I use P123 mostly for discretionary research and stock picking and research. IN looking at ranking outputs, i notice a couple of anamolies. I am not sure if this is a bug or an error of understanding on my part. I ranked using the same ranking system, with NAs set to neutral and
when i input the a list of larg cap tickers and make russell 1000 the universe - a couple of tickers are missed (CHKP, BUD). Both are definitley large cap. They are ranked when all fundamentals is used as the universe.
More alarmingly, the order of ranking changes when different universes are used. I would think this should not be the case as Prussell1000 should be a subset of all fundamentals.
If you redo this with a single ranking factor I think you will find that the order is preserved. I could be wrong and maybe you are only using one factor already.
But switching of the order when additional stocks are added and there is more than one factor is a know (and normal) thing. Although it is definitely counterintuitive.
I know this is not your only question. Perhaps P123 can correct me on the above if I am wrong and give a complete answer to all your question. They are well informed on this (obviously) and can expand upon or correct what I have said.
Just wanted to echo the comment that in multifactor ranking models relative rankings will change when the universe changes. There’s not one global rank of all possible companies, but rather a new rank is calculated starting with each new universe. I have some factors that are fairly specific to the universe included. For example, say you have a factor that prefers stocks that rank highest around 50th percentile in revenue growth. I’m just making up numbers here - but that 50th percentile of growth may be 2% sales growth in large cap universe and might be 4% sales in a larger group that also includes more smaller companies. (Again, just making up numbers - I haven’t tested this for actuals - but it will very likely be different). I have factors that are dependent upon the distribution of factors in the universe. There are alot of factors where below the 20th or 30th percentile perform poorly and might recieve a negative flag in the ranking - but if I build a universe, or restrict my universe to exclude most of those companies - then that same 20th or 30th percentile rank calculation isn’t applicable in the different restricted universe.
thanks Guys… that was helpful. But i still do not understand why the order of rank changes. I can understand ranking score change. I will test with a simple 1 factor ranker.