Question about Conditional Rank Node

Research
1

I’m a bit confused about how the new “Conditional Rank Node” feature works, the key question is:

  1. Is the value of the formula under each branch (true, false) calculated and then the root node ranks based on the formula result.

OR

  1. Is the universe of stocks under each branch treated independently, ranked accordingly and then the rank is propagated up the tree?

In the latter case, if the universe under one branch is nearly empty compared to the first branch, then does it mean the ranks assigned will be normalised across the range of the small universe, e.g. 10,20,30…100 if you have 10 stocks?

2

Edit: What I get from the help section is inconclusive. Good question.

3

Jim - do you think that means at no point will the forward PE be compared with the trailing PE?

4

Oliver,

Edit:

I understand your basic question better. One stock can get a rank in one branch and another stock can be a ranked in the other branch. How are those stocks ultimately compared. Can there be 2 stocks with a rank of 100? That is a great question!

5

I have never really been successful using the conditional nodes because there is no insight as to what is actually going on within the ranking algorithm. However, I’m willing to bet that you have to use conditional nodes in conjunction with the buy rule FRank (…,…,#GroupVar) to get anything useful. The GroupVar was thrown into FRank some time ago with no real explanation of what its use was.

Steve

6

Any chance for responce form P123 ?

7

A regular composite node takes all the scores of nodes it contains, sorts them, and does a percentile score (re-stretch values back to 100-0)

A conditional node is like a simple switch to decide where to get the score from. It takes the scores either from the TRUE branch or the FALSE branch depending on how the stock evaluates the condition. Then it does the same thing as a regular composite node.

We typically use the Conditional node with a condition that is the same for all stocks. For example our “P123 Multi-Market rank” uses a bullish condition (S&P500 estimates are trending up) to either rank based on bullish factors or quality factors.

Which help section was inconclusive so we can clean it up? Thanks

8

I’d like to build a conditional switch that behaves one way when the benchmark is up, and another when the benchmark is down (so the rule is “BenchClose(0)>(SMA(365,#bench)”. When I build it, I get this error: “Benchmark not set for function BenchClose()” The benchmark is set in the “General” tab, but doesn’t seem to be recognized in the rankings. Any thoughts on how to set it in the rankings?