Commentary

Ivy Schmerken
Traders Magazine Online News

MiFID II Transparency Puts Stress on Data Architecture

Buy-side firms are facing huge changes in disclosure and transparency requirements, which could upend their data management architectures, according to this guest commentary from FlexTrade.

Traders Poll

Are you concerned about foreign ownership of a U.S. stock exchange?



Free Site Registration

November 1, 2012

By By Peter Chapman, Gregory Bresiger, Tom Steinert-Threlkeld

> Tracking Trades to Cut Bills

Bonaire Software Solutions is offering a fee calculation engine to broker-dealers and other capital market firms to help them better manage-and trim-their trading bills.

Large broker-dealers with multiple trading desks and multiple exchange connections may be overpaying exchanges on the trades they make, explained Bonaire chief executive Chris John.

The higher cost is due to the fact that contracts with exchanges are typically signed at the level of the individual desk, John said. That can preclude the broker-dealer from leveraging the overall volume of its trading, to win discounts from the exchanges.

The monthly bills also may not be split properly between desks. That's because end-of-the-month bills are allocated to individual desks based on simple percentage formulas, John explained, rather than by the amount of trades they actually made.

The firm's software, Revport, solves both problems with precise measurement, John said.

The Boston-based vendor has deployed the fee-tracking system at one large broker-dealer with multiple trading desks and multiple exchange connections. The software tracks every trade every desk at the unnamed broker-dealer executes with every exchange. Then, the software reconciles the expected fees from those trades against the exchanges' monthly bills.

Bonaire has been marketing Revport to mutual funds and other asset managers for the past nine years to help them manage fees they charge their clients as well as those they pay for distribution.

-Peter Chapman

>Citi Dark Pool Seeks Active Traders

Citi officials claim they're now ready to bring in the business of active traders.

The big broker has just introduced Citi Cross, an alternative trading system that is a dark pool designed to match buy and sell orders from traditional investors and financial institutions with those of highly active traders.

Citi wanted to offer a trading mechanism designed for demanding active traders-clients whose needs they haven't been able to satisfy-said Hannes Greim, head of Citi Cross ATS.

These active traders are those who may be trying to count profit before the day ends, rather than the average retail or institutional investor, with investment horizons measured in months or years.

Citi Cross contains an allocation matching algorithm that will change the usual dark pool selection process. Normally, the first orders at a price point are the first to be executed, Greim says.

But with Citi Cross, all orders that qualify at a price point receive equal treatment, Citi officials say. The system will find the point in the spread that will maximize the number of shares matched in a single print.

"Citi Cross," Citi officials said in a press release, "offers an alternative to the latency arms race by removing the notion of the queue position."

-Gregory Bresiger

> Thinking Algos

The algorithmic trading management unit of Cowen Group is developing algos that learn from trading.

This will lead to "algos that can really think for themselves and think on the fly,'' said chief executive Doug Rivelli.

In tests, Cowen used three approaches to predict upward or downward movement of a stock:

First, a form of analysis called logistic regression, where the dependent variable is the stock's price. Other variables range widely, from price movement to order book data.

Second, a form of machine-learning called a "support vector machine," which plots key factors as points in space, trying to establish a clear gap that is as wide as possible. Then, it predicts a result, based on where new data appears in that gap.

Third, "random forest" methodology, which uses decision trees to produce a series of predictions about what will happen next. The approach then averages the predictions.

The models, described at TradeTech West in September, were trained on six months of data and recalibrated every five minutes.

"Random forest" produced the best early results. That technique was correct 58 percent of the time, when it got a strong signal to act. And it produced a 1.75 basis point improvement in the price, according to Cowen.

-Tom Steinert-Threlkeld