A detailed exposé on how the market is rigged from a data-centric approach

If you read "Flash Boys" by Michael Lewis, he describes the creation of the IEX, which attempts to do exactly this. RBC created a tool to try and synchronize order flow, which worked for a time. Part of the solution that IEX uses is to put large spools of fiber-optic line in between the servers to delay order flow long enough to negate the HFT systems.

Somewhat farcically perpetuating the metaphor, is it not the case that Bill gets to know that Joe is interested in buying sheep once Joe has bought a few of them? And at that point Bill can outrun Joe and make money from his (very near) future purchases?

That was my reading of the article - it seems that either there is a flow of information from the trading events to the fast traders, or the scenario portrayed in the article was very unlikely (though I guess that couldn't be ruled out given how much trading there is). noonespecial's metaphor seems to apply, what have I misunderstood?

A solution: Discrete double auctions. Instead of continuous trading, the exchange can divide up the day into a series of small windows (say 100ms). When you come to trade in the market, you have to wait for the next window to open. You submit your order and you find out what happened at the end of the window. This way nobody has any timing advantage and the delay is barely noticeable to 'normal' traders (waiting 1/5 of a second is hardly an inconvenience). It also stops all the order-book shenanigans that HFT players get up to (where they stuff the book with orders and cancel/resubmit them at high frequency).

So why don't exchanges do this? They make a ton of money in fees, it simply isn't in their interest to prevent HFT at the moment. Change their incentives (ie. regulate differently) and they might actually do something about it.

What is happening here is really quite simple, and doesn't deserve an entire blog post.

There are two exchanges, A and B, and a market maker Jill is quoting (say) 10,000 shares on each of those two exchanges for $17.

Big institutional trader Jack sees the 20,000 shares and decides that he wants to buy 15,000 of them, so he sends two orders for 7,500 shares each to A and B. Because of various effects (network latencies, routing switching delays, whatever) his order arrives at exchange A first, and is immediately filled at $17.

Jill, who has her computer co-located at exchange A, sees that she has sold 7,500 shares for $17, and realizes that there is demand for shares. Because of this demand, she decides to raise her prices. She immediately cancels her remaining 2500 shares on exchange A and replaces them with 10,000 shares at $17.05 and sends an instruction to do the same thing at exchange B.

Because Jill has fast computers and low-latency connections, her cancellation arrives at exchange B before Jack's buy order, so Jack is told that there are no longer shares available on exchange B at $17.

RESULT: Jack is filled for 7500 shares at $17 (half of what he requested) and the new market best offer is $17.05. Jack is welcome to submit another order for $17.05 if he wants to buy at that price. Jill is now short 7500 shares at $17, and will try to buy them back at a lower price (she may or may not succeed - until she does, she is exposed to the risk of further price rises).

Jill was able to use her speed advantage to detect that there was additional demand to buy this stock, and raise the price at which she was willing to sell it before Jack had finished buying all that he wanted to. This is exactly the way that an efficient market is supposed to work - it reacts to fluctuating demand (and other information) to set appropriate prices.

I think there are several things that get glossed over while people are working themselves up about this -

1. Jack is upset because he couldn't buy 15,000 shares at the price he wanted to buy them. But Jack has no god-given right to be able to buy shares at the price he likes best. He is subject to the laws of the market, just like everyone else.

2. The only reason that Jill has a speed advantage over Jack is because she has paid for it! She has paid to co-locate her server at the exchange, and she has paid to use high-speed connections between exchanges. Are we going to declare that paying for a competitive advantage is suddenly immoral?

3. If Jack doesn't like this state of affairs, he has several options. He can invest in high-speed infrastructure as well. He can use smarter order-routing logic (e.g. adding delays to his orders so that they arrive at the exchanges approximately simultaneously, or splitting his large order up into multiple smaller orders). Or he can use a broker who will do these things for him. If Jack doesn't want to pay for any of these things, then he has to put up with lower quality execution. As much as he might wish it, the ability to buy as many shares as he wants at the price he wants them is not a universal human right.

Very true! But the reverse is also true: Critics need to show that something immoral is going on. Nanex's analysis does not, to my eye, even try and make this case.