Garbage In => Garbage Out

Everything in trading system design starts with the raw data: GI=>GO.

For my purposes at this time, I do not require live tick data nor do I require greater granularity than EOD data.

To date, the best source I have found is CSI. They claim very low error rates and the data is reasonably priced. It comes with a good software package, Unfair Advantage, which gives you control of roll logic and splicing method to create series for back-testing.

For $37/mo or $313/yr plus a $60 setup fee you get access to 400 futures contracts and all the ForEx pairs you could want, plus all the available history for each. You can download up to 49 contracts per day. You could, I guess, get on a 9-day cycle and have the entire data set available for testing. At this time, Covel’s Turtle Trader website has a 10% discount coupon for CSI.

There are numerous other data vendors, but none seem to offer the nice front-end that CSI does – you are on your own to do all the leg-work of constructing back-adjusted continuous contracts.

A sub-set of this CSI’s data is available for free at Curtis Faith’s TradingBlox website under “Resources – Data Providers – Free Historical Data”. It is updated daily. It includes 40 separate futures contracts with a nice mix of currency, commodity, interest rate and equity contracts plus 19 ForEx pairs. The futures data includes the unadjusted close (which means you can change the splicing method if you want) as well as back-adjusted OHLC, open interest and volume.

This is what I currently use. I don’t know if it matters, but I try to use this resource sparingly – I would hate for it to go away for bandwidth reasons. When the time is right I will subscribe to CSI.

Michael Covel’s Turtle Trader website makes available data from the Price-Data website. It includes 33 contracts. Each contract file has to be down-loaded separately and each download includes a separate file for each contract month. The advantage is that you can set-up whatever roll rules you want (the decision has been taken out of your hands on Faith’s site), and you also have all the overlapping contracts available in case you are interested in far-off contracts rather than just front-month. However, it is more work!

This post is part of a longer series including:

1. This Post
2. Creating Good Price Series
3. Data Validation
4. Generalized Extreme Value Distribution
5. Some additional protocols.