Geographic strength

[D+F+P=Hadouken!]

Originally Posted by Mr Epee

Why am I not surprised that you'd want all the juicy details about another man's butt infection? GROSS! Let's just say that it was few years ago, and I recovered.

I'm seriously starting to worry about a few people on this board...

A butt infection? Thats what you get for fooling around with other men.

[mrbiggs]

Do you factor in that people shouldn't be counted twice if they're on more than one list?

[KD5MDK]

Originally Posted by PeterGustafsson

Hi!



OK then. May I suggest this algorithm:
1. Compile a list of the X best in each region for men´s epee regardless of age, X being a number 10-20 or so.
2. Compile a list of all bouts fought within the latest year where both fencers are on the top X of different regions.
3. For each pair of regions, compile the list of bouts fought between top-X of those two regions. You will get a results pair, numbers Y-Z, where Y denotes #bouts won by fencer from region 1, and Z denotes # of bouts won by fencer from region 2.
4. Redo step #3 for all possible region pairs
5. Put in all Y and Z into a big matrix.
6. Use the matrix to find the interregional ranking. This can be done by using eigenvectors, or many other methods. The literature on this subject is extensive, I recommend you to google "Condorcet Voting" and "Smith Set".

I think the scale issues there will come into play quickly. Also, suppose there are 5 regions. Each of them has 20 of the top fencers, evenly distributed. However, some of the regions have dozens of other high level fencers outside of the top 100. Do they matter?

[Army Fencer]

Originally Posted by Goldgar

If I were you, I would include Veterans point standings in my calculations, but not weighted the same as Senior point standings. Or maybe you'd use a cutoff -- only the top N places on the Vet point lists count.

I think that's about what I'm going to do. It's what I've done with all the lists except SME; it's now just a matter of deciding how much weight to give them.

Originally Posted by mrbiggs

Do you factor in that people shouldn't be counted twice if they're on more than one list?

With the significant exception of the Hurleys, this isn't much of an issue, but I'm still working out how I want to deal with this.

[Army Fencer]

Originally Posted by PeterGustafsson

Are you going to propose boundary redrawings? If so, I foresee problems - one set of borders may be the fairest for epee, who inappropriate for ROW weapons. How do you find the best borders in that case?

I have no desire to open that pandora's box. But if I were, why should I care about the other weapons?

[Army Fencer]

Originally Posted by smurfette

I hope you are also taking into account who ACTUALLY fences somewhere. Lots and lots o' people fence in the Metro NYC division. A number are not actually there very often.

That brings up another important issue that I'm trying to deal with. A fencer can be tied to a region through three main ways:

1. Where they train
2. Where they compete
3. Where they grew up/what region had the greatest impact on that fencer

Often these are the same, but the more elite the fencer, the more likely that they aren't. "Where they train" is most interesting to me, and so I've assigned fencers primarily according to that.

[Redblade]

In my latest Heroclix(tm) acquisition, each Fencing Region figure has a limited edition chase variant that incorporates veterans in its power matrix. And in each battle we've had so far, the veterans-inclusion figures have trounced the non-veterans sets.

(This is my roundabout way of curiously addressing the underlying question, "What is your purpose for measuring the 'strength' of a region?")