Index futures back testing?

[rb7]

There are three features which can cause a difference in backtest vs. realtime results.
1) Set statements: SetStopLoss, SetProfitTarget, etc. These statements operate INTRABAR...so in Backtest mode, Multicharts does not know which came first...the high or the low.
2) IOG set on (IntrabarOrderGeneration=True). This setting will cause ALL ORDERS to be fired anytime within the bar vs. fired at the end of the bar.
3) Renko bars - the current implementation is buggy as I have witnessed. To resolve this, I switched to the newer FlexRenko bars and all is well when backtesting.

The trick is to backtest with tick data.
But I guess, none of the available commercial solutions support that.

 

[mason macgregorson]

There are three features which can cause a difference in backtest vs. realtime results.
1) Set statements: SetStopLoss, SetProfitTarget, etc. These statements operate INTRABAR...so in Backtest mode, Multicharts does not know which came first...the high or the low.
2) IOG set on (IntrabarOrderGeneration=True). This setting will cause ALL ORDERS to be fired anytime within the bar vs. fired at the end of the bar.
3) Renko bars - the current implementation is buggy as I have witnessed. To resolve this, I switched to the newer FlexRenko bars and all is well when backtesting.

good to know -thanks

 

[Bad_Badness]

The trick is to backtest with tick data.
But I guess, none of the available commercial solutions support that.

Yeah the engines are not able to deal with per tick resolution. In fact they barely deal with then Open-Close-High-Low resolution consistently. To wit you can play forward bar by bar, go back one bar and it recalculates resulting in "finding" new OCHLs that were always there. Just something you have to work around.

 

[lmseldin]

What I do, is look at the associated ETFs. Not the same, but does not need to roll over. So for S&P 500 I use SPY. Years ago, when I used eSignal, they had symbols for futures that were combined into a single asset symbol. Have not used them in years, so not sure how current this is.

 

[lmseldin]

"Is there any software that allows for backtesting if index futures? And any good platforms for algo trading index futures?"

As far as back testing software, I recommend looking at WealthLab. Can be used for easy building of basic strategies with canned indicators and can do complex programming. It has been around for many, many years. It has full set of metrics for analysis, etc.

 

[farmerjohn1324]

"Is there any software that allows for backtesting if index futures? And any good platforms for algo trading index futures?"

As far as back testing software, I recommend looking at WealthLab. Can be used for easy building of basic strategies with canned indicators and can do complex programming. It has been around for many, many years. It has full set of metrics for analysis, etc.

Is WealthLab also a brokerage? If not, what trading platforms can it be used with?

 

[ph1l]

Is WealthLab also a brokerage? If not, what trading platforms can it be used with?

https://www.wealth-lab.com/Software/Features

Trading

• Automated trading with major brokers (Interactive Brokers, TD Ameritrade, Alpaca)
• Integrated with Kraken and Binance for trading Crypto currencies
• Automated Paper trading and manual order placement are fully supported
• Trade multiple strategies real-time with multiple brokers!
• Load open orders from your broker into Wealth-Lab

 

[virtusa]

The trick is to backtest with tick data.
But I guess, none of the available commercial solutions support that.

I backtest based on tickdata. But I do this manually in NT8. Takes a long time but gives most accurate results. I need 4-6 hours to test one week of data/trades. Years ago I tested over 10,000 trades to build a good system. And after 20 years that system still works.

An additional, but for me very important advantage, is that I saw in slow motion what happened. From that slow motion I learned alot about how the markets work and how my system reacted on it. At least for my trading system.

I don't like the "IT" approach where you write code, run it and receive a summary without even seeing how your system created that summary. Kind of a magical blackbox. Most people then try to optimize the parameters. And a few days later have to optimize the parameters again, and again...

A well build strategy does not need a second optimization as that strategy would adapt automatically to changing market conditions.