Haim Bodek, Trade Tech 2013, REG NMS & HFT, here. Interesting presentation. You can sort of see that the code for these strategies is not looking very classically floating point intensive. Lots of poorly predicted branches in the code. The opportunities to vectorize will depend on the frequency of market updates from the Colos. As you trade in the less liquid names there is less data to work with and so less opportunities to vectorize. But Bodek’s point that the latency is of secondary importance to the timing of the execution order type is interesting. It sort of makes sense why folks are taking a hard look at Swap trading over the SEFs early on in the cycle of Regulatory implementation and rollout. They want leveraged influence over the development of the order types in the evolving market structure, similar to what Bodek describes at the start of REG NMS implementation.
The other thing, which I will … now that I explained that it’s all about shifting the distribution of trade events, when that toxic trade comes in that’s going to sweep against me, there is another concept embedded in this, to try to detect when that’s happening and to cancel. You know, you hear everything about HFT is canceling all the time, cancelation is extremely important. There is just an opportunity cost. If I get out, I can afford to cancel 20-30 times, to get out of the way that big landmine coming in, I have a lot of false positives, lot of false alerts, then I can get out of the way, and it only cost me market share. So there is an emphasis on avoiding negative alpha, and by avoiding, that’s kind of a market making mindset. If I avoid that toxic flow coming in, my distribution is going to bias towards positive mean in terms of my PNL. So this concept of sweep risk, where a large order, that’s … you, guys’ maybe, comes in and takes out the whole price point that is bane of this whole strategy. That strategy will lose money on that trade if it trades against it. What is going to happen, usually, is when a sweep is detected; you try to cancel, as it’s sweeping the market. Let’s say I am on one market, I get hit there, I get off the other market, that’s why you see a rapid withdrawal liquidity, and what I am going to try to do is either hit some dummy behind me, let’s say I traded, I am going to hit the dummy behind me, and think about an inverted market like, you know, like what’s it Ajay, right? I can hit a dummy behind me who is posting on Ajay, exit that bad trade, which I got rebate for, and I actually make a net profit on my bad trades if I manage sweep risk correctly, but what that does, since everybody does it in the HFT space, you know, the top five-six firms are all responding to the exact conditions, They are responding to each other, I’ll give you a very clear thing. If someone’s at the top of the queue and that’s fast guy, and he is always at the top of the queue, and he cancels, well they can see that over the price feed, and they are like “Whoa, I am going to get out of the way, something’s coming in” so there is this concept of looking at the electronic crowd and trying to forecast that a sweep risk is coming in. Run away, and that, if there are no other algorithmic trading traditions in size in that name, that will cause a cascade effect, and in some cases, you’ll get the mini flash crash when there is basically no algorithmic strategy in there that can correct prices that are chargeable from kind of say a statistical standpoint. So if there isn’t a stat to dive in there to trade that bad price, you are going to get a mini crashes, a crash is when 6-7 participants immediately flip out of the market.