Among the criticisms of high frequency trading is the charge that colocating trading strategies in the same data centers as exchange matching engines gives HFT firms an unfair speed advantage.
While most long-only buy-side firms do not execute latency-sensitive strategies, those that develop quantitative strategies will increasingly find themselves trading against the big HFT shops, says Prashanth Nandavanam , an expert on quantitative and algorithmic trading with EMC Consulting's financial services practice, who predicts that quantitative asset managers will colocate their trading strategies.
Nandavanam recently spoke with Advanced Trading about the buy side's evolving need for low-latency connectivity.
What is the difference between colocation and hosting in a low-latency data center?
Nandavanam: Colocation is when you sit right next to the exchange's ticker plant. For example, Nasdaq has a data center in Cartaret, N.J., and for a certain price they will let you put your machines as close to the ticker plant as you can get without get-ting into Nasdaq's private networks.
That means the data that is pumped out by Nasdaq's ticker plant will get to you with close to zero latency. The low latency data centers provide low-latency access to data but don't sit right next to the exchanges on the same network.
What types of asset managers are motivated to colocate their execution strategies near exchange matching engines?
Nandavanam: From a traditional asset manager's standpoint, they obviously need access to the venues but not necessarily with ultra-low latency. So they are happy to just go with being hosted in a traditional data center that is not necessarily colocated.
The traders who are really colocated are firms whose strategies run algorithms that go in and out of the market with such high frequency that being colocated actually makes a difference for them. Those are the guys who are the most likely to colocate. But it's also a matter of the cost, as well.
Is cost a factor in the buy side's decision whether to colocate?
Nandavanam: Colocation is significantly more expensive than traditional hosting. You pay a significant amount of money because the space is [at a premium], and the equipment that you need is probably just a bit more expensive from a networking point of view - the switches and the routers and so forth.
If a traditional asset management firm wanted to move into high-frequency trading, is it difficult for a newcomer to break into the colo game?
Nandavanam: The GETCOs of the world cornered this market a long time ago. And in many cases, they are so far ahead of the market because of the hundreds of man-years of effort they've invested in building this out. And some of the smaller shops in the low-latency, high-frequency trading market - those guys run the risk of going up against these big players.
Unless the smaller shops' strategies are that unique, unless they are playing in a different segment in the market, they run the risk of trading against the big firms day in and day out, and the big firms will happily take their money. That primarily applies to the high-frequency trading arbitrage strategies.
What other kinds of trading strategies can benefit from colocation?
Nandavanam: There are plenty of strategies that require the ability to react quickly to market movements because there is plenty of liquidity in the market for a number of instruments. If you are trading the SPY [an exchange-traded fund, or ETF, based on the S&P 500] or another index, their average daily volumes are in the tens of millions of shares. There's enough of that to go around.
That's why there is a lot of need for colocation, for speed and for algorithms. So it's really a matter of what people see as the edge that they are going to get by colocating that they don't already have by being in a relatively low-latency data center.
How is the buy side's need for low-latency access evolving?
Nandavanam: The more traditional asset managers have a huge amount of tolerance built-in for time-related lags. This means the need to absolutely buy at the lowest price by shaving off a penny is not necessary. If you take an insurance firm - which will be as far away from the high frequency market as you imagine - they are extremely conservative when they invest. They have no interest in high frequency trading because that creates volatility in their books.
That is the extreme end of the investment marketplace. They are happy to farm out the order and pay the bid/ask spread to their counterparties. Could they benefit from running a quantitative strategy that uses less human intervention and could use computation models to get better prices? Absolutely. But they tend to be fairly conservative; they will stay away from the quantitative side of price discovery.
That being said, many of the traditional asset managers are considering a quantitative element to their portfolios. These guys use a lot of fundamental analysis to drive their trading strategies. For these firms, being able to get better price discovery or even to get slightly better margins through a quantitative strategy is attractive, but they have to be very conservative in how they approach it.
But will the quantitative asset managers adopt techniques, including colocation, used by the high-frequency players?
Nandavanam: Many more buy-side shops will invest in quantitative trading models for some amount of their portfolios. But that doesn't necessarily imply high-frequency or colocation. It will certainly require something that provides them low-latency access.
If you are going to be trading in the quantitative world, the frequency does increase a bit. There's no sense using a quant model if you are going to be doing two or three trades a day. And when you start to trade more frequently, you need to move closer to the market, and that is when you are dealing with the latency issues.
The people who end up needing colocation are those who believe that there is intrinsic value in being as close to the market as possible, because that is where they see the value of their strategy. They feel that being able to detect small movements in the markets - being able to detect microtrends in the market - is where they are able to make money.