do we need a crackdown on stock trading bots?
I know, I know, this blog just went from discussing porn and medical research to a post on high frequency trading and its implications from a business and computer science perspective. But that’s just how we roll around here, right? The jump off point for today? Felix Salmon of Reuters has an article in podcast form about the benefits and dangers of high frequency trading “algos” and wonders whether drilling under the melting Arctic ice to speed up the global internet so we can trade even faster is a good idea. I’m sure that the Arctic fiber channel is going to be put in place to improve the global communications infrastructure by creating a fat data pipe between the Far East and the West about 5,000 miles shorter than the one connecting them today, and that it will happen regardless of the impact on high frequency trading because it’s simply a good move for the world’s economy. But Salmon worries that the HFT algos will take advantage of the new route to make snap decisions and cause more stock market flash crashes.
Can an HFT algo get carried away? Absolutely, in fact, as shown in the previous link, we saw it happen in the real world. But consider the end result. After the crash, the algorithms rebuilt the market value of the funds they were trading within minutes to balance out the massive monetary void they carved out. On a human timescale, these are extreme changes happening all too fast and should be seen with suspicion and worry. For a computer, a millisecond is actually a decent stretch of time in which all sorts of calculations and decisions can be made, and those decisions are actually the same decisions that human traders would make because human traders gave it the basic formulas to judge the asset’s momentum on the exchange. And that may be the really unnerving part of all this. These are the same choices made every day by human traders who say they’re experts in the world of finance being done at hyper speed, and rather than being an unpredictable cyber-wilderness, the HFT algo landscape is just a time lapse of what we’d like to pretend is a carefully controlled and predictable process.
Since Salmon doesn’t seem to know how these programs are built — and they’re programs, not algorithms; these words have different meanings in computer science — he assumes that banks and other major users of this software are finding them too unpredictable to use for anything but small automated transactions triggered by individual investors. However, the real story might be that the banks are realizing the risks of releasing a buggy HFT program and noticing that rather than help them make money, the programs and their human traders are actually interfering with each other. As the program is deciding what to do with the stocks, humans change the value or momentum on rumors or drive the stock down on bad news. Both actions change the value and make it difficult to predict where the stock will end up. The software has an easily replicated and concrete model. Humans add random spoilers and obstacles, yanking the stock out of its handy mathematical fit for the HFT program to digest. If they do that enough, the software makes lots and lots of bad or unpredictable trades, or doesn’t make any trades at all.
Basically, it may come down to banks choosing to predominantly do one or the other. Humans can’t keep interfering with the software and the software can’t translate the interference into an alarming stream of bad trades that interfere with the humans. They will either have to trade on different exchanges or deal with different stocks. But in the end they’ll basically be gambling with billions of dollars using little more than the market’s momentum as their primary guide. The big formulas they use can be extremely complicated but as we constantly see, simply knowing how stocks and commodities should be quantified doesn’t mean that a bet on them will pay off. When the fundamental game is to make an educated guesstimate and hope it turns out right for a few seconds while everyone else is making their own guesstimates that can change the outcome of your bet, it really doesn’t matter who pulls the trigger or how fast. The results will still be hard to predict and what makes money today could easily lose money tomorrow. But that’s the nature of today’s stock market, plagued with wild mood swings, hoaxes, fear, rumor, and hype…