Does anyone know how to use Excel to read ranking results from the P123 web page? I am working with the theory that different factors work well for different markets. I am therefore trying to test ranking systems in different market environments.
Over the last year or so very few factors have worked as well as they used to. For this reason I stopped all trading based on P123 since February. (No I wasn’t smart enough to stay out of the market altogether. I lost some money using a different system 
). Most of the factors stopped working even before the latest market correction. It was reminiscent in a way of how PSR stopped working in 07 before the crash. I have a few theories to partially explain these things and I hope to find some reliable patterns. My goal is to make ranking systems that complement each other (especially a long/short that works well under all markets), and/or ranking systems that “turn off” certain ranking factors under certain conditions.
Using Excel for existing ranking systems shouldn’t be too difficult, but is not something that can be done with Stefano’s Excel add-in or even with Jerrod promising new add in - yet. I would like to test how existing ranking systems work over different periods and for different sectors.
Thanks,
Chip.
Hi,
You can run a ranking performance test with historical returns. Then you can download the results to a spreadsheet and look at how each ranking bucket performs over time.
Don
I was hoping to be able to test each ranking system in 6 different periods. Even with only 10 ranking systems, that would be 60 different manual downloads, and each one would first need to be downloaded as a separate spreadsheet and then the information consolidated into another spreadsheet which measures the correlation of each system and each period with a theoretical perfect ranking system… It’s just too much work to do by hand.
There must be a way to do this with Excel automatically.
Hi,
If you run 10 different ranking systems across the entire history, then you can break down the different time periods in the spreadsheet and add the correlations. Doesn’t seem that hard but maybe I’m not clear on what you are trying to do. How do you get a theoretically perfect ranking system?
Don
Hi Don,
I apologize if I wasn’t clear. I don’t have an easy time explaining what I mean in a clear way but I will try. If something is still not clear please feel free to ask me.
When backtesting a ranking system I like to compare how well the system did compared to what an ideal ranking system would have done during the same period. Obviously the best possible ranking system would give you different results for an up market as compared to a down market, but the point is to see how well it does compared to what was possible during that period. Therefore each market period uses a different benchmark.
I am working on a reliable long/short ranking system that always has positive returns no matter what the market is doing. To do this I would like to first test how all factors perform under various markets. But imagine building Dan’s Spreadsheet by hand! Excel can easily automate the process if I find out how to do it.
If anyone knows of a way to do it or at least enough to point me in the right direction I would be glad to share my code with him.
Chip