Automated Backtesting

Hello fellow FF members. I have a solid EA and am looking to backtest its viability over several years of market activity. How can I go about doing this? Metatrader, as we all know, does not create accurate backtests. I have heard TradeStation is an option, however I am not a U.S. citizen. Any suggestions?

Thanks

There

Hey everyone!

Say I've downloaded M1 history data from Alpari more than a year ago and I download M1 history right now is there a way to merge these two or an easier way to go about this?

P.S. Thanks Raygun for the 99% modeling document Great stuff!!

Heny

Thanks Bro

Would you consider a backtest using these methods reliable enough to include in a business plan?

You can try.

But I suggest you run it by me or any of the other programmers here first. Give us the opportunity to analyze the backtested report and tear it apart as it were.

If your EA is NOT a scalper and goes for fairly big profits (>20pips) and trades are easily over 10 minutes. The 90% Alpari data can really help and is pretty easy to setup. (The longer your average trade and greater your TP the more accurate the backtests will be.)

My trades last over an hour and go for max pips until I detect trend reversal. I designed a system that could be reliably tested using this data. Not being able to accurately backtest will severely cripple your EA. If your creating a scalper you better go for the 99% tick instead.

http://tinyurl.com/backtest90

Here is a preliminary report.

Modeling quality if N/A

LOL!! - N/A. I just got spanked by trusting a 70% backtest on a currency I dont usually trade (so i had no broker downloaded history). Been so long since I did that I forget that anything less than 90% is PURE CRAP!!! (now i dont backtest on anything but Alpari databank data - 90% and Zero mismatched errors)

N/A is less than junk. Try again grasshopper.

Hi Raygun,
Can you tell me if this is correct:
To use tick data from gain capital, you need to use an earlier version of Metatrader that has the recalculate box to uncheck (build 208 I think)

The data from Gain all needs to be joined into one csv file and then transferred to an fxt file

Also, when using the script to transfer file to fxt, does the file need to be in a specific directory?

Thanks

The gain data is crap, it's full of holes and outliers.
Nothing good comes for free.

Be Brave - Show the rest of the stats. (This was generated with 90% Alpari Databank Data. Sweet Stuff and VERY accurate for my trading system. Avg trade length 58 minutes.) - I write EA's targeting 100+% per month. (Hoping to get 100% every 2 months in Worst Case but truthfully the Alpari data nails the trades EXACTLY!!)

Thats why I design systems that can be backtested with the 90% data. I may spend months writing the EA but the secret sauce is the correct settings. I use settings for at least 2 months that have diff markets (trending vs. ranging) and try to find settings that will work in both markets. Nov 2008 (choppy )and Dec 2008 (trendy) are a nice testing ground. Get a EA to work in both months and you're getting close to a winner.
_

Visual mode backtests seem to be highly accurate. I did one but unfortunately my EA didn't perform very well.

impressive performance what was the biggest lot size?

1st Trade - 1.00 LotSize
36th Trade - 1.40 LotSize
73rd Trade - 2.10 LotSize

fantastic its a custommade EA?Do you run it on live acc?

Yep, custom made EA (Its a greatest hits routines of EA's I've written for the last 2 years). I run it LIVE for myself - Testing it across multiple brokers right now. Still tweaking the settings at bit for best balance across diff months. I'll PM you some info after the market gets back to normal trading the begin of year - These holiday breaks are slowing my final tweaks.

Thanks, that's good to know.
Do you think using the gain tick data is a better backtest than the 1 min Alpari data?
Which tick data source do you think is the best for purchase?

Thanks

Fulltime, Am I reading your result correctly, your utilizing 15m bars and your backtest is only for the month of November?

This is the same EA. Yep that was only a Nov. Backtest - This is a Forward Test on an account that signals LIVE trades. The EA does 100%+ for Dec too. Here's the pic and post from EURUSD. I like to see how my EA compares to manual traders. I never touch it but its fun to see my results vs. LIVE trader opinion (most that don't believe that successful EA's can be written) - Also I backtest on 15min since most of my trades average about an hour or less. They bunch up too much on the chart if I use 30min or greater for chart views.
------------------------------------------

Wow! What a day - Is it still Xmas?! My SUPERTrend EA is cleaning Up!! Just wanted to show that EA's can trade on full auto. Of course it depends on the programmers skill but it can be done. FX-SUPERTrend is a EA thats a collection of my greatest hits routines from EA's I've been writing for the last 2 years.

EA's ARE available that work. They're not in the public space for purchase but a working strategy can be programmed to trigger without emotion if its simple enough. (this EA has a 60%+ Win Ratio and a 5.0+:1 Win/Loss ratio. It cuts losses pretty quick and lets winners run. Typical <15% DD)