Backtesting on 1 tick data

F

farmerjohn1324

Realistically, backtesting on data that tests every tick is the only way to get 100% accurate results. And why would anyone want results that are less than 100% accurate?

Is this realistic to expect NinjaTrader to do this? I am just starting on NinjaTrader. Previously my only backtesting experience was on MetaTrader 4 for Forex. I don't recall if the backtesting was every tick or not.
 
I've tried twice. It says it's running the backtest, then eventually the program closes. I've never seen it close because I get busy and do something else. I can't watch it forever waiting for it to potentially do something.
 
How are you defining a tick? Trade price? Bid ask change. Top 5 bids/ask? Data set gets enormous very quickly.

There is a quote from a top Google programmer. "I look to solve difficult problems by trying to get 98% of the answer with 2% of the work."
 
farmerjohn1324 said:
I've tried twice. It says it's running the backtest, then eventually the program closes. I've never seen it close because I get busy and do something else. I can't watch it forever waiting for it to potentially do something.
More...

Hi, there - Please feel free to contact NinjaTrader support at support@ninjatrader.com. Our support team will be happy to help walk you through what is possible and troubleshoot any issues!
 
farmerjohn1324 said:
Realistically, backtesting on data that tests every tick is the only way to get 100% accurate results. And why would anyone want results that are less than 100% accurate?

Is this realistic to expect NinjaTrader to do this? I am just starting on NinjaTrader. Previously my only backtesting experience was on MetaTrader 4 for Forex. I don't recall if the backtesting was every tick or not.
More...
You won't get timeseries with tick-by-tick data which have irregular interarrival times. Measuring characteristics with irregular timeseries is much more sophisticated, there is much less literature on this matter.
 
I would suggest that you use tickdata as the last example for your backtests because it takes much more time than 1 minute data for example. If you start everything based on tick data for testing then this is very time consuming. Therefore you need to take into account also time management when backtesting.
 
farmerjohn1324 said:
How can you over fit? Don't you want it fitted as well as possible?
More...

One of the problems of backtesting is that we assume perfect execution and fills. The market does not behave like that, I doubt that you have the infrastructure in place to get filled perfectly at tick value every single time. The reality is that you will most likely be late to that value with a retail account.

You need some kind of an average to enter and exit the market and that is only achieved with an OHLC snapshot in time. Then if you choose the worst value of that candle for your algorithm you might be close to how that market would behave.

If you only use tick data you can't choose from a set values at the moment you enter/exit, so you are running a wonderful and perfect algorithm that will yield amazing profits. Then when it is run live the story changes quite a lot.

Tick data is expensive not because of its value but of its size. It is huge dataset that has to be stored somewhere. Because it has a price people gave it value, but it doesn't help you as much as you think. It is actually useless.
 
You must log in or register to reply here.
Back
Top