Friday, July 24, 2009

Thoughts on High Frequency Trading and Stock Market Manipulation

This article from the New York Times does a nice job of giving an example of how high-speed trading algorithms can front run markets. It helps to explain why traders who only buy or sell after strength or weakness has been manifested are often the ones buying the high tick or selling the low one.

It seems to me that some of the traders who are most vulnerable to these machinations are very active traders (including prop houses) who frequently bid and offer for stocks. The algorithms are reading the order book ahead of others, which tips the hand of these traders.

Because the high-speed algos are buying and selling quickly as a rule, their effects on the markets longer-term are unclear. A stock may still travel from point A to point B, but the computers will affect the path from A to B. This may help explain why traders I work with who are more selective in their intraday trades and who tend to hold for longer intraday swings on average have been doing better than very active daytraders.

When up to half of all stock market volume consists of these algorithmic trades, one has to wonder about the edge of very active traders. Interestingly, those that are successful may be trading new patterns that have emerged since the onslaught of the high-frequency computers. My hunch is that these new patterns would involve a keen reading of order flow, catching the shift in the bidding/offering and the location (bid/offer) of transactions in real time.


Charles said...

A lot of these algorithmic servers are also located on the exchange itself, furthering their speed advantage. As "human" traders, we just gotta adapt.


MicroView said...

Your t-cost (transaction cost) per trade is increased by a fixed amount (due to the scavenge of algo front running). So yes, the price path might stay the same from point A to B, but the more frequently one trades during that window, the more t-cost one pays to grab alpha in his time frame. It's a tax on per trade basis (rather than profit)

heywally said...

"This may help explain why traders I work with who are more selective in their intraday trades and who tend to hold for longer intraday swings on average have been doing better than very active daytraders. "

And that is the way to take advantage of stuff like this - extreme patience, waiting until buying or selling has been exhausted at more significant levels and then 'fading' for a move the other way. Lately, with the trend up, I only do this on the long side.

Markus said...

Dr. Steenbarger,

the order book has been full of games for quite a while and there has been a time when real human beeing could play this game very successfully like Paul Rotter (have a look at this interview with him, but I doubt that any trader can be as fast as an algo operating in a co-located computer. Nevertheless I agree that watching the tape can be very useful for getting a good execution.


JimRI said...

Since markets are a manifestation of human behavior and psychology, the increased use of computers to make the trading decision has the potential to fundamentally change the market itself. Without the human component, does the market itself make sense?

heywally said...

"Without the human component, does the market itself make sense?"

We still have the human component as we are doing the programming and parameter/level setting for the buy and sell programs; it's just that everything is accelerated and potentially more volatile. That's where extreme selectivity as a retail trader becomes very valuable.

Matt Fahmie said...

Dr. Bret,
I have noticed very similar patterns in my performance statistics. I am tending to do better on my longer intra-day strategy than on a shorter micro intra-day basis. I believe this is because of the rapid shifts of supply and demand occurring by these high frequency programs. It is obviously very hard for a human to execute at the speed of a black box, so I have cut down on some of my shortest term strategies and have now begun to focus on longer intra-day holding positions where I attempt to position myself early. Literally all day we have seen shifts between 10000 shorts coming into this market and immediately being reversed and 10000 longs enter to erase every short that just enter shortly before.

I have created a video on the trade at the open, and if you take a look, notice at how quickly these orders are being put on and pulled off. Looking for accumulation and distribution within a short consolidation range has become increasing hard on a micro scale. But ultimately we as traders must be able to adapt to our environments. A new paradigm is currently being established, being stubborn and and unwilling to adapt will remove you from the game.

JimRI said...

Yes humans are writing the programs to outmaneuver humans. But could this reach the point where humans are just noise in the system and the programmers are writing programs to outmaneuver other programs - ie other programmers. This is still human, but the tools are so specialized and expensive that only a very select few could particiapte.

Cap said...

You are so right about HFT; coincidentally, I had just posted about this as well ... check out chart ...

John said...

this has been going on for years.

The guy over at did a piece on this a while back. He has been right for many years.

I think they should be jailed for is a crime and overlooked, quite deceitful really.

Caveat Bettor said...

High frequency traders do cost non-HFTs, but less than any other financial intermediary. What legitimate trader would complain about improved liquidity and spreads?

Here's my take: