zerohedge.com (thanks to J. Pitera's thread for the link)
TCR: So what are some other ways that HFT shops make money?
GARRETT: There are many different strategies. Some take advantage of rebates, which are financial incentives the exchanges offer for being a market-maker.
Here’s where I should clarify that not all HFTs are bad. I’m very sour on HFTs in general because I’ve seen the havoc they can wreak, but there are good ones. Market-makers increase liquidity and make the markets more efficient. That’s great. There are good HFTs.
Some HFTs try to read and process the news quicker than everyone else. There are algorithms designed to read newspaper headlines, search for key words and execute trades based on what they read, all in seconds or less. I wouldn’t say this is particularly nefarious, because the HFTs in this case are just doing what humans do – trading the news – but faster.
That said, it can create problems. Awhile back, there was an errant news release about Boeing going bankrupt, and the HFTs started selling because they saw the keywords “bankruptcy” and “Boeing.” The story turned out to be an error.
In that situation, most human traders would pause and think, “Wait a minute, I’ve never heard a thing about Boeing going bankrupt. What’s going on here?” But the computers don’t think. They just execute their instructions, and in this case, it caused a crash.
Then there are the manipulative algorithms, the ones that prey on other investors.
TCR: Can you give us an example?
GARRETT: Sure. Many HFTs will make near-simultaneous trades on different exchanges and profit because of the delay in one of the exchanges. An example will help me explain: let’s use the NASDAQ and EDGE exchanges, and say that ABC stock is trading at $1.00.
The HFT will send a bunch of quotes (offers) to NASDAQ and EDGE, trying to sell ABC stock at $1.01. Once the NASDAQ order is accepted, the HFT can simultaneously cancel the $1.01 sell order on the EDGE exchange and replace it with a buy order at the original price of $1.00. EDGE immediately accepts that $1.00 order, because its system has not caught up to the new price of $1.01, and the HFT’s net position becomes zero.
This is possible because of latency, which is jargon for delay in the system. The net result is, the HFT captures a $0.01 arbitrage.
By scalping this tiny amount from many trades, the profits add up quickly.
A second example: HFTs can model other traders’ behavior. When someone trades through Scottrade or Interactive Brokers, their order has a unique number attached to it – the same number every time a client places an order. This number is bundled with all relevant trade information (time, price, etc.) and sold as an encrypted “enhanced data feed.” An HFT can then use those past results to predict the trader’s behavior.
TCR: So HFTs try to predict what you’re going to do before you do it. Do the brokers admit to selling this information? Can traders opt out?
GARRETT: This data is standard and available to anyone who wants to buy it, so it’s not that HFTs are purchasing illegal information. But the data set is huge and is only of practical use to players with very fast and powerful computers – meaning HFTs. And yes, most brokers I have encountered will allow you to opt out of having your unique number attached to your information.
To be clear, I’m not saying HFTs track your individual account and literally jump in front of you right before you trade. But they do use this information on the aggregate to model traders’ behavior. So an HFT could have a very good idea of when traders on, say, E*TRADE’s book will enter into a certain transaction.
TCR: Many defenders of HFT claim that it is a net-positive force in the market because it provides much-needed liquidity and tightens the spread between bid and ask. Are those claims true?
GARRETT: As I said earlier, there are many different HFTs that do many different things. But in my experience, in the aggregate, both of those claims are false. High-frequency trading will reduce liquidity when we need it most, and will flood the system with nonsense at other times.
Case in point, computers regularly withdraw liquidity just before news releases. Oil is a great example. The other day, there was a status report scheduled at 10:30, and around 10:28-10:29, the buy orders on USO (United States Oil Fund, an ETF that aims to track oil) dried up. That doesn’t happen with human traders.
So anyone who wants to get out of USO before the news release is out of luck; they can either take a bad price or wait until liquidity comes back.
Contrast that with the end of most trading days, when HFTs are unwinding their positions; I actually turn my platforms off for the last 10 minutes of the day because the action is confusing and useless. Sure, there’s plenty of liquidity as the HFTs unwind, but the action is just nonsensical. There’s no new information being introduced, no price discovery. It’s just scalping.
The whole liquidity argument is just a justification. On net, HFTs hurt liquidity more than they help.
I also don’t buy the argument that HFTs keep the bid-and-ask spread tight. I’ve seen algorithms that quote as far away from the NBBO as they are legally allowed to.
TCR: Can you expand on that?
GARRETT: SEC rules say traders can quote up to 8% from what the National Best Bid and Offer is, and they’re allowed to “drift” another 1.5%. So legally, traders can trade 9.5% above or below the NBBO.
Well, there are algos that probe the market, starting by submitting an order close to the NBBO, then working out to the fringes. These orders only last for milliseconds – they are not intended to be hit, only to sniff out other traders’ orders. So the algo works its way out, trying to get a bite on a price further away from the NBBO, and thus more favorable to them.
That is not a recipe for a tight spread. Now, the spread might look tight on your screen, but when you actually go to fill an order, you won’t get it, because the order has already been withdrawn.
Think of it like a dying star. When a star dies, we still see its light here on Earth, because the light is still traveling to earth. When an HFT cancels an order, your comparatively slow computer still sees the order for awhile. Then you try to fill it and it’s “Sorry, that order no longer exists.”
TCR: So are the quotes on Google Finance or Yahoo Finance reliable?
GARRETT: They are reliable enough to use as a broad snapshot. But I would not trade on them.
TCR: What markets are least affected by HFT?
GARRETT: I don’t know the answer to that. I see HFT the most in equities, but that’s just because I trade equities. It’s also prevalent in futures and Forex.
Within equities, HFTs tend to focus heavily on ETFs. The manipulation is far less in most individual stocks. |
