[Date Prev][Date Next][Thread Prev][Thread Next]
[Date Index] [Thread Index] [Author Index]
Return to main CEDA-L Archive Page

Re: Pick 20+

In general I choose not to participate in the "grand discussion" about 
CEDA policy because I have the constitutional responsibility to help 
enact it.  In the case of the national tournament, I have had the further 
responsibility of "programming the computer" and operating it.  So before 
I begin I want to assure the community that I will always endeavor to 
implement the will of the community in terms of national tournament 
management.  At the point that I cannot implement a decision of the 
community or at the point that the President chooses to use a different 
package, I will gracefully step aside.

I do feel, however, that the impacts of Pick 20+ have been overclaimed by 
both sides and that we need to take a step back from the rhetoric of the 
debate.  First, I am concerned by the discussion that suggests that 
tabrooms (and perhaps by extension the national tournament tabroom) will 
intentionally discriminate against some program or another.  I want to 
assure the community that during the time that I have been associated 
with the tabroom at nationals it has operated with absolute integrity, 
both in the operation of the computer and in all decisions made by tournament 
directors and operations managers.  If Pick 20+ or any other type of MJP 
were adopted it would similarly occur without any intervention.  
Concerning Professor Rogers bad experiences, it should be noted that MJP 
does not in and of itself insure non-intervention by the tabroom.  Both 
my program and Rich's program permit computer-assisted judge assignment 
where the computer identifies AA, AB, BB, etc. judges and the computer 
operator selects from the list the judge which will be placed in the 
round.  Up until perhaps 5-6 years ago, this was the most frequent model 
for judge assignment, particularly in tournaments that used Rich's 
program.  So it is possible that a tournament that advertises MJP could 
still use tabroom intervention in the assignment of judges (though most 
tabrooms would still display high integrity in doing so).  Tournaments 
that I have run, whether MJP or not, do NOT use tabroom intervention.  
The computer assigns the judges using the protocol established.  I would 
suggest that all tournaments that use computers indicate whether the 
computer will automatically assign judges or simply assist in the task (I 
honestly believe that most do it automatically for time reasons if 
nothing else).  Nationals will use automatic assignment regardless.

The second issue that is perhaps misunderstood by both sides is who will 
get first crack at the judging pool.  Tuna notes that Rich's program 
starts in the middle of the bracket.  The protocol outlined in the 
advisory referendum does not.  Since the protocol says that from round 3 
on the computer will assign judges in the following order: DOWN 3, 2, 1, 
0, 4, 5, 6, 7.  In practical terms the following happens:

	Round 3:  Since the "worst" team has 2 losses, the computer starts 
at the bottom and ends at the top.  If we assume that 50% of the field 
would get mutual A's (optimistic if teams pick 20-25 judges and regional 
constraints are maintained), and given that their were 108 rounds at 
nationals last year, roughly the bottom half of the tournament would 
receive mutual A's, the rest B's (this is never absolute, however, given 
different preferences that teams might have - e.g. a pairing might get a 
mutual A even after 25 pairings fail to receive one).

	Round 4:  Since the "worst" team has three losses, the protocol 
would again start at the bottom and end at the top.

	Round 5:  Last year 9 pairings in round 5 involved teams that 
both had four losses, so the computer would start at pairing 99 and 
perhaps make it as high as 45 (a 2-2 pairing).

	Round 6:  Last year 21 pairings in round 6 involved teams that 
both had 4-5 losses, so the computer would start at pairing 88 and 
perhaps make it as high as 34 (a 3-2 pairing).

	Round 7:  Last year 37 pairings in round 7 involved teams that 
both had 4-6 losses, so the computer would start at pairing 71 and 
perhaps make it as high as 17 (a 4-2 pairing).

	Round 8:  Last year 54 pairings in round 8 involved teams that 
both had 4-7 losses, so the computer would start at pairing 54 and 
perhaps make it as high as 1 (a 7-0 pairing).

Given this scenario there is good news and bad news.  The good news might 
be that teams at the bottom are not unduly discriminated against, 
particularly for the first 4-5 rounds.  The bad news, if you are a team 
at the top fighting for a good seed and trying to avoid being in a break 
round, is that you may never be in a good position to get a mutual A.  
Now that may well be fair, but I have a suspicion that it is not what 
many advocates would be looking for!?

Of course, the above scenario depends on a 50% yield.  Tuna has talked at 
length about a 95% yield.  Of course, Tuna is describing a different 
system.  In Tuna's model, 33%+ of the field is designated A's.  In the 
proposed model as little as 12.5% of a 160 judge field might be 
designated A's (last year we had 167 in the pool so 20 would actually be 
closer to 12% - the proposal is not sensitive to the size of the pool).
Even advocates have noted that if everyone chose only 20 the number of 
mutual A matches would be very small (with regional constraints I would 
guess closer to 35% - without perhaps as high as 50%).  Of course, it has 
been widely suggested that if the proposal passed, most "rational" 
coaches would select many more than 20 - if they were "comfortable" with 
40-50-60 they would select that many.  They would know that choosing more 
would increase their odds of getting a match (particularly if their 
oppoent cooperated by also picking 50+).  Snider, Croasmun, and Hunt have 
begun to question that logic, however (and with good reason).  ABCX 
systems don't LIMIT a coach to picking 1/3 A's but yet most don't follow 
the logic that says picking more makes it more likely that you'll be 
successful.

Croasmun's reference to the Prisoner's Dilemma is precisely a propos.  
I may pick 50+ if I beleive that you will pick at least as many or more 

than I do.  But it's a risky choice (unless I really want my judges 
numbered 21-50 - and it may even be risky then).  If I pick 50 and you 
pick 20 you statistically control the likelihood of a match.  If both of 
us get a judge 1-20 then we're all happy (but then I should have just 
picked 20).  If you get a judge 1-20 and I get one 21-50, I may NOT be 
very happy.  I may be glad that I got 48 rather than 148 but I may not be 
so happy that you got 18.  In fact I might not think the preference is 
mutual (your real A list versus my acceptable A list).  The story is 
even worse if we DON'T get a "mutual" A.  My B list excludes all judges 
above 50 while your B list includes (21-50) which might actually be just 
as acceptable to you as my 21-50 are to me.  Again I might argue that it 
doesn't seem very mutual.  To protect myself, I may limit myself to 20 as 
well (Prisoner's Dilemma)!

One might argue that none of us would ever play the game this way, but 
I'd keep your money in your pockets.

Finally, I need to comment about Tuna's 95% claim.  My experience 
confirms that mutual preference does generate a very high yield 
particularly when A's are set at say 35-40% of the field (perhaps close 
to 100%), particularly if there is some "slack" in the pool that permits 
judges to have rounds off (arguably a good or BAD thing).  But the South 
Carolina tournament had some features that suggest that we should not be 
too quick to generalize.  1) 2 divisions - during the pairing of open, 
Tuna noted that nearly all rounds received mutual preferences.  Of 
course, the existence of a secod division creates an artificially large 
judging pool from which to panel open rounds.  CEDA Nats is a single 
division tournament with historically NO slack.  All judges who owe 
rounds must hear every round that they owe (last year 2 judges were on 
standby for round 8 who arguably could be said to owe the round).  2) a 
percentage of teams did not turn in preference forms (I believe it was 
approx 10% - forgive if I'm wrong Tuna).  In this case all judges are 
coded as A's.  When one suggests that ALL judges are preferred by someone 
it is critical to consider the teams that are only preferring them by 
default.  Similar blank sheets may or not appear at Nats so the 
statistics should be adjusted accordingly.  Two years ago, with only 10
strikes available per team, a judge was struck by 38% of the teams at Nats.

In general, I am not criticizing Tuna's claims about his mutual 
preference experiences.  They are rather widely replicated.  This would 
not demonstrate, however, that they would apply to the national 
tournament (they may - they may not).  Of importance, however, is the 
fact that the referendum concerns a different system.  Advocates as well 
as detractors need to compare apples to apples and oranges to oranges.  
If we did so, MPJ would prove to be neither so good or so evil as the 
rhetoric suggests.  

Sincerely,

GARY

PS.  Tuna's "I think I found a flaw post" says substitute ABCX for Pick 
20+ (ABX).  This may miss the point.  The key difference is not B's and 
C's vs. just B's.  The issue is HOW MANY A's should be selected (mandated)?
References:
Archive created by Jonathan Stanton (jonathan@cs.jhu.edu)
Return to main CEDA-L Archive Page