Historical data for backtesting

6 posts / 0 new
Last post
gambit
Historical data for backtesting

What amount of historical price data is available to use for system development and backtesting?

Using the REST API I got no MINUTE data back for the ASX200 A$5 CFD beyond 3 months ago.

I also tried requesting a few MINUTE_30 prices for the Gold A$1 CFD about a year ago and it returned nothing.

gambit
I also emailed Client

I also emailed Client Services this question. Their response:

Some of our markets hold more historical data than others, in each timeframe. I'd recommend logging in to the web based IG platform and opening a chart for the market in question. You can scroll back in each timeframe to see how much data is being held. Everything that we have is shown in the chart, and this is the same data that you have access to from the API.

So I opened up a Spot Gold M30 chart and it only went back about 1 month.

Summary is unfortunately IG doesn't seem to provide enough intraday historical data to do system backtesting.

Chris
Historical data for backtesting

Hi gambit,

We are only able to store a limited amount of data, as our Client Services team have stated. So on some markets, especially on the lower time frames, data may not go further than a couple of days or a month.

Thanks,
Chris

gambit
With all due respect, that is

With all due respect, that is an absurd response. I can understand if IG has a business reason for not making historical data available. However ten years of M1 data is tiny compared to today's storage costs and capabilities. Even keeping that much across hundreds of markets is not a difficult proposition.

Without data you can't build and test automated trading systems. Your competitors can provide enough data, so naturally your customers are forced towards those competitors while you continue this policy.

neilcaithness
Re: Historical data and long back-tests

Your're right gambit, about 25GB for 10 years of 1 minute candle data for 100 markets, so not that much. It would be nice if IG supported client's long-time-frame modelling efforts, but I don't think that's what they do. I'm curious though, do you think 10 years of back-testing for a 1-minute time frame is relevant? I store 1-second data (strike-price only) for a hand-full of markets and I'm happy with about 2 weeks for back-testing. Do you think a much longer back-test really adds more confidence?

Cheers
Neil

gambit
Hi Neil

Hi Neil

It becomes important if the average trade length is a few days, then you need years of data to get a good sample of history and ideally covering different periods of market behavior.

Also the software I am currently using for backtesting (NinjaTrader 7) will only import tick, minute or daily data. So if you using M5, M15, H1, etc data to model with, you still need M1 data.

Regards
brad

Log in or register to post comments