Would someone be kind enough to post, in detail, how the Rank Tolerance feature works in the Screener? I can’t seem to figure it out. I cannot find any documentation. Thanks!
There’s an orange Help button in the lower-left hand corner. 
From the Screener UI glossary:
Thank you! I had been wondering about that also. Google search wasn’t finding this.
If I can ask for clarification (sorry, my mind works better in concrete terms). As an example:
Rank tolerance: 10
Portfolio positions 20
Universe size: 1000
Given the above, would a rank tolerance of 10 on a 20 position portfolio in a 1000 stock universe mean that a stock would have to fall out of the top 22 (20+10%), or would it mean falling out of the top 100 (being the top 10% of the universe) in the rankings? Or something else? thank you.
Thanks for pointing out the Help button. On my computer it is actually in the lower right hand corner. 
I have done more work on this because the explanation is still not clear to me. I would not place any bets on this but it looks to me like the rank tolerance has nothing to do with percentage change. Instead, if one uses a rank tolerance of 5, the screener will not sell that stock because of a declining rank until that decline is more than 5 ranking points. For example, you buy a stock with a rank of 86 at time of purchase and you choose a rank tolerance of 5. The screener will hold that stock until it falls below a rank of 81. Of course, there could be other rules that cause a sale but not ranking. Also, it is all about the starting rank when the stock is purchased, not a week or a number of weeks later. For my universe and my screen, rank tolerance adds the most to performance when it is set at 10.
If I am wrong (or right) I would like to know.
OK, so the button is on my other left.
I also need to correct the documentation in this case. It should be referring to percentile ranks rather than percentage points. That is, if it required a 98 to pass the filter and you had a rank tolerance of 5 then it will still pass the filter if it has a rank of 93 or higher. This is literally the only case in the screener where the backtest cares about what the result was last week.
I don’t think the backtest cares about the result last week unless the stock was purchased last week. It only cares about the rank when the stock was purchased regardless of when that occurred.
I suppose that I should clarify: For any other test, the screener backtest doesn’t care about what it said in the last rebalance period (or any other rebalance period). If it is going to have a 100% turnover, so be it. This is the only setting where it even looks at what it held in the past. (I tend to think in weekly rebalance period terms.)
Portfolio size: 10 stocks
Rank tolerance: 4
Hello, so if you were to run this strategy live, do you simply include the score of your number 10 stock (let’s say 98.5) and subtract 4 and include any stock you’re already holding that scores a 94.5 or better in lieu of the #10?
Not quite. You would first look at the new ranks of every stock you own. If the stock is in your top 10, you keep it. If the stock’s rank is within 4 of the rank it was when you first bought it, you keep it. Then, if you keep fewer than 10 stocks, you add the top-ranked stock(s) that you don’t already own.
To get a feel for how this works, run a screen backtest using a ranking system or a quick rank with Max # of Stocks set to 10 and Rank Tolerance set to 4 and the “Save Log” box checked. After running the backtest, click on “View Log.” As you scroll down the log, study what the screen does each week.