Algorithmic trading sounds like it means one specific, exotic thing. It doesn't — it's a spectrum, and most of it is a lot less exciting than the phrase suggests.
The core idea
Algorithmic (or "systematic") trading means a defined, repeatable set of rules decides when to buy and sell — not a human's in-the-moment judgment call. That's the whole definition. It says nothing about speed, complexity, or capital size. A simple rule like "buy when price crosses above its 50-day moving average" is algorithmic trading, executed by hand, on a spreadsheet, if you wanted.
What people usually picture — colocated servers executing thousands of trades a second — is one narrow, capital-intensive corner of it called high-frequency trading (HFT). The overwhelming majority of algorithmic trading, including everything relevant to an individual investor, operates on time horizons of hours to weeks, not microseconds.
Systematic vs. discretionary
The real dividing line worth understanding is systematic vs. discretionary:
- Discretionary trading: a human evaluates the situation each time and decides. Flexible, can incorporate judgment a rule can't articulate — but also inconsistent, subject to fatigue, fear, and after-the-fact rationalization ("I would have sold there anyway").
- Systematic trading: a predefined rule set decides, applied the same way every time. Consistent and testable — you can backtest exactly what it would have done historically — but rigid, and only as good as the rules and the data used to validate them.
Neither is strictly better. A systematic approach's real advantage isn't that computers are smarter than people — it's that a rule can be tested against history before you risk money on it, and its real behavior is precisely knowable in a way a human's future judgment never is.
What algorithmic trading is not
It's worth naming the things the term gets wrongly credited with:
- It doesn't remove risk. A bad rule loses money as reliably as a bad discretionary habit — just more consistently.
- It doesn't require proprietary infrastructure. Most retail-accessible systematic strategies run fine on standard brokerage APIs with hourly or daily bars.
- It doesn't guarantee the backtest is honest. A strategy is only as trustworthy as the backtest that validated it — see how to backtest without fooling yourself for what a real one requires.
What to actually evaluate
If you're looking at a systematic strategy — your own or a vendor's — the useful questions aren't "is this an algorithm." They're: is the rule set specific enough to backtest unambiguously, has it been tested out-of-sample, does it account for realistic fees and slippage, and does its live behavior match what the backtest predicted? A rule is either well-validated or it isn't; calling it "algorithmic" doesn't change which one is true.
This is the same bar Atradian holds its own strategies to — every live signal function is the exact code replayed in the backtest engine, with a per-regime performance breakdown you can inspect rather than take on faith.