Open Positions in WTT

[JulianCohen]

I have been having some issues with my WTT system. Basically the open positions change every now and then for no apparent reason. It only happens maybe 5 or so times in a year but I have just had a run of three changes in the last four weeks so it is starting to concern me.

The way I run my system is using Russell 3000 (NOT current and past) from a specified start date, 1/6/2016. This keeps things consistent and removes start date bias.

Norgate has a link specifically for this issue. It says

“My trading system backtest shows a different sequence of trades than previously recorded. How could that be possible?

You may be testing on a particular class of stocks where membership of the class is not fixed and no information exists as to historical membership. For instance, you may be testing on “Nasdaq” stocks. If a stock that was previously listed on Nasdaq changes to NYSE, then any trades previously recorded for the stock will disappear from the backtest. Similarly, if you are backtesting on stocks in a particular sector or industry.”

It would appear from this that when you run a portfolio test the interplay of all the stocks affects the outcome. If the basket of stocks this week is different from last week, that will affect the outcome. Since all the trades are overlapping, it’s a very difficult job to find the ones that have come in and affected the result, or the ones that have left that have altered it.

So my question is, is there any reason why I shouldn’t run my WTT operationally on the Russell 3000 Current and Past dataset, as that will probably eradicate the issue?

[GlenPeake]

Hey Julian,

You’re not alone, I seem to recall seeing a similar example(s)…. as you’ve described.

I.e. on my WTT on the ASX…. I was running backtests on the XAO universe and ISX.au was in the results (and I was holding it at the time) so it was matching…. When I ran another backtest recently ISX.au was not listed…. which I though was odd.

I’ll have a look through my Amibroker Reports directory to confirm, but I’m pretty sure I saw a discrepancy. i.e. Amibroker saves all your backtests in a report directory e.g. “C:Program FilesAmiBrokerReports” so I’ll check it out etc.

[JulianCohen]

Yup that’s the issue.

I have always run day to day operations on the current database, ie Russell 3000 in this case. But I can’t see any reason not to run on the current and past, unless I’m missing something, which I often do

[Stephen James]

The only issue with using Current and Past is the time of processing the scan, as it contains all current constituents, As long as you have the Norgate code active.

The current watchlist only is fine for daily management but as soon as you start signal checking with previous dates as far back as you are then Current and Past is best.

[Author]

Posts