Quality Money Management : Process Engineering and Best Practices for Systematic Trading and Investment

239

systems. These calculations are similar to sensors on a machine, used in real time to focus

the Six Sigma quality improvement process.

We will discuss the most basic concepts of risk only, due to the enormous body of

knowledge on market risk and risk calculations, discuss how to use SPC to show when

the system has changed, and then DoE to find the root cause of the changes.

26.1. LOOP 1: Specify Single Performance Controls

To effectively evaluate the performance of a trading/investment system relative to itself, that

is, to normalcy arrived at during backtesting, the team will have to specify time intervals—

by minute, hour, day, month, year—and time spans, consisting of series of time intervals,

to create a time series of performance metrics. (The product team will have defined these

metrics in Stage 1, captured them during Stage 2, and now must plan to monitor these

metrics periodically.)

The performance over a number of time intervals can be calculated arithmetically,

geometrically, or continuously. Key components in performance measurements are the

fees charged by brokers, exchanges, regulators, custodians, and auditors as well as taxes.

Performance should always be calculated net of these fees.^1

26.2. LOOP 2: Define Benchmarks and

Attribution Controls

Traders sometimes argue that they should be measured solely on absolute returns, that

the uniqueness of their strategy renders their performance incomparable. Of course, the

trader has an incentive to convince management of his true genius: if something good

happens, he must be right! While some strategies are in fact unique (especially macro

strategies), most are not.

Traders prefer to have few, or better still no, quantifiable methods of determining the

value they add. In the world of systematic trading and investment and automated execu-

tion, the trader adds no value in times of normal operation. The purpose of a trader is to

simply monitor and oversee the trading/investment machine. Only if the machine breaks

down, that is, the algorithm stops working, and SPC charts show nonconformance, may

the trader step in to “ land the plane safely ” by unwinding positions. This role is critical.

According to market lore, Leland, O ’ Brien, Rubenstein Associates (LOR) began selling portfolio insur-

ance in 1983, without any backtesting, only probationary trading in personal accounts. In 1987, LOR ’ s

clients and licensees executed the programmed sales, contributing to selling that swamped the market.

Their programs underestimated volatility and overestimated liquidity. Proper backtesting would have

uncovered algorithm failures under extreme market conditions. Several firms went bankrupt. At Wells

Fargo, however, a trader ditched the algorithm and traded on his own to manage positions. He may have

saved Wells Fargo itself. Wells Fargo lost money, but remained in business. A good trader adds value

through crash management when a system is out of control. SPC will signal an out-of-control state.

When it comes to performance and risk, the real question to be asked is the following:

which algorithm(s) is/are adding value? Most systems have effectively two algorithms.

One algorithm for position selection, and another algorithm for optimal weighting of

positions in a portfolio. The goal is not to overweight sectors, market caps, credit risk,

26.2. LOOP 2: DEFINE BENCHMARKS AND ATTRIBUTION