It's not front running; it's getting out of the way
Front-running isn't really a thing that's actually happening. Here's what is actually happening.
The HFTer/liquidity provider wants to sit there all day long buying IBM for 99.99 and selling for 100.00 making his penny a share in return for providing liquidity. But the liquidity provider has a problem. Every now and then Brad Katsuyama (or someone like him) shows up with proprietary information that, say, IBM should really be trading for 101.00. Brad tries to suck up all those shares being sold for 100.00 as fast as possible before the price moves. That's great for Brad but all of a sudden the liquidity provider has lost a bunch of money when IBM moves to 101.00 and he can no longer buy for 99.99.
So the liquidity provider tries to recognize as accurately as he can, and as fast as he can, that Brad has shown up on the scene and get out of the way by canceling all his orders and resubmitting them with a new price. That's why when Brad tries to buy 100,000 shares (from multiple exchanges) in the book he sees a bunch of his orders not get filled because the liquidity provider is running away from him.
So it's not the HFTer jumping in the middle like people think. It's actually the HFTer running out of the way of a big fish.
So how is this good for you when you're buying a small number of shares? The more often the HFTer gets eaten by a big fish the wider his spreads have to be to make an overall profit. If he gets eaten too often he has to buy for 99.95 and sell for 100.00 (5 cent spread) instead of 99.99/100.00 (1 cent spread). Spreads used to be really big: 12.5 cents or 25 cents. They're smaller now because computers can react a lot faster than humans to get out of the way of Brad. They also don't demand really big salaries so they cost less to operate.
To sum up: HFT isn't jumping in the middle. It's jumping out of the way. And because they keep getting better at that they can afford to offer lower prices for liquidity when you show up and want to by 1000 shares of IBM to hold for the next 5 years.
There are other advantages for you as well having to do with faster price discovery by the market, but I'll leave that aside for the moment as a separate issue.
(reprint from https://gist.github.com/harryh/9996284)















