3. Statistical arbitrage and market making: profiting from tiny inefficiencies

Theory 6: A survey of real strategies

There are profit opportunities in the market with no exposure whatsoever to whether the price rises or falls. They are worth fractions of a penny and they last fractions of a second.

The strategies seen so far at this level — momentum, mean reversion, pairs trading — share a common trait: with varying degrees of certainty, they bet on a future direction of price or spread.

The last two families in this module share a different approach. Instead of predicting where price is going, they seek to capture tiny, very short-lived inefficiencies in how several prices relate to one another at the same instant.

Statistical arbitrage: pairs trading at industrial scale

Statistical arbitrage, in its purest form, generalizes the logic of pairs trading to far larger baskets of assets. Not two stocks, but dozens or hundreds, grouped according to statistical models of common factors: sector, size, momentum, value.

What it looks for are small temporary deviations in relative price between assets that, according to the model, should move coherently with one another.

Scale is key here. An individual pairs trade may carry a modest statistical edge and considerable variability. But trading hundreds of similar relationships simultaneously, each with a small but positive edge, changes the picture.

By the law of large numbers, the aggregate result of the whole portfolio turns out to be far more stable and predictable than that of any individual position. And that holds even though each individual bet has a relatively modest probability of success.

It is a strategy that depends, almost by definition, on technological and risk management infrastructure considerably more sophisticated than trading a single pair of stocks. That is why it tends to be dominated by institutional quantitative funds rather than by individuals.

Pure arbitrage: no directional risk at all

Pure arbitrage differs from the statistical kind in something important: statistical arbitrage remains a probabilistic bet, whereas the pure kind looks for situations where, in theory, there is no directional risk at all.

The textbook situation is this. The same asset, or mathematically equivalent assets, quoted simultaneously at slightly different prices in two different places. Two different exchanges, for example, or an asset and its equivalent derivative.

That lets you buy where it is cheaper and sell where it is dearer at the same time, capturing the difference with no net exposure to whether the market rises or falls.

There is the scene we opened with. Modern markets are highly efficient, with thousands of participants constantly watching for these opportunities. So these discrepancies are usually minuscule — fractions of a penny — and they last fractions of a second, before someone captures them and prices converge again.

That explains why pure arbitrage is dominated almost entirely by high-frequency firms, with the technological infrastructure to detect and execute these opportunities in microseconds. Competing on that ground without that infrastructure is, in practice, unviable for an individual investor.

Professional market making: capturing the spread at scale

Market making, which we met conceptually in the level devoted to market execution, is itself a sophisticated algorithmic strategy when run at professional scale.

Instead of predicting direction, the goal is to capture the spread systematically and repeatedly. Bid and ask prices are posted continuously and adjusted in real time according to three variables:

• The order flow observed.
• The prevailing volatility.
• The inventory accumulated at each moment.

A sophisticated algorithmic market maker manages thousands of these relationships simultaneously across dozens or hundreds of different assets. It constantly balances the inventory risk we saw in the execution level against the expected return of continuing to quote in each particular asset.

Where the edge comes from, and where you cannot compete

What statistical arbitrage, pure arbitrage and professional market making share is where their edge comes from. It does not come from 'getting right' where the market is going.

It comes from a combination of three things:

• Execution speed.
• Scale: trading many small opportunities at once.
• Technological sophistication.

For an individual investor without that infrastructure, that combination is practically impossible to replicate competitively.

Understanding this is not a reason to dismiss algorithmic trading as inaccessible in general. The momentum, mean reversion and pairs trading strategies of the previous modules remain perfectly viable at a smaller scale.

It is a reason to understand realistically on which ground an individual competes, and on which one they definitely do not.