Nats Seeding

[Howard Rush]

I've been trying to stay out of trouble, but I thought this was amusing enough to post.

For the heck of it, I did another simulation of seeding vs. random circle allocation.  I was a little more scientific this time, I think, but came to pretty much the same conclusions as last time.

I took the order of contestants by actual performance capability at a given Nats, then added a seeding error to each.  I assumed that the error in how each guy is seeded is random and has a normal distribution with zero mean and some standard deviation.  I ran the simulation with standard deviations of 2, 4, and 6 places. I added this seeding error (a different random number to each guy each time) to each guy's ranking, resorted the ranking, and folded the seeds with the new ranking into the four circles per our Nats process.  I then sorted each circle by actual performance capability and recorded the five in each circle who qualified.  I compared the various seedings with random circle allocation.  I probably should have seeded about 70% of the contestants, as we do in Open at the Nats, and done the rest randomly, but I didn't know how to model that. 
   
One peculiarity is the lumpiness of the random outcome.  It didn't get any smoother with number of trials, so I guess it's something real.  It could be some error I made, but I suspect it has to do with how guys stack into the four circles. 

The upchuck is that seeding has little effect on the 20th best guy-- at least with a 40-contestant Nats.  It appears to affect the 16th-best and 24th-best guys most.  The top guys always qualify.

[Tim Wescott]

Make sure to archive this so that your numbers can be checked at the 52015 Nats.

[Perry Rose]

You need to re do it, you forgot to carry the 4.

[Crist Rigotti]

Interesting.  Thanks Howard.

[Dan McEntee]

  I thought this was going to be about improving the grass circles??
   Always nice to hear from Howard.
  Dan McEntee

[Howard Rush]

I reran two cases with 56 guys to see if the crossover at 20.5 was a function of the number of guys advancing to the next round or the number of contestants.  The crossover stays at 20.5, so I guess it's a function of the number advancing to the next round.  A statistically astute person would have known the answer, but I had to use brute force.

Feb. 27: Oops, that last plot had the wrong title.  Here's a corrected one.

[john e. holliday]

  I thought this was going to be about improving the grass circles??
   Always nice to hear from Howard.
  Dan McEntee

Boy that would really separate the men from the boys.  One official flight on the grass and one on the pavement that counts towards advancing to top twenty.   Well I guess Howard is going to keep crunching numbers.

[Paul Smith]

So who is the ultimate judge of the truly-deserving winners to determine if the seeding (or random) plan worked?

Maybe if the contest is run and NOBODY complains then the system was valid.

[Tim Wescott]

If the contest is run and NOBODY complains then you should double-check to see that you haven't stumbled into an alternate universe.

[Brett Buck]

So who is the ultimate judge of the truly-deserving winners to determine if the seeding (or random) plan worked?

Maybe if the contest is run and NOBODY complains then the system was valid.

   Look at the converse of that argument  - is one person complains, it's invalid. This is how we have made some of our biggest mistakes over the years. Just because someone complains doesn't mean the complaint is valid/sensible/legitimate, or deserves a response. PAMPA did that for years, and (in hindsight) it appears to have been a mistake.

   Brett