The power of the internet.

Jan 28, 2009 17:44

I EXPRESSLY RESERVE ANY AND ALL RIGHTS TO THE CONTENT WITHIN, AND EXPLICIT WRITTEN PERMISSION MUST BE OBTAINED TO REPRODUCE ANYTHING WITHIN THIS ENTRY, IN WHOLE OR IN PART. THIS INCLUDES ALL TEXT AND PHOTOGRAPHS.
---

On University Challenge (Illustrated)All kinds of things about the show and our three matches for anyone interested ( Read more... )

lincoln college, exeter university, oxford, 26th january, corpus christi college, pembroke college, exeter, sheffield, university challenge, jeremy paxman, oxford university

Leave a comment

anonymous January 29 2009, 14:55:21 UTC
Your idea of the viewers-at-home exercise, to make them appreciate how it actually works, is so true, esp. the habit of counting both teams' bonuses to one's credit. I would almost put it down as a slightly (but only slightly) tongue-in-cheek Wikipedia link for the UC entry. On a day where Paxman is less prone to rushing them, the contestants might get through only twenty/very low 20s starters, and not only are these crucial to any point-scoring at all, but it leaves open a large element of chance. People who did schools-level quizzes on the buzzer - probably a surprisingly large proportion of UC viewers - are likely more used to bonuses passing across to the team who failed to buzz if the team with first call on them does not get them right, a set-up which makes the stakes lower and the lowest scores higher. Or failing that, they're familiar with pub quizzes where no team's greater speed stops them writing an answer down.

Though of course some teams each series are very good and you know they will go far, for the vast majority - perhaps the whole field bar one or two - I think that, because of the element of luck, any attempts to predict how they will do by extrapolating their and their opponents' previous scores tend to fail more spectacularly than in any other competition, and that's saying something as it's never a particularly reliable indicator in anything. I never forget from Sean Blanchflower's stats site that the greatest margin in the Paxman era, a difference of 360 points, was inflicted on a team that managed to reach the semis.

At least you do have the two previous rounds to your credit. I always feel so sorry for those who go out dismally in the first round, as they have nothing to back them up in the viewers' eyes, the public being oblivious to the qualification process which has to whittle what is sometimes not far short of 200 teams down to 28, and which can only do this with questions at the harder end of even the UC spectrum. I think that many people, who admittedly give it little thought, assume that all universities can get on automatically - which unless we have the same number of institutions of higher education as during the 1930s, is hard to explain. Do they think it's a coin toss or a royal rumble or something? It's also good that you mention the slightly unspoken cap on Oxbridge colleges, as to read the New Statesman last December, or to know of the old, ongoing bourgeoisie-flagellating debate of which it is part, it is clear that notta lotta people know this.

Just one little thing: 2/4 of the Corpus team were in fact undergrads. Not that this really detracts from much of what you say.

Reply

lystellion February 16 2009, 21:00:15 UTC
Thanks for the comment! I think that the buzz-at-home exercise is absolutely essential for anyone to do who has strong opinions about people on the show. Especially at the latter stages, where the speed of the contestants is generally a lot higher and speed comes into play all the more.

If the questions were written down, then for the majority of them I would say that at least 2 of the 8 contestants would know the answers. That's where the speed is so important. It's not only easy if you know the answer. It's only easy if you know the answer before everyone else, and you or a team member hasn't interrupted the question too early, and if your confidence isn't so dented that you're too scared to buzz in.

I've corrected the Ph.D thing. I had a feeling it might be wrong. Turns out they're younger on average by three years than I thought too - which makes their performance all the more impressive.

Reply


Leave a comment

Up