If you’ve followed markets for any length of time, you’ve heard the term algorithmic trading. It tends to split a room. Critics argue it puts everyday investors at a disadvantage, while defenders point out that it supplies liquidity to markets – there is almost always a machine (or a person running one) ready to take the other side of your trade. Both camps have a point, and this article isn’t here to referee.<\/p>\n
Algorithmic trading – also called algo trading, automated trading, or sometimes black-box trading – is the use of computer programs to execute trades automatically according to predefined rules. Those rules can be based on price, timing, quantity, or the output of a mathematical model. However simple or sophisticated, the core idea stays the same: a computer follows the plan faster and more consistently than a human ever could.<\/p>\n
Before Tiingo, I traded at a hedge fund, so this world is close to home for me. Below is what algorithmic trading actually is, how it works, the real tradeoffs, and five of the most common strategies. No hype, no scare stories.<\/p>\n
At its core, algorithmic trading grew out of the need for speed. Traders realized decades ago that a computer could react far faster than any human. Watch the market the moment unemployment numbers are released: prices move almost instantly, partly because automated systems are parsing the release and firing orders within fractions of a second. The days of traders shouting orders across a crowded floor are long gone.<\/p>\n
But speed is only half of it. The other half is consistency.<\/p>\n
An algorithm is just a strategy written down precisely enough that a computer can follow it: what to buy, when, how much, and under what conditions to get out. Humans still make every important decision – which strategy to run, what parameters to set, when to intervene – but the moment-to-moment execution belongs to the machine.<\/p>\n
What does that look like in practice? Here’s about the simplest version: a program instructed to buy 100 shares of a stock when its price crosses above its 200-day moving average, then sell if the position rises 15% or falls 5%. That’s it. That’s an algorithm. The rules reference a price and an indicator, the computer watches the market, and when the conditions are met, the order goes out.<\/p>\n
From there, the complexity scales as far as you care to take it: statistical models weighing dozens of inputs, machine learning systems retrained on fresh data, portfolios of strategies running side by side. The definition holds at every level – predefined rules, executed automatically.<\/p>\n
The consistency point deserves emphasis. Even with the best alerts and the best willpower in the world, a human watching charts gets tired, gets distracted, and misses things. A program doesn’t. It watches every tick, across as many symbols as you point it at, and does exactly what the rules say, every single time. (Whether the rules are any good is a separate question. We’ll get there.)<\/p>\n
Every algorithmic trading system, from a weekend Python script to an institutional trading desk, runs the same basic loop:<\/p>\n
Before any of this runs live, there’s a step that separates disciplined algorithmic trading from expensive guessing: backtesting. You run your rules against historical data to see how the strategy would have behaved through past markets – calm ones, volatile ones, and everything in between. It won’t tell you the future, but it will catch a strategy that never made sense in the first place. We wrote a full guide to backtesting stocks<\/a> if you want the details.<\/p>\n And underneath everything sits the data itself. Garbage in, garbage out is a cliche because it keeps being true. If your historical prices are missing dividend adjustments, mishandle splits, or contain bad ticks, your backtest isn’t testing your strategy – it’s testing your data errors. Clean data isn’t glamorous, but it’s the foundation that decides whether everything built on top of it means anything.<\/p>\n One clarification, because it trips people up constantly: high-frequency trading (HFT) is one type of algorithmic trading, not a synonym for it.<\/p>\n HFT sits at the extreme end of the spectrum: servers co-located inside exchange data centers, ultra-low-latency connections, institutional infrastructure built to shave off microseconds. It’s an arms race measured in physics and capital, and retail traders are not competing in it. That’s the reality of that terrain.<\/p>\n But most algorithmic trading is not HFT. A strategy that checks its conditions once a day and rebalances monthly is every bit as algorithmic as a high-frequency market maker. It just doesn’t need a microsecond edge – it needs sound rules, disciplined testing, and reliable data. The speed most strategies actually require is “faster and more consistent than a human,” not “faster than everyone else on Wall Street.”<\/p>\n The pros and cons here are two sides of the same coin. Automation amplifies whatever you give it – discipline and mistakes alike. Here’s a fair accounting of both.<\/p>\n So why hasn’t everyone handed their portfolio to the machines? Because every one of those strengths has a shadow.<\/p>\n None of this makes algorithmic trading good or bad. It makes it a tool – one that rewards care and punishes carelessness, at speed.<\/p>\n There are dozens of strategy families out there, and most serious systems blend several. These five come up the most. They’re descriptions of mechanics, not recommendations.<\/p>\n The same asset sometimes trades at slightly different prices in two venues at the same moment. Arbitrage means buying in the cheaper market and selling in the pricier one, capturing the gap. The gaps are typically tiny and vanish in an instant, which is exactly why this became a machine’s game: spotting and acting on a fleeting price difference is a speed problem, and speed is what computers do. In practice, the fastest infrastructure wins the race, which makes this family the home turf of well-capitalized institutional firms.<\/p>\n Mean reversion rests on one assumption: when a price stretches unusually far above or below its average, it tends to drift back. An algorithm computes the mean and defines bands around it – often using historical prices, technical indicators, and measures of volatility<\/a> – then trades when the price moves outside those bands. The assumption is also the risk. Sometimes a price is far from its average because the world genuinely changed, and it’s never coming back. Quants call that a regime shift, and it’s the scenario every mean-reversion system has to reckon with.<\/p>\n Index funds rebalance on fixed schedules to stay in line with their benchmark indices, which means large, roughly predictable flows of buying and selling around known dates. Some algorithmic traders position around those flows, trading just before or after the rebalancing happens. The catch is built into the premise: anything predictable attracts competition, and widely anticipated flows tend to get priced in quickly by everyone else watching the same calendar.<\/p>\n Volume-weighted average price (VWAP) is a different animal from the strategies above – it’s an execution algorithm rather than a prediction. A VWAP algorithm takes a large order and releases it in smaller slices over the course of the day, aiming to fill near the day’s volume-weighted average price while avoiding two problems: moving the market with one huge order, and telegraphing intent to everyone watching. That’s why institutions like pension funds rely on it – when you need to trade size, doing it quietly matters.<\/p>\n Delta trading, often called a delta-neutral strategy, offsets positive and negative deltas – typically between an asset and a derivative tied to it – so the combined position is insulated from small moves in the underlying price. As prices move, those deltas drift, and the algorithm rebalances automatically to hold the target ratio. It’s constant-adjustment, bookkeeping-heavy work that would be tedious for a human, which is what makes it such a natural fit for automation.<\/p>\n The core toolkit is shorter than you might expect:<\/p>\n Notice that data shows up twice. It’s the input to every step: research, backtesting, live evaluation. Your system’s decisions can only ever be as good as the data feeding them, which is why a dependable stock market API<\/a> sits at the center of most serious setups.<\/p>\n Here’s where I stop being neutral, because this part is personal.<\/p>\n We started Tiingo in 2014 with a simple belief: high-end financial data should be accessible to everyone, not just firms with institutional budgets. Our motto is “Actively Do Good,” and making clean data affordable is how we live it. The quants at big firms have always had immaculate data to build on. We think you deserve the same.<\/p>\n Tiingo is a data provider, not a brokerage or an execution venue. We don’t touch orders. What we do is data – and the data we have is incredible:<\/p>\n The pricing was designed around a question we’ve asked since day one: how much can we give and get away with it? The Starter plan is $0 – that’s 500 symbols a month, 50 requests an hour, and 1,000 requests a day, enough to research and backtest before paying a cent. Power is $30\/month and Commercial is $50\/month. Flat rates, no surprises.<\/p>\n How do we keep prices there? We’ve taken no outside investment – not a dollar of venture funding – and we’ve been profitable for 8+ years. Sustainable disruption, not growth-at-all-costs. It’s exactly why we can put people over short-term profit.<\/p>\nHigh-Frequency Trading Is a Subset, Not the Whole Story<\/h3>\n
Pros and Cons of Algorithmic Trading<\/h2>\n
Pros of Algorithmic Trading<\/h3>\n
\n
Cons of Algorithmic Trading<\/h3>\n
\n
5 Algorithmic Trading Strategies<\/h2>\n
1. Algorithmic Arbitrage<\/h3>\n
2. Mean Reversion<\/h3>\n
3. Auto Algorithmic Trades Before or After Index Funds Rebalance<\/h3>\n
4. Volume-Weighted Average Price<\/h3>\n
5. Delta Trading<\/h3>\n
What Technology or Resources Are Needed for Algorithmic Trading?<\/h2>\n
\n
Use Clean Data With Your Algorithmic Trading Strategy<\/h2>\n
\n