Have anyone tried out something like this with success : (Rank - RankPrev(1)) / abs(RankPrev(1)) < -0.05
Sell there is a big correction in the stock score ranking?
Ore what about land momentum: Eval(Country("NOR,SWE,FIN,DNK,DEU,NLD,BEL,FRA,ITA,ESP"), Pr52W%Chg - Pr52WRel%Chg, -1000) //
I manually check every week and sometimes midweek for any large rank drops (more than 5 in my case but similar concept) for my higher turnover account . Curious to hear others’ thoughts
I tried that in my current model and it reduced my overall returns and increased my turnover.
Yes, precisely these formulas, and also the land momentum in the ranking system, didn't work well for me either, but here the question was more about whether anyone is using something similar.
In other words, is there any reason to pay extra close attention to a company that has had a major correction in its rank recently?
Do I understand correctly that you are monitoring stocks that have fallen more than 5 positions, but over what period – the last week? And have you had a chance to test how your approach has performed in backtesting?
Something like this: RankPos - RankPosPrev(1) > 7
I monitor all my positions weekly (or daily depending on which of my short-term accounts) and sell if rank drops below say 90 and I find a better option. The way I see it why would you not want to hold the best available opportunities? Fixed holding periods are arbitrary to me and I tend to have lower turnover since a lot of effort goes into the picking.
I do not use the simulation feature a lot (just a few times a year) because it takes a very long time for me and also because I know I probably will not follow it exactly. I use mostly just the ranking where 90th and below significantly underperform 91th percentile and above across all tested holding periods. I also look into the cause of the drop. Is it a takeover? A news or earnings gap up? A huge miss with horrible guidance? If the outlook has completely deteriorated I would definitely sell. This is part of what I would test or study. The actual drivers. For example you could create a rule if bad sales surprise and downward revision and rank collapse sell. Just an idea to illustrate what I mean.
I think 95th as a threshold could be too tight for me because it still is a good-performing bucket but do use 90th more often. I do sometimes use 94/95 if i find a great higher ranking opportunity. It just depends on what the model is recommending and if I think its better.
I treat my models like analysts pitching stocks to me. Do I recommend others do this last part? No- do what works for you based on your own strengths and weaknesses and own tests