Backtest Multiple Strategies

If you are serious about trading and/or investing and are willing to commit a good amount of your net worth, you may choose not to follow just one strategy. You may be better off splitting your capital amongst not just different asset classes or stocks or etfs but also different strategies that trade those assets.

How can we simulate this?

One way is not to. You can develop good strategies independent of one another and invest in them as you see fit.
The other way is to simulate a multi-asset, multi-strategy portfolio as a whole.

You can think of a strategy as a time-series. SPY is a time-series. So is GLD, so is IBM. Just a sequence of numbers. So a strategy is it’s equity curve, an artificially made time series. You can invest in one or in multiple ones, just as if they were “assets” also.
20 years ago we should have diversified in different asset classes, now we may have to diversify in different strategies, as well.

So how can you do that? What tools to use?
There are many choices. I will briefly go through the ones I have tried. Others exists that might be better but I haven’t tried myself (NinjaTrader, TradingBlox, etc.)
_______________________________________________________________________________

  Amibroker

As you know I am a big fan of Amibroker. It is not an obvious choice for backtesting a multi-asset, multi-strategy portfolio. But as usual there’s many ways to do things in AB. The obvious choice is to backtest each strategy and export the individual equities. Then trade those equity curves as buy & hold. The downside is that it takes two steps to do this. The upside is that you can write a new script and develop rules or allocation schemes on when and how much to trade in each strategy. Another choice is to program multiple strategies in one afl script, so that both available funds as well as compound profits are taken into consideration.
This  can be done, with some limitations. If someone is interested, I can do a post with the afl code and logic.
________________________________________________________________________________

The more I work with this software, the more I like it.
In QuantShare, you first develop individual strategies. Each can trade it’s own specific basket of assets.  You can then combine strategies by using the combine-trading-systems plug-in.
 It asks you to choose which strategies to test, then combines them and returns stats and equity curves. By listing the stats of all the possible combinations, you can quickly see which combinations of strategies are better without going through a correlation analysis.
Another way to backtest multiple strategies is to write a MoneyManagment script. Using such a script (in C# or JScript) you can control multiple “categories” that have their own rules.
In the next post I will go briefly through an adaptive multi-strategy script.

_______________________________________________________________________________

I downloaded a 30-day trial and so far I am very impressed, especially with the ease it communicates with Interactive Brokers (as well as many other brokers and feeds) and the potential to run ATS (automated trading strategies) with many different brokers. I was able to set up a simple ATS system in less than 10 minutes and run it. This is definitively a contender when it comes to intra-day ATS systems.
That said, MultiChart can also backtest multi-strategy,multi-assets portfolios.

________________________________________________________________________________

Now, this is a very interesting piece of software.
a. You can perform multi-asset, multi-strategy backtesting by using “Accounts”. Each Portfolio has one or more accounts. Each account has each own instrument list, strategy, money managment script, as well as commissions and broker connection (for autotrading).
b. You can have a Master money managment script that “sees” all accounts under the portfolio and reallocates funds according to set rules.
c. You can have a Master risk-controlling script that “sees” all accounts and ,for example, rejects positions if different strategies tend to buy the same one stock. You can automate all this, not just backtest. As an exampe: Let’s say you have two brokers: Interactive Brokers and MB Trading. Under Portfolio I can create two Accounts: InteractiveBrokers_1 and MBTrading_1. Both can be auto-traded from IQ.  So even if you are a bit paranoid and don’t trust your broker,  you can execute through them only half your strategy! 🙂
e.IQBroker uses C# and allows you to import dll references.
f. It’s currently free for individuals.
So, what’s the downside?
A bit of a steep learning curve and no support (unless you pay).

Quantshare Trading Software on Amazon’s EC2

I wrote an article on how to automate the process of updating quotes, updating multiple trading strategies and e-mailing next day signals form the cloud to our own e-mail boxes. This time I am using QuantShare as the trading software and an Amazon EC2 micro instance as the host cloud.

You can read the article at the QuantShare Blog.
More…

QuantShare is a fairly new piece of software that can be used to develop and backtest trading strategies. It has many features (I will get into them in another article/review). I will just go through a few that interest me. Keep in mind that most of my system backtesting and development has been done in Amibroker.

1. It has a similar “language” and array based logic as Amibroker. That means one can develop quickly in Amibroker and then “detail” the strategy in QuantShare. More…

2. You can write a Money Management script in C# to extend the rules of the main strategy.
 More…

3. You can write your own indicators in C# and then call them from the array based language as functions.

4. It has a genetic optimizer that works with a “Rules” list. You can combine rules and “auto-generate” systems in minutes. Or better, take your existing system and “look” for filters or additional rules that might improve it.
 More

5. QS uses custom databases. They can hold quotes, fundamentals  Yahoo news, tweets, COT data, whatever. It has a downloaded tool that can be customized to handle all kinds of downloads and update the database. Ready made downloaders exists for different markets. These historical databases can then be called from the backtester. More…

6. There is a community and ready made “scripts” that can be downloaded through the “Sharing Server“.

7. You can “combine” systems. So you can see the equity curve of trading MySystem1 with MySystem2 and… MySystemN.  More…

8. Reasonably priced.

So what’s the downside?
The real-time version is still in Beta as of this writing ( 28/10/2012).  So you cannot connect to a broker or auto-trade, yet.

Amibroker – From Backtest to Trading – Going Live

Going Into Production 

So you have a system in Amibroker that you backtested and you are happy with. You now want to go “into production”.
How do you get your signals? How do you keep track of open orders, position sizes, ranking rules, etc. It’s not as easy as it may seem.

Explore or Scan?

If you try to run scans or explorations you will get Buy/Sell signals but how many? If you have 500 symbols and you trade the top 10 and your size varies with each stock’s ATR, you will have to rank the first 10 and compute, atr’s and position sizes and bla..bla..bla.. Do that day in day out… You will make a mistake.


Just Backtest!

In every Backtested system there are assumptions. Position Sizing, Ranking rules, skipping signals or not, etc. One way to keep track of all those things is to let Amibroker do it for you.

The following example assumes 
a. You use no stops
b You trade End Of Day at tommorow’s open or tommorow’s close.



Here are the steps.
1. Go to Amibroker and backtest your formula as usual. Check everything is ok, Commisions, Initial Equity, etc… Check that your backtested equity curve is what it should be.

2.Open the Backtester Setting.

3. In the Portfolio Tab, check the “Add artificial future bar” box.

** If you are trading ETF’s you might want to set “Limit trade size as % of entry bar volume” to 0.

4. Under Report select “Detailed  Log”.

5. From to Dates: Start today or a month ago depending on wether you are starting from 0 or synching with the current state of your system** (i.e. it is a long term system that holds Gold for 3 years now).
End day can be sometime in the far future: 1/1/2020

6. Check your margin if you are trading stocks. Max should be 50 (i.e.2x leverage) or higher if your broker allows for that. If you want no leverage set it to 100.

6a. Run the backtest. You should see a list. The last item should be for tommorow’s trade.

Here you have the orders ready for you to follow. The Position sizing rules and ranking are done for you. All you have to do is synch to the system (the first time) and then follow it every day (or week or month depending on your system).

Automating it

7. Save the .apx file that incorporates all your setting.
While the Analysis Window is selected and you are happy with all the settings go to Amibroker menu, and select File—>Save As.  A window will ask you where to save the file. Choose the folder where your AFL system formula is (or any other folder you want). Give it a name (My_System_1_Live) and save it as .apx, the default. Now you should have a My_System_1_Live.apx file that is all you need to backtest this strategy.

8. Write a Jscript.
a. Create a new NotePad test document
b. Paste this code and save it not as txt but as .js. Replace the file locations with your own.

//——-Code adapted form Thomaz’s code

var WshShell= new ActiveXObject(“Wscript.Shell”);
AB = new ActiveXObject( “Broker.Application” ); // creates AmiBroker object
try
{
// opens previously saved analysis project file
    NewA = AB.AnalysisDocs.Open(“C:Program FilesAmiBrokerFormulasMy_System_1_Live.apx ” );
    // NewA represents the instance of New Analysis document/window
    if ( NewA )
    {
         NewA.Run( 2 ); // start backtest asynchronously
         while ( NewA.IsBusy ) WScript.Sleep( 500 ); // check IsBusy every 0.5 second
        NewA.Export( “C:UsersUserMeDesktopLiveSystemsMy_System_1_Live .csv” );
         NewA.Close(); // close new Analysis
     }
else
     WshShell.Popup (“NewA problem”, 3,”Active X problem”,0);
}
catch ( err )

{
   //Email_Event(” failed to run”,”Exception: ” + err.message);
     WshShell.Popup (“Active X problem”+”Exception: ” + err.message, 3,”Active X problem”,0);
}
//end code

For example save it as “Run_Backtest_Sys_1.js
Saving it as a “.js” make it an executable JScript under windows.

9. Double clicking on it will run run it. It will launch Amibroker, open the .apx file you specified, backtest your system and export a report at the location you also specify.

10. Open the .csv file. You should see the backtester’s results. The last lines should be tommorow’s trades.

11. (Oprional) Parse the .csv file.
If you are good at jScript or VBscript you can write a script that will go to tommorow’s entries and parse them into arrays. You can then export clean csv’s or e-mail yourself the signals or drive a virtual system at Collective 2 or email your broker.

What about Stops?  
Using the Custom Backtester Object.

There’s many ways to do teh same thing. Here’s teh code for retrieving tommorow’s trades (and stops or other variables) using Amibroker’s Custom Backtester.

//-One way to get tommorow’s signals – 
//Make sure “Add artificial bar for tommorow”, under Backtest Settings–>Portfolio Tab is checked to //get tommorow’s trade
///////////////Put your system HERE/////////////////////
Buy=Cross(30,RSI(15));
Sell=Cross(RSI(15),70);
SetTradeDelays(1,1,1,1);
BuyPrice=SellPrice=0;
/////////////////////End System ///////////////////

////////////Variable that you need to report/////////////
stoploss=C-3*ATR(10);
StaticVarSet(Name()+”SL”,(stoploss));//place the stoploss inside a staticvar array. Include Name!
/////////////////////////////////////////////////

////////////////////////////Code for custom report///////////
//Make sure “Add artificial bar for tommorow”, under Backtest Settings–>Portfolio Tab is checked to //get tommorow’s trade
idx = BarIndex();
d = Day();
m = Month();
y = Year();
trace=1;
action=””;
Datestr=””;
report=””;

SetCustomBacktestProc( “” );
if ( Status( “action” ) == actionPortfolio )
{
    bo = GetBacktesterObject();
    bo.PreProcess();
    for ( bar = 0; bar < BarCount; bar++ )
    {
        bo.ProcessTradeSignals( bar );
        CurEquity = bo.Equity;
//if bar is the last bar on the chart set Date
     if(bar==BarCount-1)  Datestr= “n” + “n” + d[bar] + “/” + m[bar] + “/” + y[bar] ;

        for ( openpos = bo.GetFirstOpenPos(); openpos; openpos = bo.GetNextOpenPos() )
        {
SL=LastValue(Nz(StaticVarGet(Openpos.Symbol()+”SL”))); // Get the stop loss from the static array
if(Openpos.IsLong()) action=”Buy”; else action=”Sell”;

              if(bar==BarCount-1)  //only continue if we are at the last bar (i.e tommorow’s trade)
{
report= action+”  “+Openpos.Symbol()+ ”  “+ openpos.Shares()+” shares”+” Set Stop Loss @: “+ SL+”n” ;

PopupWindow(datestr+” n “+report,”Your Trades”,10);
Say(report+” and please remember to enjoy the rest of your day”, purge = True) ;
//AlertIf( True, “EMAIL”, datestr+” n “+report, 1 );

}
            
        }
    }

    bo.PostProcess(); // Finalize backtester
}