[prodigal_erik]

How can HFT serve any purpose other than front-running a trade that was about to happen anyway? Is there any other reason latency matters? The correct allocation of society's resources did not actually change within those few milliseconds, so their profits can only come from exploiting some flaw in the way the market clears.

[Imprecate]

Front-running is knowing an order is going to be inserted into an exchange's book, and sending your order before it; an example would be if a broker saw a client making a big trade and sent his order before the client's, hoping to profit when the market moved.

In most modern stock markets there is not tiered access. My order is just as good as anyone else's, and no participants are given unfair notice of another's orders. Firms can spend money and be more competitive in these markets (e.g. better analysts, faster systems, co-location), but anyone has a fair shot at competing. If an order is public, there's nothing wrong with me taking action based on it before you if my only advantage is superior technology.

[rdtsc]

> In most modern stock markets there is not tiered access.

It seems if you are the broker and also trade for your own company, you would see your clients' orders before they go in. You can send in your order before theirs, thus creating a tiered access.

[Imprecate]

That's called dual trading and there are already regulations in place that prohibit the scenario you described. A broker must execute his clients' orders before his own.

In any case, most high-frequency operations are only trading for themselves.

[BrentRitterbeck]

Is there any other reason latency matters?

Let's say I have a statistical model that is able to identify and exploit statistically significant price deviations. Furthermore, let's say someone has a very similar model. If I'm able to execute my trade before my competitor, I get the profit, and he doesn't. That's one argument. There are many others.