Feature Request: FMin(), FMax() and others to augment existing cross-sectional functions

[primus]

I am requesting additional cross-sectional functions to assist in the analysis of aggregates and sets.

One potential use-case is covered in a recent thread “MORE DETAILS ON HOW THE DYNAMIC POSITION SIZING WORKS IN THE BACKGROUND. AN EXAMPLE…”. But the use-cases do not stop there; I am confident that other users will find these additional functions handy.

Specifically, I am requesting the addition of the following functions:

1. FMin(“Formula” [,scope, excl_zero])
2. FMax(“Formula” [,scope, excl_zero])
3. FPercentile(“Formula”, k, [,scope, excl_zero]):

Returns the formula value of the kth percentile within set. A generalization of several functions which functions identically to Excel function “Percentile.Inc” with addition of optional P123 scope parameters. For example, FPercentile(“MktCap”, 50, #All, False) would return the same value as FMedian(“MktCap”, #All, False); FPercentile(“MktCap”, 100, #All, False) would return the same value as FMax(“MktCap”, #All, False).

Likewise, it is the inverse of the Rank function. For example, ‘MktCap > FPercentile(“MktCap”, 90, #All, False)’ would function identically to 'FRank(“MktCap”,#All,#Desc) > 90.
View the feature request here.

[primus]

Add to this:

LoopRank(“formula”,“k”,iterations[,start,increment,noNAs,break])

This would be useful for assessing a security’s history as its own “universe”. For example, a stock can have an attractive earnings yield with regard to an absolute threshold (e.g., NetIncBXorTTM/MktCap > .1) or to a cross-section of the broader market (e.g., FRank(“NetIncBXorTTM/MktCap”,#All,#DESC) > 90), a peer group (e.g., FRank(“NetIncBXorTTM/MktCap”,#Industry,#DESC) > 90), or to its own history. Currently, I think assessing a securities valuation relative to its own history is under-regarded on this platform. Sure, there are workarounds, but nothing so robust as being able to perform a non-parametric rank test.

For example, how do we know that 10% is an attractive earnings yield for a stock like AMZN which has no (GAAP) earnings? Well, if we wanted to look at the its current P/E versus where the stock has traded over the years in weekly increments, we could do as follows:

LoopRank(“NetIncBXor(Trunc(CTR/63),TTM)/SharesFD(Trunc(CTR/63),TTM)/Close(WeeksIntoQ*5+CTR)”,“NetIncBXorTTM/SharesFDTTM/Price”,51,5,0)

So, AMZN’s current earnings yield of 0.3% would be assessed relative to the company’s history which we can see, from the following graph, has ranged between .3% to .6% over the previous year. Thus, a very simple visual examination indicates that AMZN is expensive relative to its own history. However, there is no currently not a way to perform this easy test using programmatic methods.

//dpa

---

[hemmerling]

I agree that a min and max ‘factor’ function is vital. As it is - we need to use a massive series with fhist and then apply a min and max to that series.

But a min or max with a formula would be so much faster and better. My only addition would be to specify the look back period and how frequently to sample. So you could have a one year lookback and weekly sampling or 5 year lookback and quarterly sampling and so forth.

[WalterW]

FMin & FMax shouldn’t be much of a stretch for P123. After all, it’s kinda available in the Custom Series tool as Universe Operation UnivMin and UnivMax.

As for LoopRank, I like it very much.

At one time I proposed loop stochastic as a convenience function. It would replace the stuff on the right with a call that looks something like the left-side.

LoopStoch(“GMgn%TTM”,12,4) = (GMgn%(4,TTM)-LoopMin(“GMgn%(CTR,TTM)”,12,4)/(LoopMax(“GMgn%(CTR,TTM)”,12,4)-LoopMin(“GMgn%(CTR,TTM)”,12,4))

The LoopRank seems more flexible since it accommodates the datum to rank (as ‘k’, if I’m reading things right).

Walter

[primus]

I’d like to expand this request to include Daniel’s suggestion for a function which returns the k-percentile rank of a value x. See: https://www.portfolio123.com/mvnforum/viewthread_thread,11438

It is the same as Excel’s [font=courier new]PercentRank(array, x[, significance])[/font].

Translated it might look like:

FPercentRank(“formula”, value [, #Scope, #Sort])