade it live, if at all, only after it passes, and compare live results with the backtest every month. Daily or hourly bars are a practical starting point because the edge per trade is larger relative to costs.

Is systematic trading the same as algorithmic trading?

No. Systematic describes how decisions are made (by fixed rules); algorithmic describes who sends the orders (a program). A trader can follow rules by hand, and a program can execute a discretionary trader's clicks. Most systematic futures strategies run as algos because a program does not miss signals or hesitate, but the program adds discipline, not an edge.

Is systematic trading suitable for retail traders?

It is suitable when the average trade after costs is several times the round-trip cost and the account can hold the backtested drawdown with room for a worse one. Trade frequency decides most of this: a rule that trades a few times a month on daily bars pays the round trip rarely, while a 5-minute rule pays it many times a day. Check the division before anything else.

Do systematic strategies stop working?

Yes, and the three usual causes are overfitting, costs and regime change. A rule that described noise in the history fails on new data; a rule whose edge is smaller than the round trip fails as soon as slippage rises; a rule built for a trending market fails when the market turns sideways. A monthly review against the backtest and a stop rule written before launch decide when to switch a strategy off.

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.
  • Covel, M. W. (2007). The Complete TurtleTrader: The Legend, the Lessons, the Results. HarperBusiness.
  • Pardo, R. (2008). The Evaluation and Optimization of Trading Strategies (2nd ed.). John Wiley & Sons.