What is algorithmic trading? A plain-English guide

Algorithmic trading is trading where the decision to buy or sell comes from explicit rules rather than a gut call. The rules are precise enough that a computer could follow them: "buy when this number crosses that number, sell after five days, never risk more than 2% of the account on one position."

The rules do not have to run on a computer to count. What matters is that they are written down completely. That one property, that the strategy is specified, is what lets you test it against history, measure it, and improve it. A discretionary trader can only remember how their judgment performed. A systematic trader can replay it.

#Why traders use rules instead of judgment

People turn to rules for three practical reasons.

1. Testability. If a rule is precise, you can apply it to years of past data and see what would have happened. That is a backtest, and it is the central tool of systematic trading. We cover how to do one honestly in How to backtest a trading strategy.
2. Consistency. Rules do not get scared after a loss or greedy after a win. Many discretionary losses come from abandoning a plan at the worst moment.
3. Scale. A rule can watch hundreds of markets at once. A person can watch a few.

Rules have a cost too. They only know what you told them. When the world changes, a rule keeps doing what used to work until you notice.

#The anatomy of a trading strategy

Whether it trades stocks, futures or prediction-market contracts, nearly every systematic strategy has the same six parts.

Beginners tend to spend all their time on the signal. Experienced traders know that sizing, costs and risk limits decide whether a modest signal makes money or loses it.

#Common strategy families

Most strategies fall into a handful of families. Each rests on a different belief about why prices move.

• Trend following (momentum). Assets that have been rising tend to keep rising for a while. Buy strength, sell weakness. These strategies lose often in small amounts and win occasionally in large amounts.
• Mean reversion. Prices that move far from a typical level tend to come back. Buy dips, sell spikes. These strategies win often in small amounts and can lose badly when a "dip" turns out to be a new regime.
• Statistical arbitrage. Two related prices usually move together. When they drift apart, bet on them converging.
• Market making. Quote both a bid and an ask, and earn the spread between them. The risk is being on the wrong side when the price moves.
• Event-driven. Trade around scheduled events: earnings, economic releases, elections, games. Prediction markets are event-driven by construction.
• Mispricing and calibration. Look for prices that are systematically too high or too low compared with how often outcomes actually happen. In prediction markets this is a direct, measurable question. See Market calibration.

#The research loop

Systematic trading is mostly research. A typical loop looks like this:

1. Ask a specific question. "Do contracts priced at 80 cents settle YES about 80% of the time?" is testable. "Can I beat the market?" is not.
2. Collect the data. Get every relevant observation, not just the ones that are easy to find. Missing the losers is one of the most common ways to fool yourself.
3. Define the rule precisely. Write down exactly which price you trade at, when, and with what costs.
4. Backtest it. Apply the rule to the data and record every hypothetical trade.
5. Measure it. Compute expected value, win rate, drawdown and how much of the result depends on a few trades.
6. Try to break it. Test on data you did not use to design the rule. Change parameters slightly. If the result disappears, it was probably noise.
7. Decide. Most ideas die here, and that is the point. Research is cheap; losing money to discover a bad rule is not.

#Prediction markets as a place to learn

Prediction markets such as Kalshi list contracts that pay $1 if an event happens and $0 if it does not. That structure makes them unusually good for learning systematic research:

• The price is a probability. A contract trading at 62 cents implies roughly a 62% chance. We explain the details in How prediction market prices work.
• Every contract settles. You get a clean yes-or-no outcome, so you can measure how good the prices were.
• There are many independent-ish events. Games, economic releases and weather markets produce large samples quickly.

The same research habits, such as defining rules precisely, counting costs and testing out of sample, carry over directly to stocks and futures.

#Do you need to code?

Not to start. Professional desks use Python and C++ because they process huge datasets and trade fast. But the thinking of algorithmic trading, which is writing precise rules and testing them on data, fits naturally in a spreadsheet:

• Each row is an observation (a market, a day, a trade).
• Each column is a field or a derived value (=IF(B2<0.7, "buy", "")).
• Summary cells compute the results (=AVERAGEIFS(...), =COUNTIFS(...)).

The advantage of a spreadsheet is that every step is visible. You can click any number and see where it came from. That transparency catches mistakes that hide inside scripts.

Jordan is built on that idea. You describe the analysis in plain English, and Jordan collects the market data onto a sheet and writes the formulas, showing you each change before it is applied. It is a research tool, not a trading bot: it cannot place trades.

#Frequently asked questions

#Is algorithmic trading legal?

Yes. Using rules or software to make trading decisions is legal in the US and most other markets. Specific practices, such as manipulating prices or trading on material non-public information, are illegal whether a human or an algorithm does them. Each exchange also has its own rules on automated access.

#Is algorithmic trading profitable?

Some strategies are, many are not, and edges tend to shrink as more people find them. A backtest showing profit is not evidence of future profit unless it accounts for costs, avoids lookahead bias and holds up on data that was not used to design it.

#How much money do I need to start?

You need no money to research. Backtesting and calibration studies use historical data. When you do trade, start small enough that the costs of learning are affordable, and remember that fees and spreads matter more, proportionally, on small positions.

#What is the difference between algorithmic trading and high-frequency trading?

High-frequency trading is a narrow subset of algorithmic trading that holds positions for milliseconds to seconds and competes on speed. Most algorithmic strategies hold for minutes to months and compete on research quality instead.