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

146 CHAPTER ◆ 1 4 Gather Historical Data

We recommend you buy fundamental data from a single vendor and single wholesale

product and that you do not mix between wholesale products. If you need to mix, our rec-

ommendation is to initially purchase a software product whose sole purpose is to consoli-

date data for backtesting, such as Factset, FAME, or MarketQA. After the system is up

and running and successful, we strongly recommend the building of your own proprietary

database and consolidation engine, allowing you the ability to create your own custom

calculations and data into your backtester.

14.4. STEP 1, LOOP 3: Document Data Maps

and Results Tables

By this loop the in-sample test is done and the team should be preparing for the gate

meeting. The goal is to complete a data definition document and a data map. Within the

context of a database management system, a data dictionary begins as a simple collec-

tion of column names and their definitions and data types, but should evolve to include

in-memory representation of data elements such as integer, string, and date formats as

well as data structures and a Rosetta Stone. The goal of a data dictionary is consistency

between data items across different databases and different tables.

Data mapping is the process of laying out data elements and conversions between dis-

parate data models, between data sources and destinations. Here is an example of what

we mean. A real-time API we have worked with gives data as a string with no decimals.

Of course, in order to use this data for calculations, it must be converted and conversion

rules and techniques must be documented.

Multiple databases, data vendors, and data formats can really create problems. Documenting formats

early will save countless hours of debugging later on.

A data flow map will show the flow of data into and out of the system, how it is

cleaned, and how it is being stored. For example, a data map will show what data flows

into and out of the calculation engine, the risk engine, and the accounting engine. These

documents will eventually be used by programmers for implementation in Stage 3. The

documents should be appended to include the cleaning algorithms.

Real time

API

Calculation

Price feed “120575” 1205.75Price feed engine

Convert string representation of mini-contract

price with no decimal into the correct format

as type double

FIGURE 14-2

14.5. Summary

A backtest is a test of the ability of the trading/investment system to meet the specifi-

cations set forth in the Money Document. Backtesting requires lots of data, which may