Frequent Batch Auctions

Implementation details: http://faculty.chicagobooth.edu/eric.budish/research/HFT-FrequentBatchA…

No one?

@"IlliniProgrammer" @"justin88" @"peyo212" @"trazer985"

I have a much weaker opinion on this than I do on the SEC's proposal for 5 cent tick sizes. A one second interval auction format might be able to provide broader liquidity, especially during a market event, at least in theory. It arguably democratizes the game without getting much in the way of execution. But the problem is that the Cayman Islands and other offshore markets that will invariably form will provide a leading signal if they open up an exchange. Everyone will be giving and getting their signals from the continuous markets and then going back to arb the NYSE and CBOE.

The one interesting thing is that the market is operating with small enough latencies that distance matters (speed of light). The CBOE/CBOT/CME are all at least 200 miles from Canada. The NYSE and I believe the Nasdaq are 200 miles from Canada, too. In the eastern direction, we can force any offshore stock exchange on a boat from sitting within 200 miles of the US coastline.

Given that the speed of light is 186,000 miles per second, the SEC could get away with having millisecond interval auctions. It wouldn't be fair to human beings, but it would level the playing field for anyone who was halfway decent at C++, and could in theory rent a server from an amazing ISP, within a few miles of the exchange and get DMA from a broker with amazing latency. If the SEC were trying to democratize the HFT game, that would be one step.

I think the conservative thing for the SEC to do, if they choose to intervene, is to impose a millisecond auction format. It does not risk killing or reducing the primacy of the exchanges for US trading, so if it turns out to be a horrible idea, we can roll it back without giving rise to an unregulated competitor.

I have no idea what the result of a 1-second or 1 millisecond auction exchange format will be. I do know that if the Feds say "You can't do this (under our jurisdiction)" and doing this makes enough money, other venues will be found, if physically possible. So run the 1 millisecond auction format; if that works, announce plans to increase it to a wide enough radius to exchanges in Canada or Europe (~25 milliseconds) with the understanding that the SEC can throw the switch back overnight if volumes start migrating away. After a few years at that level without problems, we can try 1 second auctions.

If the SEC signals that it understands basic physics and game theory, it can prevent a lot of regulatory arbitrage and protect the US exchanges (and therefore its control over the trading of US-issued securities.)

He actually came to my firm and gave a talk on his paper. My main issue with his argument is that it's too idealistic. Of course everyone would like to stop spending exorbitant amounts of money on the HFT arms race, and I think most people in the industry would agree that his idea definitely solves that problem.

However...

Every exchange/trading venue would have to adopt the exact same rules for this to work. Even then, each exchange would have a great incentive to deviate (prisoner's dilemma, since it costs less technologically for all exchanges to have batch auctions than it does in the current state of the world) because traders would want to trade on an exchange that didn't have batch auctions (for the speed advantage) if everyone else was batching. Also, if not all exchanges participated in batch auctions, there would be even more time arbitrage than there is now, which is something that Budish wants to reduce in his world.

Then you might argue, "Well what if there is regulation to make sure all exchanges follow this new standard?" Think about how hard that would be. Even if we got past all the lobbyists and the United States actually follows through with this, it still wouldn't be enough. If we had batch auctions, then the big market makers that profit from time arbitrage would just go to other places that didn't batch and the U.S. markets would suffer from a great shortage in liquidity. How feasible is it to have every government in the world to agree to this plan and not be willing to deviate (even though deviating would be good for their country) for the sake of the greater good?

Even though it may be a good idea, frequent batch auctions won't happen any time soon. This academic theory is just not practical.

@"peyo212" what if we had an auction every millisecond or every five milliseconds? What if, by the time an order got to London or Paris or even the Cayman Islands, there would have already been an auction in the US? (The speed of light is about 180-190 miles per millisecond)

The fastest players would limit their business to foreign exchanges, sure, but I'm not sure transactions in US-based securities or transactions looking for US liquidity would really follow.

With auctions every five milliseconds or even one millisecond, in theory any regular Joe with a python script, a server near the exchange (not necessarily colocated at it) and a broker offering exceptionally fast DMA can compete in the markets. And market takers with information originating in the US have no incentive to spend 40 milliseconds waiting for an order to hit Asia.

One second is clearly too much time. But a few milliseconds puts severe geographical limitations on peoples' ability to get faster or better execution outside of US jurisdiction.

I don't have strong opinions about this. If HFT were eliminated (somehow), it would hurt retail investors who currently benefit from tighter spreads.

HFT makes its money off of large traders with sloppy execution, i.e. pension funds and even banks and hedge funds. That said, execution is cheaper than ever, partially due to competition in the HFT space.