253
respect to the system and its performance. Quality practitioners solve basic problems with
seven basic tools:● Fishbone diagram
● Pareto chart
● Check sheet
● SPC charts
● Flowchart
● Histogram
● Scatter diagram.Chance causes are those that are continuously active in the process and are built in.
Assignable causes are those that can be detected because they are not always active in the
process. If the variations in the product (caused by materials, machines, etc.) are due to
chance causes alone, the product will vary in a normal, predictable manner. The process
is said to be stable. We want to know how the product varies under normal, stable condi-
tions. If any unusual change occurs, we can see this change in our normal distribution
curve. We can say that it is the result of an assignable cause and not due to chance causes
alone. When assignable causes are present, the curve has shifted.
If you make no effort to measure or monitor the variation normally expected, you could
find yourself in a lot of trouble. All processes that are not monitored go downhill. So, it is
necessary to measure the output of any process to know when an input pro cess has changed.
The source of variation in a process can be found in one or more of five inputs: price data,
valuation data, fundamental data, calculated data, and economic data. We recommend
using a fishbone diagram, which can be useful for searching out root causes of trouble in a
pro cess. The variation when measuring outputs will result from chance (or system) causes
and assignable (or special) causes.
By using statistical tools, the operator of a production line can discover that a sig-
nificant change has occurred in the production line and investigate the root cause of the
change. He may even stop production before parts go outside specifications.28.2. Process View
A trading/investment system is a machine and we apply SPC to each and every one of
the outputs of the machine. Now, most trading/investments machines at best explain
40–70% of the variation of the process, hence, we recommend using the early quality
control tools, those developed when manufacturing produced large amounts of variation.
The best man-made algorithms have an r 2 .7. Therefore, we believe that SPC, which
is meant to control stochastic processes, is the proper tool to measure, watch, and control
the trading machines that run on stochastic processes.
If you are running a quantitative hedge fund with a long–short or other multifactor
system, under-/overweighting is normally a direct result of either portfolio optimization,
which can be controlled through construction of appropriate constraint functions, or the
portfolio selection algorithm. Because both of these are machines, you would expect the
outputs of these machines to be stable and measurable using SPC. When the machine
drastically over-/underweights a sector, or your take out ratio changes or excess returns
drastically change, you should see that in control charts, which means the environment28.2. P R OCESS VIEW