High Frequency Trading Explained, and Why You Cannot Do It

High frequency trading is automated trading that competes on speed. Positions last from milliseconds to minutes, individual profits are tiny, and the volume is enormous.

It comes up constantly in retail discussion, usually as an explanation for why trading is hard. Some of that is accurate and some of it is not, and the distinction is useful because it tells you where to spend your effort.

What HFT is actually doing

Several distinct activities share the label.

Market making. Quoting both a buy and a sell price continuously and earning the difference. Enormously repeated, tiny margin per round trip. Most HFT volume is some version of this.

Statistical arbitrage. Two instruments that normally move together diverge briefly, and the trade bets on convergence. The window is short, so speed decides who captures it.

Latency arbitrage. The same instrument prices differently on two venues for a few microseconds. Whoever sees and acts first takes it.

None of these involve prediction in the sense a discretionary trader means. Nobody at an HFT firm is deciding whether a chart looks bullish. They are exploiting structural facts about how prices propagate between venues.

What it costs to compete

The entry requirements explain why this is closed.

Co-location. Servers physically installed in the same data centre as the exchange's matching engine, because the distance light travels matters at these timescales. Exchanges sell this space.

Direct feeds. Raw exchange data rather than the consolidated feeds retail platforms use, which arrive later and pre-aggregated.

Engineering. Code where microseconds matter, sometimes implemented in hardware rather than software, maintained by teams of specialists.

Capital and scale. Margins per trade are minuscule, so profitability requires volume that requires capital.

That is not a budget a retail trader closes with effort. It is a different kind of business.

Which retail complaints are true

Worth separating, because the accurate version is more useful than the general grievance.

True: some edges disappeared. Anything that depended on reacting to public information faster than others is gone. If news breaks, a machine has already traded it. Strategies built on being early to obvious information do not work.

True: order flow is visible to those who pay for it. Firms that buy direct feeds and, in some structures, see retail order flow, know things about short-term supply and demand that you do not.

Overstated: HFT causes your losses. Most retail losses come from position sizing, chasing entries, and abandoning plans during drawdowns. Those are unrelated to market microstructure, and attributing them to HFT prevents the correction that would help.

Overstated: markets are less fair than before. Spreads on liquid instruments are much narrower than in the era of human market makers, and that saving is real on every trade you place. The pre-electronic market was not friendlier to small participants, it was just less visibly automated.

Mixed: flash events. Automated liquidity can withdraw suddenly and amplify sharp moves. That is a genuine structural concern and it affects everyone, including the firms involved.

The practical conclusion

Do not build anything that needs speed. That single rule saves an enormous amount of wasted effort.

Concretely, it rules out strategies whose entry depends on being first to a price, on reacting to a data release, or on capturing a difference that closes in seconds. Whatever backtests you produce for those, the fills will not exist.

What remains is substantial, and it is where the structural situation favours you:

Longer holding periods. A move that develops over hours or days cannot be arbitraged away by latency. The firms that dominate microseconds have no advantage over a multi-day position.

Capacity-limited situations. A fund deploying large capital cannot participate in something that only supports a small position. Nobody is competing away an opportunity too small to matter to them.

No mandate. This is the advantage worth internalising. You can be flat for three weeks. A firm with investors expecting activity cannot, and being forced to trade in poor conditions is a cost you do not pay.

Where the effort actually pays

If speed is unavailable and information is unavailable, what is left is judgement about situations that take time to resolve — and the discipline to execute a plan when it is uncomfortable.

Both are learnable, and neither can be bought with infrastructure. That is not a consolation prize. It is the only part of this market where a small participant has a real chance, and it happens to be the part most people neglect while worrying about the part they cannot access.

The practical version of that is unglamorous. Write your rules down. Test them on data you have not optimised against. Then practise the decisions with the outcome hidden, so you find out where your reasoning fails rather than where your latency does.

Review checklist

Is your strategy competing where you can win?

• Its edge does not depend on acting faster than anyone else
• It does not require data you cannot obtain
• It holds positions long enough that microseconds are irrelevant
• It works at a size a fund would not bother with
• It permits you to do nothing when conditions are poor
• Its failures are traceable to your reasoning rather than to your fills