Back From the Drawing Board

OptiMark Says It Has a Bulge Bracket Solution

OptiMark is back.

The most spectacular technology flop on Wall Street is giving it another go.

This time, though, instead of taking on the New York Stock Exchange, it has set its sights on the program desks of the bulge bracket.

OptiMark has injected $30 million in cash, technology and intellectual property into a moribund purveyor of VWAP crosses formerly known as Ashton Technology.

In the process, it has transformed Ashton from a passive order-taking ATS into a risk-tolerant program trading house.

Now known as Vie Financial Group, the resuscitated shop is charged with elbowing its way into the business of guaranteed VWAP executions.

Traditionally dominated by such mammoth trading houses as Deutsche Bank, Morgan Stanley and others, the service of guaranteeing executions at the day's average price is aimed at buyside desks trading large lists.

To outbid the big boys, Vie employs a sophisticated proprietary algorithm that breaks down huge blocks into little chunks and feeds them piece-by-piece into the market. The algorithm's job is to trade the stock at or better than its volume weighted average price during a given time period.

Heading up the new enterprise is Dean Stamos, recruited from the alternative trading system NYFIX Millennium. "We are trying to create a hybrid," Stamos said. "We offer both anonymity and nearly zero market impact like an ATS, but with the fill rate of a traditional broker dealer."

Technically, Ashton was reborn through the efforts of OptiMark Innovations, the venture capital arm of OptiMark Holdings. OptiMark Holdings is the reconstituted OptiMark Technologies, the notorious brainchild of Bill Lupien.

The firm spent about $350 million over four years to build an ATS for the buyside.

The OptiMark system was intended to allow institutional desks to trade directly with each other, bypassing the Big Board. Apparently ahead of its time, it was considered too difficult to use and never got off the ground. Its operation was terminated in 2000.

OptiMark regrouped, however, as OptiMark Holdings, parent to two subsidiaries. One, OptiMark, Inc., offers consulting services to stock markets and others. Nasdaq's SuperMontage was built with the help and technology of OptiMark, Inc. The other, OptiMark Innovations, forms partnerships, makes acquisitions and invests in various businesses.

Ashton is OptiMark Innovation's first incubation. Over a period of seven years, Ashton itself spent a large sum – about $90 million – on an alternative trading system called eVWAP. Although easier to use than OptiMark, eVWAP never attracted sufficient order flow.

Vie Financial, unsurprisingly, is not an ATS. That could be interpreted as a sign that the Wall Street mania for creating black box ATSs has come to an end. Most have flopped. OptiMark is dead. eVWAP shut down last November. The Arizona Stock Exchange is defunct. Linx never got off the ground. Via-Net is still just talk. Liquidnet and NYFIX Millennium, which are praised by some pros, do not have heavy volume. Another respected player, POSIT, which is the granddaddy of them all, has seen its volume decline this year.

ATSs have two flaws. First, they generally execute no more than five or ten percent of their order flow. That frustrates traders. Frustrated traders stay away. That makes it difficult for the ATS to build critical mass.

Second, while the order sits in the ATS, the stock may move adversely.

Traders get burned.

Stamos, who obviously knows a thing or two about ATSs, notes there are two key variables that prevent crossing networks from reaching critical mass.

"Price and time," he said. "Both customers must agree upon a price at which to cross. Both buyer and seller must be there at the same time to put up the print."

Vie solves this, according to Stamos, as any large broker would: It guarantees the fill.

Here's a possible scenario: At 9:30, a customer shoots Vie an order to buy 100,000 shares of XYZ at the VWAP. There are no sellers of XYZ at 9:30. The algorithm breaks up the order and starts feeding it into the market. Then, at 1:00, with 50,000 shares remaining to execute in the buy order, Vie receives a sell order for 50,000 shares of XYZ at the VWAP. The algorithm is turned off. The two orders cross at the VWAP in Vie's matching engine.

"Price is eliminated because everyone wants the same price," said Stamos, "the VWAP. Time is eliminated because the absence of another natural side is mitigated by the algorithm."

So far, Vie is not crossing much of its order flow. Only about five percent of the orders match. The algorithm handles 70 percent. A group of "liquidity providers" trades the balance. To date, Vie itself is the largest liquidity provider. The firm, though, has made arrangements with an anonymous group of small market makers to step in if necessary.

Needless to say, the algorithm is at the heart of Vie's business. OptiMark contributed the algorithm – together with some of its vaunted matching technology – but it was not a part of the original OptiMark trading system.

OptiMark Innovations acquired the algorithm shortly before the Ashton transaction of February 2002. The seller was a noted, but undisclosed VWAP "scientist," says Vie's chief financial officer Jim Pak. He adds that the algorithm was used profitably for five years by a group of institutional investors.

That run of luck did not hold for Vie, though, during the first three months of its usage. The broker dealer lost about $114,000 in the final quarter of last year using the algorithm to provide liquidity to customers.

Despite the losses, it is this algorithm that Vie is betting will give it a leg up on the bulge bracket. It did not enter the guaranteed VWAP business to compete on the basis of commissions. Firms such as Morgan Stanley or Deutsche Bank may only charge a quarter-cent per share or even nothing to execute their baskets.

The big program desks make up for the meager commissions by their trading. They figure they can trade the list at a price better than VWAP. The customer pays or receives VWAP. The broker pockets the difference.

Vie argues it can deliver a cheaper execution to the buyside despite charging a higher commission. That's because the big program desk is likely to negatively impact the prices of the securities in the list through clumsy trading.

As large members of the NYSE, they will take the order down to the floor. Knowledge of the order will leak. The price and, therefore, the VWAP for that period will move adversely.

Vie argues its algorithm trades more nimbly than the bulge bracket desks. On average, according to Stamos, it executes only 20 percent of its listed orders on the floor of the New York. The rest is spread out over the five regional exchanges and Nasdaq's Intermarket. The anonymity and non-disclosure inherent in this methodology creates a lower market impact.

The bottom line is the customer gets a lower all-in cost on his trades. That is, the sum of the explicit (commission) and implicit (market impact) charges will be lower than those of a big Wall Street program desk.

"By scattering the order across the national market," Stamos explained, "we're denying any type of information leakage. We're making it practically impossible. No specialist will be able to gauge whether or not there is a buying or selling interest in the stock."

When Vie's orders do hit a specialist post at the Big Board via DOT they are in small amounts and spaced out in four or five minutes gaps, explains Stamos. That way the specialist is unable to determine whether or not they are part of a program. Specialists can generally tell when a program trade comes in through DOT because the intervals between orders are short, typically measured in seconds.

"It's not like a 150-share order coming in every ten seconds," Stamos said. "That pattern can be picked up very easily."

Vie is working with Elkins McSherry, a transaction cost measurement outfit, to produce monthly cost reports. Once they are available, customers will know the full cost of their trades. "We can't wait to get the numbers," said Stamos.

Vie's business model is something of an anomaly – a hybrid – in the world of VWAP trading. Most broker dealers offering VWAP executions roughly break down into three groups. Bulge bracket firms offer guarantees and assume the trading risk. Crossing systems, such as those of Instinet and the now defunct Ashton, attempt to match orders on a best efforts basis.

Finally, brokers such as Bernard L. Madoff Investment Securities and ITG utilize technology that feeds the trade into the market piecemeal. They attempt to "catch" VWAP. Again the trade is done on a best efforts basis. Fills are not guaranteed.

Buyside traders using the former get a price guarantee, but no control over that price. Those using the latter get control, but no guarantee.

In a twist, Vie plans to give its customers a guarantee, but will also cede them some control. Vie policy requires a buyside trader to leave his order with Vie for a minimum of two hours. But Vie will permit the trader to cancel the order if prices move adversely. It seeks to combine the peace of mind offered by the bulge bracket with the flexibility of electronic trading methodologies.

Vie's algorithm is similar to Madoff's TimeSlicing technology and ITG's SmartServer, but it has chosen to operate in a dealer, rather than agency, capacity. (Vie has recently started to offer a best efforts VWAP service.)

That, of course, entails more risk. As a dealer, it guarantees price and hopes to meet or beat that price through its trading. But if its algorithm executes the order at a price worse than VWAP, Vie eats the loss. Of course, if the algorithm executes the order at a price better than VWAP, Vie pockets the difference.

Such auto-trading is typically predicated on historical price and volume patterns. But those patterns may not be present on any given day. The risk is especially acute with less liquid stocks.

Vie's team of traders and technologists is constantly recalibrating the algorithm, according to Stamos. "We're not doing the same thing today that we were a month ago," he said. "The algorithm changes by the week. It is a constant effort to tune it and adapt it."

He added: "You're trying to predict volume throughout the course of the day. And obviously volumes aren't consistent. So, you constantly change volumes and predictions of volumes."

In addition to volume, the algorithm takes into account intra-day volatility as measured by the CBOE's VIX index and the absolute price level of the securities traded.

"Last year, in the August-September time period, the VIX was at a historical high," noted Pak. "So, what worked last year, won't necessarily give you the same result this year."

All the tweaking may be paying off. Vie does not disclose volume, but claims to be growing at 45 percent per month. Reported Nasdaq volume for Vie's two NASD-registered units, Vie Institutional Services and Vie Securities, spiked to a combined 10 million shares in January. That's up from negligible the previous month. Nasdaq shares account for half of Vie's volume.

In the final quarter of 2002, only two customers – one was Fiduciary Asset Management – gave it two-thirds of its revenues. In late February, no single customer accounted for more than 25 percent of total revenues. Vie is close to breaking even, Stamos noted, something it did not expect until May.

If true, that is a dramatic reversal. Ashton never made a profit in its seven years of existence. "We've made this transition from a shop viewed as providing VWAP transactions," Stamos said, "to a viable source of liquidity. A customer may not even care about VWAP. He just wants liquidity without market impact. We have a number of customers who could care less about VWAP."

Despite Vie's heavy reliance on its algorithm, its goal is to increase the percentage of volume crossed naturally. To speed the process it is actively pursuing sellside order flow. In the final quarter of 2002, the buyside accounted for 100 percent of Vie's revenues. Today the mix is 50-50, buyside-sellside, according to the firm.

"We are moving aggressively to aggregate as much sellside VWAP volume as possible," Stamos said. "Because everyone wants a natural."

The Calc at Ashton

Ashton Technology failed but not through lack of trying. The trading house even went to the trouble of patenting its own VWAP calculation. Based on every trade on every exchange, Ashton's VWAP calc was more accurate than most. Most VWAP calcs simply take into account all the trades on a security's primary exchange. In most cases that is the New York Stock Exchange.

Unfortunately for Ashton, a VWAP calc based on just the primary is what most buyside traders want. That's what they see all day on their Bloombergs. They don't want a fancy VWAP. Vie took the hint. It scrapped Ashton's patented calc despite the fact that it executes the majority of its trades on the regionals. It gives customers the primary.


Not to Move the Market

A typical customer at Vie trades lists. A typical order ranges from 700,000 shares to two million or three million shares. The average order size is 40 symbols of about 67,000 shares apiece. Vie does not get the multi-stock orders associated with indices such as the S&P 500.

Vie cautions its customers against sending orders in which the total number of shares in any given name exceeds 20 percent of the stock's average daily volume. Exceeding that threshold is likely to engender market impact.

"The customer is really making the VWAP at that point," Stamos said. "His order is impacting the price, what the benchmark will be at the end of the day."