Backtester - how to smooth out random high CAR years

AFL Programming
1

Hi, I am optimising a portfolio over 20 years.

Sorting results by CAR/MDD gives some very high CAR years and other years that are negative - however, over 20 years giving a good CAR.

This is untradeable in real life, however, because it is hard to sit through poor return years to hit the good years.

Is there a metric to use that smooths out the year on year return and gives the best smoothed equity curve slope?

Thank you

2

You could check out the Ulcer Performance Index, which many find to be a better reward/risk metric than CAR/MDD or even Sharpe Ratio. And of course you could define your own master performance metric, as authors like Van Tharp and Howard Bandy have done.

2 Likes
3

Thanks mradtke.

Does ulcer performance index help smooth the years that are still positive, but a low positive, eg: YOY:

12%--7%--7%--9%

is better than:

20%--1%--2%--15%

4

I am experimenting with ideas for this, and am going to put this out there.

I shall test it further.

Anyone else, please try it out and feel free to amend - and post any comments.


SetCustomBacktestProc("");

if( Status("action") == actionPortfolio )
{
    bo = GetBacktesterObject();
    bo.Backtest();
    
    eq = bo.EquityArray();
    st = bo.GetPerformanceStats(0);

    BI = BarIndex();
    t = LinearReg(eq, BI);
    yy = StdErr(eq, BI);
 
    // Higher is better
    bo.AddCustomMetric( "newMetric", LastValue(t) / LastValue(yy));
		
}
1 Like
5

For what it is worth, standard error of equity is included in built-in stats System test report window, so there is no need to calculate it the way you do.

6

@traderuk99 it seems like you might also be interested in something like the "K-Ratio" by Lars Kestner.

2 Likes
7

I had forgotten about Kratio - thanks!

8

This topic was automatically closed 100 days after the last reply. New replies are no longer allowed.