Good backtest score?

[cjsfireman]

I have been building my automated strategy for a few weeks, but I do not know what is considered a high enough score to start trading real money with? I'm not if if the scores are considered good or great. The back test is on a single contract, 10 minute time frame for the past 2 months. I've attached a copy of the scores. Is there a way to do a lot longer back test ?

Thanks for any help or advise
CJsFireman

[zacharydw00]

Looks pretty good compaired to what the futures brokers are selling. Check out this link to see how well the "professional" automated trading systems did http://www.ampfutures.com/system_details.html. That said, personally I wouldn't trust 2 months of backtesting. Those could be the only 2 good months of the year. If you are using Zen-fire or TT data feed you can reload/download 1 year of historical data.

[darckeen]

Those results look really good but like zach said your using a relatively short time span to backtest the system. The ideal way to test a system imo is to use random segments from long price history. This method allows you to examine varying market conditions and results are not dependent on a perticular start date. I'm new to NT so I don't know if/how you can do that.

Also be aware that those futures are essentially the same thing, broad equity indexes that are highly coorelated. Running a system on them at the same time wouldn't provide much diversification at all.

[cjsfireman]

Can someone let me know the best way and proper method to backtest? When I do a historic reload for a year with Zen-Fire it does not ever pull more than a few months of data. Thanks for any help!

[NinjaTrader_Ben]

Hello,

You should get more than a few months of data. What instrument are you charting? Did you increase the Days Back field in the Format Data Series window?

The best way to backtest is explained here:

http://www.ninjatrader-support.com/HelpGuideV6/BasketTest.html

[mainstream]

If you haven't watched this video yet it's very interesting and is a good baseline for determining how to interpret results of a system.

h**p://**w.mirusfutures.com/webinars/TruthAboutDrawdowns/TruthAboutDrawdowns.html

cut and paste, remove and replace the **

[cjsfireman]

Thanks for your insight

Thanks to you all for your insight. I will increase my back testing time and I have a few minor changes I want to address that I believe will increase my winning percentage even a tad more.

[mgbloomfield]

I have been building my automated strategy for a few weeks, but I do not know what is considered a high enough score to start trading real money with? I'm not if if the scores are considered good or great. The back test is on a single contract, 10 minute time frame for the past 2 months. I've attached a copy of the scores.

To me, the backtest results that you showed looked great. But you really need a much longer backtest with at least 1,000 trades before reaching any conclusions.

Is there a way to do a lot longer back test ?

Yes, merging older contracts months into the front month.

Or, with OpenTick as my data feed, I was able to use "ES ##-##" to get a continuous contract feed.

[darckeen]

I agree that you should to test it on alot more trades. Also try to test it under varying market conditions. The last two months the market has been in a very strong down trend. Try it out in sideway and uptrending markets too.

The most widely accepted stat for measuring the performance of a system is the sharp ratio which basically normalizes the returns by standard deviation. So a system that makes only half the returns of a system that has a three times higher deviation is better since you can use leverage to achieve higher returns.

Dont forget to include commision and slippage beofre you go live although with your systems low trade frequency they shouldnt be much of a factor.

[r2kTrader]

I don't believe you can do that unless you did it manually. Meaning you have to do the test, save results, export to Excel, repeat (for each segment of time).

Also, you can not get true DrawDown results as NT only considers the DrawDown to be Close of Equity. Or more specifically, the amount of money you cumulative lost on your trades for a given period, excluding how much you really may have drawn down while the trade was up. So take your risk factor and throw that out too.

Those results look really good but like zach said your using a relatively short time span to backtest the system. The ideal way to test a system imo is to use random segments from long price history. This method allows you to examine varying market conditions and results are not dependent on a perticular start date. I'm new to NT so I don't know if/how you can do that.

Also be aware that those futures are essentially the same thing, broad equity indexes that are highly coorelated. Running a system on them at the same time wouldn't provide much diversification at all.

[phillionair]

Very good backtest score!

cjsfireman,
I have good backtest scores too but I am not live trading yet. Have you done any live trading yet? Did your strategies work in live trading? Thanks,

[NinjaTrader_Bertrand]

Hi phillionair, good job! Did you also test those with commissions and a tick slippage?

[Trevor.Richter]

Can someone let me know the best way and proper method to backtest? When I do a historic reload for a year with Zen-Fire it does not ever pull more than a few months of data. Thanks for any help!

You can purchase historical data here:

http://disktrading.is99.com/disktrading/

I've bought data from them several times and have historical data back to 1999 that I use for backtesting. They're cheaper than anyone else I've come across.

Trevor

[ctrlbrk]

phillionair,

In the 'trades' tab of the backtest report, do you have any trades on "Bar=1"? This indicates entry/exit on same bar. NT doesn't know the order of the OHLC, so for back test purpose you need to code around this because the results won't be realistic. I have several threads here about it.

[zacharydw00]

FYI - A personal experience

Since my last post here, I have purchased ES data going back to 1997. After back testing from 1997 to current I discovered my stratagy lost money in the long run. Previously I was back testing from July 2006 to current, and my strategy was making lots of money. However, after back testing it for 11 years I discovered the strategy really only started to make money from June 2006 to current. So, the 18 months of data I was back testing on gave a false positive. The moral of the story... it's well worth purchasing 10+ years of data. Two low cost sourses I have found are http://disktrading.is99.com/ (as mentioned below) or www.tickdatamarket.com

I have posted a multi time period backtest framework code for Strategy Analyzer that will help achive more accurate results when an entry and profit/exit happen on the same bar. http://www.ninjatrader-support2.com/vb/local_links.php

P.S. - If anyone would like to trade their strategy idea for ES tick data, PM me.