Backtesting futures

[marcoheimann]

Hi

I'm looking for the best way to backtest a strategy on futures. The strategy trades intraday (1min) around monthly HLOC. Data provider is KinetickRT.

I tried to backtest on 3-13 and merge backadjusted, which gives completely differnt results from ##-## contracts. Unfortunately I couold not find out how KinetickRT constructs ##-## charts.

My intuition is to rely on backadjusted single contracts, mainly because this avoids the gaps. However I don't know wheter the montly data called from within the strategy is calculated with the the same offset values then minute data. Since backadjusting shifts contracts to unrealistic prices I would need to have the same adjustments in montly and minute data.

Thanks,

Marco

[NinjaTrader_Lance]

Hello,

Go to tools -> Account Connections -> choose your kinetick connection and click change

Click next until you get to the Login/password page. Here you can choose to check or uncheck use back adjusted data. By default it does not back adjust.

Please let me know if this does not make your results match in the testing.

[marcoheimann]

Hi Lance,

Thank you for your response. Can you tell me what is the proper way of backtesting multiple contracts.

So far I used continuous contracts (##-##) to backtest, but I read here that continuous contracts are useless for backtesting.

I'm particularly worried because I use intraday and monthly data, which as I understand come from a differnent database. If daily (that generates monthly bars) data has differnt rollover dates my backtests become worthless.

Is there any "best practice" note or someone who has experience with backtesting on futures?

Marco

[NinjaTrader_Lance]

Hello,

When using Kinetick with daily bars (monthly) and minute bars you may see differences in the Closing price as the daily bars use the settlement and the minute bars will use the last traded price.

With the Continuous contract you're note able to individually set offset amounts when back adjusting but otherwise it should come down to preference. Many other data providers do not support the ##-## contract.

There is not a "best" method and it will come down to personal preference and you're data goals. If you need the values from the minute closes you could always generate monthly bars from minute data (i dont know off the top of my head how many minutes that would be). Just keep in mind this would take longer to build.

This post you linked to be more in terms of whether or not you use backadjusted or non-backadjusted data and not whether you use the ##-## or 03-13 etc

[marcoheimann]

Lance,

First of all, my personal preference is to be as close to market conditions as possible. But your suggestion to simulate monthly bars with minute data does not seem straightforward to me. simply because a month can have between 27 and 31 days. Can you think of any other solution to assure that both data have equal values?

I remember that when downloading data you can set "generate daily bars from minute data'' to true. But I'm not sure if this would overrwrite my existing daily data.

The post I referred to earlier provides details about "perpetual" or "continuous" contracts and back-adjusted contracts. Continuous are the ##-## contracts in NT and are calcualted by the data provider (Kinetick) through a weightened average of two contracts. Back-adjusted contracts can be obtained in NT when testing the front month (3-13) and setting tools>options>data to "merge backadjusted". It further shows that strategies using previous high/lows cannot be backtested in perpetual/continuous/##-## contracts.

Yours,

Marco

[NinjaTrader_Lance]

Lance,

First of all, my personal preference is to be as close to market conditions as possible. But your suggestion to simulate monthly bars with minute data does not seem straightforward to me. simply because a month can have between 27 and 31 days. Can you think of any other solution to assure that both data have equal values?

You could create a time filter that allowed you to store the values that fall within a month into a custom DataSeries however this might be more work than it's worth.

http://www.ninjatrader.com/support/helpGuides/nt7/index.html?dataseries_class.htm

because your working with multi series you would have to initialize the series differently

http://www.ninjatrader.com/support/forum/showthread.php?t=3572

More on using DataSeries: http://www.ninjatrader.com/support/f...ead.php?t=7299

I remember that when downloading data you can set "generate daily bars from minute data'' to true. But I'm not sure if this would overrwrite my existing daily data.

Go to tools -> Historical data manager -> Edit -> select all daily data for the desired instrument by expanding with the + box (or just delete all the data for it if you intend to re-download all of it) -> right click -> delete

then re-download as you mentioned and check mark generate daily bars from minute data.

http://www.ninjatrader.com/support/helpGuides/nt7/index.html?importing.htm

The post I referred to earlier provides details about "perpetual" or "continuous" contracts and back-adjusted contracts. Continuous are the ##-## contracts in NT and are calcualted by the data provider (Kinetick) through a weightened average of two contracts. Back-adjusted contracts can be obtained in NT when testing the front month (3-13) and setting tools>options>data to "merge backadjusted". It further shows that strategies using previous high/lows cannot be backtested in perpetual/continuous/##-## contracts.

Kinetick data feed will use the offset for its continuous contracts

The only way you're going to get non adjusted market prices is if you do not merge contracts and were to test one at a time. If you're testing a large chunk of time you're going to have to decide which method of adjustment is most suitable for you.

Please let me know if I can be of further assistance.