8 algorithmic trading strategies: the upside and downside of each
“Algorithmic trading strategies” is a name that covers a lot of very different things. Two managers can both tell you they run one and be doing completely different things with your money. The name alone tells you almost nothing.
What matters shows up later. Does the strategy still work with real money on the line? How much can it handle before it stops working? What does it cost to run, and how careful is the team behind it?
For each of the eight main families you’ll get three things: what it is, in plain terms; the one problem most likely to trip it up; and how much you can check from the public record before you trust anyone’s sales pitch. That last part matters more than it sounds, and it is where most of the money gets lost.
- The main algorithmic trading strategies, in one list
- Other terms you’ll hear: HFT, arbitrage, breakout, AI
- What “algorithmic trading strategies” means
- Why the name of a strategy tells you so little
- The eight families at a glance
- Trend following: expect years with no profit
- Mean reversion: almost nobody runs it alone
- Statistical arbitrage: the more it grows, the less it makes
- Market making: earning the gap between buy and sell prices
- Volatility and carry: steady income until a crash
- Event-driven: little proof the automated version works
- Factor investing: great on paper, weak for years
- Momentum: it works, but it can crash hard
- What changes when you trade these strategies on gold
- Why results are worse with real money
- What weak providers hide
- Questions to ask about any trading strategy
- Common questions
The main algorithmic trading strategies, in one list
There are eight main families of algorithmic trading strategies. Here is each one in a sentence, plus the catch: the specific thing that tends to go wrong with it.
- Trend following. Buys what’s going up and sells what’s going down. The catch: it can go years without making money, and that is normal for this strategy.
- Mean reversion. Bets that a price which has moved too far will snap back. The catch: almost nobody runs it on its own at a large scale, so a fund that does needs a good reason.
- Statistical arbitrage. Makes money from small, short-lived price gaps between investments that usually move together. The catch: the versions that make the big returns can’t take much money, so the ones you can buy earn far less.
- Market making. Quotes a buy price and a sell price all day and earns the small gap between them. The catch: it depends on fast technology, and plenty of firms use the name without having it.
- Volatility and carry. Sells a kind of financial insurance that pays a steady premium in calm markets. The catch: when the market crashes, you’re the one who has to pay out.
- Event-driven. Trades around known events like company earnings or changes to a stock index. The catch: there is very little public proof that the automated version works over the long run.
- Factor investing. Buys baskets of stocks that share a trait, like “cheap” or “high quality.” The catch: a style that works over fifty years can still lose money for fifteen of them.
- Momentum. Buys what’s been rising and sells what’s been falling. The catch: run on its own, it can fall hard when the market suddenly turns.
One thing to hold onto as you read. A strategy having little public proof doesn’t make it bad. It makes it something to ask harder questions about, because the record won’t answer them for you.
Other terms you’ll hear: HFT, arbitrage, breakout, and AI
You’ll also run into four other terms: high-frequency trading, arbitrage, breakout trading, and AI or machine-learning strategies. None of them is a ninth family. Each is a different angle on the eight above. Here’s where each one fits, and where the word can fool you.
- High-frequency trading (HFT) is about speed rather than a strategy of its own. It’s how market making and some fast arbitrage get done. The money comes from technology and being first in the queue, far more than from any clever call on price. We cover the machinery on our high-frequency trading page.
- Arbitrage, in the textbook sense, means a risk-free profit. In the real world almost nothing sold as arbitrage is risk-free. It’s usually statistical or “latency” arbitrage under a cleaner-sounding word. When someone says arbitrage, ask which kind they mean.
- Breakout trading is trend following’s way of getting in: buying the moment a price pushes past its recent range. On its own it has the same weakness trend following does. Most breakouts fizzle, and the wait between the good ones is long.
- AI and machine-learning strategies describe how the trading decisions get made, rather than what’s being traded. A smart model still ends up inside one of the eight families, and it inherits that family’s limits. The name doesn’t change what it is underneath. Our AI trading page covers what it does and doesn’t add.
What “algorithmic trading strategies” means
An algorithmic trading strategy is any set of rules a computer follows to decide what to buy or sell and when. Some make a direct bet on where a price is heading. Others make money from the gaps and relationships between related investments, without caring much which way the market goes overall.
The name covers a huge range. Fast-trading arbitrage books, slow trend programs that hold positions for months, market-making desks, factor strategies tucked inside big stock portfolios, and insurance-like products that feel safe until they suddenly aren’t. All of it gets sold with the same three words.
Most algorithmic trading strategies marketed this way never make it into big institutional portfolios. The menu you’re shown is far larger than the list worth owning.
Why the name of a strategy tells you so little
Two managers can both say “we run mean reversion” and build something almost unrelated. One might hold a balanced set of paired bets for a few days at a time. Another might trade a fast-moving basket of stocks and be out within the hour. Same name. Different businesses. Different risks.
Underneath, every one of these algorithmic trading strategies is three decisions stacked together. First, a signal: what to trade and when. Second, sizing: how big each position gets and how concentrated the whole thing is. Third, execution: whether the idea still makes money once you count the cost of trading it.
Weak providers spend nearly all their time talking about the first one, the clever signal. Experienced buyers learn to push hardest on the other two, because that is where good ideas quietly die.
The eight families at a glance
The table below lays out the same eight algorithmic trading strategies side by side, in plain words.
Two columns matter most: how much public proof exists that a strategy works, and whether it is ever sold on its own. Thin proof and a “rarely” are the signals to slow down and ask more questions.
| Strategy | Money it can handle | Trading speed | Market-swing risk | Public proof it works | Sold on its own? |
|---|---|---|---|---|---|
| Trend following | Large | Slow | Medium | Strong | Yes |
| Mean reversion | Moderate | Fast | High | Thin | Rarely |
| Statistical arbitrage | Limited | Fast | High | Good | Sometimes |
| Market making | Moderate | Extreme | Low | Strong | Yes |
| Volatility & carry | Moderate | Medium | High | Thin | Rarely |
| Event-driven | Moderate | Medium | Medium | Thin | Rarely |
| Factor investing | Large | Slow | High | Strong | Yes |
| Momentum | Large | Medium | Medium | Some | Rarely |
Reading the table: “Large” capacity means it can take a lot of money before it stops working. “Thin” public proof and a “Rarely” in the last column are the families to question hardest.
Trend following: expect years with no profit
Trend following is easy to describe: buy what’s going up, sell what’s going down, and stay in until the move runs out. The bet is that big price moves tend to last longer than pure chance would suggest. The hard part isn’t the idea. It’s the waiting.
- Man AHL has run trend strategies for more than thirty years. In a field where most products don’t survive a single bad stretch, lasting that long says more than any single year’s result.
- What an experienced investor looks at first is how a manager handled the flat, boring years, rather than last year’s headline number.
A strong trend strategy can sit flat for three to five years between its big winning runs. AQR’s 2023 research on this kind of strategy makes the case that those flat years are exactly when it’s doing its job, waiting to protect you when the next big move arrives.
The payoff shows up when it’s needed most. 2022 was trend following’s best year in over a decade: as both stocks and bonds fell together, the SG Trend Index, which tracks the big trend funds, returned 27.3% (Hedgeweek).
The gains came from riding falling bonds, a rising dollar, and climbing energy prices while most portfolios sank. The managers worth a meeting are the ones who warn you about the flat years up front.
Mean reversion: almost nobody runs it on its own
Mean reversion bets that a price which has stretched too far will come back toward normal. The idea is old and easy to grasp. Making money from it reliably is hard, and very few firms sell it as a standalone product.
Look through the well-known quant firms and you’ll struggle to find many whose whole business is mean reversion. Most of it lives inside bigger statistical-arbitrage operations, as one signal among dozens, with the risk controlled by machinery built for a wider purpose.
- Almost no institutional money runs mean reversion on its own. It nearly always sits inside a larger statistical-arbitrage operation.
- So if someone offers you a pure mean-reversion fund, that’s an unusual choice. The first question is simple: why run it alone when almost everyone else runs it embedded?
A good firm will have a real, structural answer to that question. A weaker one will change the subject. That difference tells you a lot.
Statistical arbitrage: the more it grows, the less it makes
Statistical arbitrage, or “stat-arb,” makes money from tiny, short-lived price gaps between investments that usually move together. The maths is clever. But the thing that decides what you take home is size: the more money a stat-arb strategy runs, the harder it becomes to find enough of those gaps, and the lower the returns fall.
The clearest proof comes from one firm running both versions side by side. Renaissance Technologies’ Medallion fund is the most famous quant fund in history. It closed to outside investors in 1993, caps itself near $10 billion, and hands its profits back every year to stay small.
A fair pushback is that maybe Renaissance is being choosy rather than hitting a real limit. So look at what the same firm does with everyone else’s money.
Medallion is fast, heavily traded, and open only to employees. Renaissance also runs large funds anyone can buy, which hold positions for months instead of days. In 2020, Medallion gained 76%. That same year, the public funds lost money: Renaissance Institutional Equities fell 22.6%, and its Diversified Alpha fund fell 33.6%. Same firm, same researchers, same year, opposite outcomes.
- Medallion returned about 76% in 2020, while Renaissance’s own public funds lost 22.6% and 33.6% (Institutional Investor). Medallion holds trades for days and is capped near $10 billion; the public funds hold for months and run far more money.
- Medallion has compounded near 39% a year. The scalable version, the one you can buy, earns a fraction of that and can fall hard in a bad year.
So stat-arb does scale. What doesn’t scale is the version that produces the huge numbers. The firm grew by gathering outside billions into its slower, lower-return funds while keeping the crown jewel small and shut.
When someone offers you a stat-arb strategy, don’t ask whether the maths is good. Ask which end of that trade-off you’re buying: the small, fast, closed kind, or the big, slower kind that gives most of the return back.
And 2020 is a reminder that even the best quant shop’s public fund can drop a third in a hard year, so “quant” and “safe” are far from the same word.
Market making: earning the gap between buy and sell prices
Market making is different from everything else on this list. The money doesn’t come from guessing where prices go. It comes from quoting a price to buy and a price to sell all day, and earning the small gap between them thousands of times over. What it takes is fast technology and enough capital to hold inventory.
- Virtu Financial, a public market maker, reported a single losing trading day out of 1,178 between January 2009 and September 2013, and made $2.145 billion in adjusted net trading income in its 2025 financial year.
- That pattern, thousands of small wins and almost no losing days, is the clearest sign you’re looking at a real market maker rather than something using the name.
The business has only grown. In 2024, Jane Street, a private trading firm most people have never heard of, earned a record $20.5 billion in net trading revenue and handled roughly a tenth of all US stock trading (Bloomberg). That is the scale of a real market-making operation.
If a product is sold as “market making” but its results look more like an ordinary fund, with good months and bad months, it probably isn’t market making at all. The shape of the profits gives it away, and firms using the words without the technology behind them are usually selling something else under a better-sounding name.
Volatility and carry: steady income until a crash
Volatility and carry strategies sell something people want most when things go wrong, a kind of financial insurance. In calm markets the premiums roll in and it looks like easy, steady income. In a crash, you’re the one who has to pay the claims.
The standalone products built to run this strategy were mostly wiped out in a single day in early 2018, when volatility spiked and the bets went the wrong way fast. What survives today mostly runs quietly inside larger multi-strategy funds, rather than as a product you buy on its own.
- On 5 August 2024, the VIX, Wall Street’s fear gauge, spiked to around 65 during the day, one of its sharpest jumps on record (BIS). The trigger was a sudden unwind of the yen carry trade.
- Sellers of volatility who had banked months of calm premiums gave much of it back in a single morning. The 2018 lesson repeated, six years later.
So a standalone version being offered today is a deliberate choice that should come with a clear explanation of how it survives the next spike. A careful manager has that answer ready. A weaker one steers the conversation elsewhere.
Event-driven: little proof the automated version works
Event-driven strategies trade around specific events: a company’s earnings, a merger, a stock being added to or dropped from an index. There’s a hands-on version, where humans make the calls, and that’s where most of the money in this area sits. The fully automated version is the smaller, less-proven cousin.
Of all eight families, the automated event-driven strategy has the least public evidence behind it. Long-running firms built purely on it are hard to find. Most famous event-driven investors are people making judgment calls on mergers or troubled companies, rather than machines.
- The automated version of event-driven trading has the weakest public evidence of any family here. Long-running firms built on it are rare.
- Most well-known event-driven managers rely on human judgment. The pure automated form is the exception, so it deserves extra scrutiny.
There’s a reason for the thin record. Many of these patterns, like a stock slowly rising after good earnings, have faded as more people piled in to trade them. And there aren’t many big events to learn from, which makes it hard to prove a system holds up. So if you’re offered a purely automated event-driven fund, the fair question is why automate it, when the people doing this best still do it by hand.
Factor investing: great on paper, but it can lose for over a decade
Factor investing buys baskets of stocks that share a trait, like being cheap, high quality, smaller, less volatile, or recently rising. Researchers have studied these traits for decades, and the long-run case for them is solid. The problem is what happens in between.
A trait that pays off over the very long run can lose money for painfully long stretches. FTSE Russell’s research found one such stretch that lasted 184 months, from July 2006 to November 2021, where growth stocks beat value.
- FTSE Russell measured a single 184-month stretch, July 2006 to November 2021, where one popular style lost to its opposite. Fifteen years is most of an investing lifetime.
- Even well-accepted styles can lose for over a decade. Turning a proven idea into a fund you can live with is the real work.
Be wary of a factor product that shows you a fifty-year chart while hoping you’ll invest based on the last good year. A straight-talking manager tells you plainly that long losing stretches are part of the deal. A weaker one lets the chart imply otherwise.
Momentum: it works, but it can crash hard
Momentum buys whatever has been rising and sells whatever has been falling. The evidence that this works is some of the strongest in all of finance. Yet almost nobody sells it as a standalone product, and there’s a good reason.
Nearly all the momentum in big portfolios is blended together with other styles like value and quality. Blending smooths out momentum’s ugliest habit: when the market suddenly turns after a big fall, momentum can crash hard, because it’s betting against exactly the beaten-down names that suddenly rocket back.
- Nearly all institutional momentum is blended into multi-style portfolios rather than sold on its own. Blending is the norm for a reason.
- Pure momentum keeps the nasty crash risk that blending softens. A good manager can explain how they handle it. A weaker one can’t.
So a pure momentum fund is an unusual choice. The fair question is why go standalone, when both the research and the professionals point toward blending?
What changes when you trade these strategies on gold
Gold is a single asset, not a market full of them. That rules out several of the strategies above straight away. Statistical arbitrage compares one thing against another. Factor investing and momentum pick the best few out of hundreds. Market making quotes prices across a whole book. With only gold to trade, there is nothing to compare it against.
That leaves trend following, mean reversion, volatility and event trading. Gold behaves in its own way: it jumps on news, sits quiet for weeks, and can move further in a day than most currency pairs. The same rules need different settings here, and some stop working altogether. Our page on gold (XAUUSD) trading strategies goes through the ones that do work, and what to ask a seller for.
Why algorithmic trading strategies do worse with real money than in a backtest
A “backtest” is a test run of a strategy on past data, showing how it would have done. The gap between a great backtest and the real-world results of algorithmic trading strategies is the most misunderstood part of this whole subject. The problem usually isn’t a faked test. It’s that trading for real is harder than trading on paper.
Take a mean-reversion strategy that looks brilliant across ten years of history. Run it live and the cost of borrowing shares comes in higher than assumed, your orders get worse prices on wild days, and the opportunity fades before you’ve finished buying. What looked like a clear advantage turns into noise once real costs land.
Fast stat-arb hits a similar wall. On paper it glides between hundreds of positions for free. In the real world, trading that much moves prices against you and your place in the queue starts to matter. Even small, realistic trading costs can flip the whole thing from profitable to pointless.
It cuts the other way too. A plain, medium-speed trend strategy on futures might look boring next to something more exotic. But if it trades things that are easy to buy and sell, follows stable rules, and is run by a disciplined team, its real-world results can be far more dependable. The boring one is often the one you can own.
Costs, trading volume, size limits, and speed aren’t small technical details. For a lot of algorithmic trading strategies, they decide whether there’s any advantage left at all once you leave the spreadsheet.
What weak providers hide
Weak providers of algorithmic trading strategies rarely admit a strategy is fragile. They tuck the fragile part somewhere the sales deck doesn’t point. Here’s where to look, and what a strong answer looks like next to what a weak one tends to bury.
| What to check | What a strong answer looks like | What a weak one hides |
|---|---|---|
| Costs | Every cost shown period by period: trading, borrowing, financing | Only the final “net” number, with no breakdown you can check |
| Testing | Results on data the strategy never saw while being built, kept separate | One long history chart, no separation, no discussion of stability |
| Signals | A short list of reasons the strategy works, each one explained | Dozens of signals and no story, with complexity sold as proof of genius |
| Risk | How bad the losses can get, and how it behaves when markets seize up | A single daily “wobble” number and nothing about the worst case |
| Size limits | A real number for how much money it can handle, and the reasoning | “Plenty of capacity” with no figure and no method behind it |
One quick tell. When a manager explains their success mostly through how clever the model is, rather than how carefully they run it day to day, that’s often where the weakness lives. The research matters. So does who’s allowed to change the settings, how mistakes get handled, and what happens after a bad month.
Questions to ask about any trading strategy
Judging algorithmic trading strategies isn’t about one killer question. It’s about a pattern. The people worth trusting describe their weak spots clearly and the same way twice. The ones to avoid get vague exactly where it matters. Here are the questions that surface that pattern, grouped by what they tell you.
Is the track record real?
- How much of the record is real trading versus a simulation?
- How did you decide what counts as a fair “unseen” test before you started?
- How does the strategy behave in conditions that look nothing like the recent past?
- How many versions did you try before landing on this one?
Does it survive the real world?
- What exactly does it trade, and how easy is that to buy and sell in a panic?
- How do you measure real trading costs, not the ones you assumed?
- How much do the results depend on perfect timing?
- Who’s allowed to change the settings or switch it off, and what gets reviewed after a bad month?
Does it fit the rest of my money?
- What job is this strategy meant to do alongside everything else I hold?
- How does it behave when stocks fall, when cash gets tight, or when rates jump?
- Do the returns come from one lucky condition, or from several different sources?
- Is the fee fair for how complex and how limited the strategy is?
Want to go deeper on any of these? We cover the day-to-day discipline and safety controls on our risk management and operational discipline pages, how to trust a backtest on our backtesting page, the modelling behind the signals on our quant trading page, and the firms that run these strategies on our quant managers page. The tighter your questions, the faster the weak offers fall away.
No one answers every question perfectly, and you shouldn’t expect that. What you’re listening for is honesty about the weak spots, said the same way twice.
Common questions about algorithmic trading strategies
Most algorithmic trading strategies aren’t worth investing in
Take trend following, the family with the best public record of them all. One industry index counted 356 active programs in 2025. The benchmark that tracks the ones big institutions trust holds just ten. That is roughly thirty-five marketed strategies for every one the professionals rely on, and that is the best-documented family. The thinner ones are almost certainly worse.
That gap is the reason The Review exists:
- Curated. Only algorithmic trading strategies that have earned their place make the list (most do not).
- Capital protection first. Chosen for protecting your money and growing it at a steady pace rather than chasing the biggest headline number.
- Clear on access. Each entry says plainly whether it is open to the public or closed to new money.