Program traders, deluged by data, are scrambling for a solution. The surge in quotes of the past few years is being compared to a tsunami – a tidal wave – by the Financial Information Forum, the New York-based market data think tank. Peak message traffic in the equities and options markets hit 31,000 messages per second earlier this year, up from 900 in mid-1997, according to the FIF. And it says that number could double next year.
Cheaper computing power and the willingness to use it has both brokers and exchanges spewing out more data than ever, says the FIF. Decimalization, which overnight created 100 price points for every dollar of a stock's price, is also responsible for the surge, others note.
Traders now use computers for both order placement and up-to-the-minute analysis of market conditions. Smart routing determines where to trade. Sophisticated analytics determine what, when and how to trade.
But, with the data deluge, getting a firm grip on the what, when and how is becoming more difficult. Unsurprisingly, traders have no tolerance for delays in receiving data. Yet that's what many experience as their databases are unable to handle the onslaught.
Now, however, several vendors have developed a solution: the tick engines. Also known as tick databases, these systems can store large amounts of data in real time and allow traders to test their theories without pause. The technology nestles between a trader's front-end trading system and his market data infrastructure. Many prop desks, program desks and hedge funds are finding they can't live without it.
That has vendors smiling. Several offer the technology including KX Systems, Vhayu Technologies, Neovest, TimesTen, Townsend Analytics, Leading Market Technologies and Radial Systems.
For Vhayu, which began life in 1998 as an Internet service for daytraders, the wall of data couldn't have come crashing down any sooner. The collapse of the semi-pro market nearly brought the company down. But a decision to focus on the pros and a $6.5 million cash infusion last year appear to have revitalized the vendor.
Los Gatos, Calif.-based Vhayu (pronounced "Vie-You") says it has six customers, including the Japanese giant Nomura Securities. Another seven or eight are said to be in pilot. Employees number 24, up from six at the time of the financing. Vhayu expects to end the year with 35 to 40.
Gary Manton is head of sales and marketing at Vhayu. The exec sat down with Traders Magazine technology editor Peter Chapman to discuss the problems facing traders as well as the mechanics of the Vhayu technology.
Traders: We're seeing a lot of data these days.
Manton: No question. Data volumes are double what they were one year ago. Raw data for Level One equities is about four gigabytes a day. That's up from 2.5 gig one year ago. Take that up to Level Two and you're are at 20 gig. It's massive volumes.
Traders: What does this mean for traders?
Manton: Reducing latency is the biggest issue. Storing large amounts of real-time and historic data is another. And then there is the issue of handling reference data, corporate actions, corrections, and cancellations. These are all issues traders haven't had to address in the past.
Traders: Decimalization is a major driver?
Manton: And program trading. People are making smaller trades. They're trading fewer shares, but more trades.
Traders: Traditional databases can't cope?
Manton: Databases such as Sybase can't handle the volume of incoming data as well as all the queries simultaneously. Databases choke. I was at a Tier One firm recently… We did the VWAP calc on the S&P 500. We did all 500 stocks for the last sixty minutes. You can imagine the volume of ticks. Vhayu calculated VWAP in 0.3 seconds. They tried to do it for five stocks and locked their database. It crashed. That is our strength.
Traders: Who is using tick engines?
Manton: Hedge funds all the way up to the biggest Tier Ones. Prop desks. Program desks. Market data groups.
Traders: It is understandstandable why prop desks and hedge funds would find it necessary, but why program desks dealing with customer orders?
Manton: They're going more to auto-trading for their clients as opposed to manual intervention. Because every time you put manual intervention in place, you are missing opportunities. Anytime an order comes to an individual trader they may not be able to execute at the price that order was initially placed. You have lags.
Traders: O.K.
Manton: But if you start automating this, your customers can now say: 'whenever this situation occurs, whenever the 60-minute VWAP in Microsoft is eight cents off the current price, I want you to buy 5,000 shares.' That is the customer strategy. I've talked to a number of program desks. Their customers are saying: 'here's the strategy we want to trade.'
Traders: So, the customer directs the broker?
Manton: The customer can find that opportunity, send it to the trader, and then the trader must execute. That eight-cent variance may no longer exist. If that is now an automated signal and it comes off a system which sends a FIX message to the execution venue or OMS, you've eliminated a manual step. We're seeing that to be more and more the case.
Traders: Nomura is using Vhayu for its own account or for customers?
Manton: Could be both. Right now this is for prop trading at Nomura. A lot is their own stuff, but they are looking to extend the usage to a wide variety of desks internally.
Traders: Nomura is using the Portware portfolio trading system on its prop desk. Portware does not store data?
Manton: Portware doesn't do anything from a storage perspective. But our trading strategies are running their proprietary analytics. As soon as it comes up with an opportunity, it sends it over to Portware. Portware takes that and takes care of the executions, direct connections, etc.
Traders: And they can create a query in Portware?
Manton: Right, such as: 'show me Microsoft from the opening bell to the current time.' We have everything stored and can send that information back to Portware for display.
Traders: Do you also provide any canned strategies?
Manton: We have some canned analytics, but most traders, most hedge funds, won't be trading VWAP out of the box. There will be some modification to a canned strategy, otherwise everyone is trading the same strategy and no one has any advantage. Take VWAP for example. Block trades could skew the VWAP numbers, so let's normalize it down to the 75th percentile. Take all those block trades and normalize them down, then calculate VWAP. You are customizing VWAP to your approach.
Traders: What is MATLAB?
Manton: We've integrated with statistical packages like MATLAB and S-Plus. We are having conversations with SAS. These are statistical software applications that many quants and traders use to build trading models. A lot of the quants and PhDs have developed trading strategies using these pure stat libraries. They run them against our engine for back-testing. In other words, I have this idea for a trading model. Let's look at past five years.
Traders: Can you give me an example?
Manton: In MATLAB, you say: 'Should the VWAP be eight cents off the current price and today's volume exceeds the 30-day moving average by five percent, we want to buy 10,000 shares.' That MATLAB query will run in real-time on our engine. Should an opportunity arise based on that query, Vhayu will send a FIX message to a Portware.
Traders: Then what will the front-end do?
Manton: It will say – 'we don't want to buy 10,000 shares of Microsoft at one time. We want to do these in five blocks of 2,000.' This is to avoid impact. But you use MATLAB to create the query.
Traders: Vhayu does not decide where to send orders?
Manton: We could do that, but we don't have the direct connections that the Portwares have. We don't have that direct access. They are also tracking the orders: position management, what has gotten executed, what has not.
Traders: Bypassing the market data vendors and taking in data directly from exchanges has become popular today.
Manton: That's right. That's tied into the reduction of latency. People now want to start going directly to the exchanges for data. They want to eliminate the consolidators which have a minimal amount of latency. They want to eliminate one hop.
Traders: Does Vhayu take in data?
Manton: We do. Real-time. TAQ. Fundamental. We build feed handlers for all these different data sources.
Traders: You build feed handlers like Wombat?
Manton: Well, that is all Wombat does. We'll partner with Wombat. If someone has Wombat, we'll still do it. Our focus is on the data storage. But, if someone doesn't want to buy Wombat and wants a direct feed, well we've built feed handlers for Reuters, Bridge, Triarch, Tib, Nasdaq, NYSE, GL Trade. We have experience on staff. But we don't necessarily want to compete with Wombat.
Traders: Thanks, Gary.
