Re: On Powers and Bonus Conversion

jpahk <*no_reply_at_yahoogroups.com*> · Thu, 13 Feb 2003 19:46:32 -0000

--- In quizbowl_at_yahoogroups.com, "bucktowntiger <jdh22_at_c...>" 
<jdh22_at_c...> wrote:
> Check this out... and this could happen realistically...

depends on your definition of "realistically"--to me, your example 
seems extremely far-fetched. certainly there are elements of realism 
here, but...

> Let's say 
> that you have a team which does not have a great breadth of 
> knowledge, but has profound depth at the subjects in which it 
knows.  
> Let's say that said team doesn't get many tossups in general and 
> goes, say, 4-8, but out of the tossups they get, over 80% are 
powered.

sure, let's *say* that, but have you ever actually *seen* a team 
power 80% of its tossups? have you even ever seen a team power half 
of its tossups over the course of a tournament? no, of course you 
haven't; and the only teams that even come close are the super-teams, 
who are pulling in 80 powers in a tournament.

> Of course, the team won't get to hear many bonuses, but let's say 
> that they happen to get all of their bonuses in their areas of 
> expertise, and they average over 25 points per bonus.

if this team is dominating enough in its specialty areas to win four 
games, they must be answering about 1/4 to 1/3 of the questions, or 
at least 5 or 6 tossups per game. so that's about 70 bonuses they've 
been asked in this hypothetical tournament, and you are saying that 
they "happen" to get all bonuses in the roughly 1/4 to 1/3 of 
subjects where they have very deep knowledge? the odds against this 
are astronomical. (in fact, astronomy itself doesn't come close to 
having numbers this outrageous; the age of the universe is only on 
the order of 10^17 seconds). even the odds of 2/3 or more of their 70 
bonuses being right up their alley are vanishingly small.

seriously, unless a team is answering so few tossups as to make their 
S-value entirely moot, their bonus conversion rate will be a very 
accurate reflection of their actual knowledge coverage across all 
areas. power TU rate, as several people have pointed out, are 
somewhat less reliable, especially given that they suffer more from 
small sample-size issues.

fortunately for us, it looks like NAQT is using a healthy measure of 
just plain points per tossup heard (tossup and bonus points), and 
being confident in their formula correctly adjusting for strength of 
opponents.

i'm posting too often. see you guys in a month or so.

joon