What Haim's referring to are a set of order types used on many exchanges to allow traders to place orders on a level that can't be displayed for regulatory reasons. There's nothing nefarious about them.
Reg NMS prohibits exchanges from displaying quotes that would lock (bid==ask) or cross (bid>ask) another marketplace.
As an example, say BATS is 10.01 bid 10.02 offered and the 10.02 offer trades out completely. Some HFTs want to be the first to form the new 10.02 bid to earn the spread + liquidity rebate, so they send post-only bids at 10.02. BATS sees stale 10.02 offers on ARCA, so they re-price the trader's order to 10.01 bid, or reject it (behavior depends on the exchange). In response to this, HFT algos would repeatedly re-submit their orders until the exchange let them in. This led to enormous stress on their matching engines during price moves.
In response, exchanges created order types that would let the trader bid 10.02, but only display it once the away markets faded from their point of view. It's basically like having a callback API rather than a polling one. The end result is basically the same, except less load is placed on the exchange's matching engine.
There's nothing "rigged" about this behavior. Traders forming new price levels on public markets still incur substantial risk that the price will move against them when filled. There aren't any order types that let you get execution priority in front of a large queue like he describes.
These order types and their behavior were well-documented on public websites and anyone concerned with the microstructure of how their trades executed could easily use them. The game he's describing is of little concern to anyone other than HFT MMs and execution algos that depend on favorable queue position. The entire process of a level ticking away and a new price being formed in liquid tick-wide names plays out in microseconds these days.
I have read Patterson's book which scratches the surface in "Dark Pools", Haim's book "The problem of HFT" and his regular pieces on Zerohedge. I cannot help but feel that Bodek has an ulterior motive "for exposing the system"... I don't buy his narrative largely because all this so called adverse behaviour is widely described in literature especially in exchange/venue published documentation, support and educational material.
A prime example is on page 93 of his book [5] which is a scalping scenario based on inter-market sweep orders (he attributes to a significant loss of capital at Trading Machines) to this strategy... Come on Haim really? you never considered the adverse affects of placing vanilla limit orders? The books [1], [2] and [4] touch on these exact scenarios.
I don't buy for a second that he couldn't have picked up a phone and called his account manager/support technician at X venue, or his broker to ask why he was getting an abysmally low rate of fills, or seek colour on orderbook oddities. Despite his lofty background it appears he was out of his depth at "trading machines" and now is attempting to sell his ideas to anyone that will listen. I do commend him though on shedding light over the field and on getting me interested in HFT/execution.
For anyone interested in learning more, some books I would recommend:
[1] Dark Pools - Erik Banks (2010).
[2] Empirical Market Microstructure - Joel Hasbrock (2007).
Exactly. The vast majority of "exotic" order types fall into generally two categories:
1) Orders that help deal with the asynchronous nature of the market. Removing the restriction on locked markets can reduce the need for these orders, but with that restriction in place these orders are a net benefit IMO.
2) Orders that help a participant control how an order is filled. Post only, etc, to ensure one gets a passive fill vs taking liquidity.
There are no secrets here. These orders are all public. Whether you can use them or not is up to how you get to the market.
I'm not knowledgeble about HFT by any stretch of imagination, but I watched this epsiode of VPRO Backlight when it aired on dutch TV, and what I got from it (not just from Bodek himself, but also the Nanex guy) that initially, these 'special' order types had been a close-kept 'secret' only known by insiders. Maybe not their existence, but how they could be used to 1-up every other HFT algo that wasn't in on this trick.
The video isn't only interesting because of Haim Bodek's allegations by the way, there's quite a few fairly high-profile people (formerly) active in the HFT world commenting on how markets are effectively broken and unstable because of opaque HFT algorithms, stock exchanges actively promoting HFT to drive up the volume of trades (and hence their profits), and regulators not really doing anything to reduce the risk of flash crashes or penalize parties that completely break the market every once in a while.
These order types are highly regulated by the SEC and FINRA. None of this is a secret - any new order type has to be submitted to the regulators in a public process before it is authorized. Exchanges and trading venues are required by law and regulation to be transparent.
I don't think the actual order type was a secret, just that some HFT firms figured out a way to use it which gave them an advantage over other HFT firms. The firms who didn't understand it lost out.
The whole point is that these 'smarter strategies' are only smarter because they are exploiting arbitrary API conventions, not smarter because they improve the utility of markets to society in general. Which is a tragedy, because if all these geniuses were working in the fields they trained in - computer science, physics, or bioinformatics - they would be contributing to society on a more fundamental level.
There are several fallacies to this sort of argument:
1) The myth of the HFT geniuses. Like any other industry there are smart people in HFT, there are also dumb people & bad software developers. In some ways the industry is very insular and has many bad habits.
2) We could harness all these people to something that provides more utility. It would be great if that were true, but how do you do that? Historically, centrally planning what careers people are required to follow has led to pretty bad outcomes. Conversely, if you think you can provide more utility by hiring away some of these smart people, nothing is stopping you.
Your first point isn't a fallacy to the argument. If they are good enough to to a job there, they're good enough to do a job elsewhere.
Your second point is trivially answered: you let the market decide where they work.
That's not obvious to you because you conflate "provide utility" with "make money". There are lots of things that make money that provide no utility. A large part of the history of business regulation is basically restricting those so people focus on providing utility rather than making money at things that provide zero or negative societal value.
The argument you respond to is basically suggesting that HFT is not generating any value for society. If you want to demonstrate that it's false, you have to show that HFT creates value in line with its costs.
Fine, but why doesn't every new industry have to make this same argument?
It's pretty obvious to me that social networks have some pretty dangerous attributes (decrease in privacy & personal securty, etc) to them. Do these dangerous attributes outweigh the value that twitter provides? Who decides and why?
Fine, but why doesn't every new industry have to make this same argument?
Because most new industries aren't within the financial industry, messing with the fundamental operation of markets. Those that are have to make that argument.
I think every new industry does have to make this argument. And they have to make it to their industry regulators and to the people who employ them. That is we, the people.
Myth? The finance industry, where the HFT folks are near the top, pays epic salaries for the smartest people. They do this to hire the best and the brightest, and would not continue to do so if it did not work.
In the documentary, Haim relates a story where the team that he was on included PhDs in bioinformatics, physics, and mathematics. This does not look like a myth to me. The mathematical knowledge required to understand and be competitive in the HFT environment, coupled with the software development knowledge is very rare. This industry consumes the time of geniuses in order to function and pays them well for it - that this attracts dumb people and bad software developers that are occasionally hired hardly makes the existence of these geniuses a myth.
> 2) We could harness all these people to something that provides more utility. It would be great if that were true, but how do you do that?
Disclaimer) I work in HFT and this is all anecdotal.
I have found very little correlation between advanced degrees and ability to determine valuable trading strategies. In fact, the strategies that most people bemoan the most with HFT require the least amount of mathematics because they are very simple.
I've also worked outside of trading in both big institutions & start ups. The ratio of genius to average to bad seems about the same in all of these environments.
As far as the common pool resource approach, the problem with that is that it will most likely tilt the balance of power in the markets even further into the hands of a few giant institutions. Why it is popular to protect the poor investment bankers from the ravages of HFT baffles me.
I agree that there are improvements that could be made to the HFT world. Most notably, eliminating the subpenny rule, which acts as a minimum wage for HFT:
I don't think that someone as apparently smart as Haim is spent a year of his life troubleshooting an API call. There's got to be more to it than that.
Reg NMS prohibits exchanges from displaying quotes that would lock (bid==ask) or cross (bid>ask) another marketplace.
As an example, say BATS is 10.01 bid 10.02 offered and the 10.02 offer trades out completely. Some HFTs want to be the first to form the new 10.02 bid to earn the spread + liquidity rebate, so they send post-only bids at 10.02. BATS sees stale 10.02 offers on ARCA, so they re-price the trader's order to 10.01 bid, or reject it (behavior depends on the exchange). In response to this, HFT algos would repeatedly re-submit their orders until the exchange let them in. This led to enormous stress on their matching engines during price moves.
In response, exchanges created order types that would let the trader bid 10.02, but only display it once the away markets faded from their point of view. It's basically like having a callback API rather than a polling one. The end result is basically the same, except less load is placed on the exchange's matching engine.
There's nothing "rigged" about this behavior. Traders forming new price levels on public markets still incur substantial risk that the price will move against them when filled. There aren't any order types that let you get execution priority in front of a large queue like he describes.
These order types and their behavior were well-documented on public websites and anyone concerned with the microstructure of how their trades executed could easily use them. The game he's describing is of little concern to anyone other than HFT MMs and execution algos that depend on favorable queue position. The entire process of a level ticking away and a new price being formed in liquid tick-wide names plays out in microseconds these days.