Tuesday 9 April 2013

University Challenge: Semi-Final Preview

So, after 34 matches, some nondescript, some truely memorable, we've finally found the final four. They are:
  • University College London: Adam Papaphilippopoulos, Tom Tyszczuk Smith, Simon Dennis and Tom Parton (895 over four matches) 
  • New College Oxford: Remi Beecroft, India Lenon, Andrew Hood and Tom Cappleman (770 over four matches)
  • Manchester: David Brice, Adam Barr, Richard Gilbert and Debbie Brown (955 over five matches)
  • Bangor: Adam Pearce, Mark Stevens, Nina Grant and Simon Tomlinson (830 over five matches)

So, those are the raw stats, but they only tell half the story. What about the actual performance records?

Well, U.C.L. have been the most consistently strong team overall, scoring well and strongly over their four matches, with their only real 'scare' being when they fell well behind Bangor in the quarters, but they soon recovered from that. All four players have contributed well over the matches, and the bonus work has been fair as well.

New College were fairly strong early on in the contest, scoring well in the first two rounds on both buzzer and bonus. But, in the quarter-finals, they seemed to drop down a gear, relying mainly on buzzer strength, and not doing too well on the bonuses. And Tom Cappleman has been getting most of their starters, raising a question over whether the rest of the side can make up if he has an off day.

Manchester got off to a shaky start, trailing Lincoln College Oxford until the last half second of the match. But they've more than made up for that 'scare' since, and have been a strong team on buzzer, and fair on the bonuses, with no real standout player.

Bangor also got off to a shaky start, winning a low scoring affair over St Andrews. But they've since shown that they weren't just lucky to get a lucky draw first time around, and have been a strong consistent team ever since. Most of their starters have been courtesy of Simon Tomlinson, which, again, raises questions about if he has an off-day. The team also comprises two blog readers in the form of Adam 'Welshguy' Pearce, and Nina 'pocketmouse' Grant.

So, those are the stats, and the over performance record. What about the draw?

Well, for the first time under the new quarter-final system, the two teams who came straight through without losing a quarter-final, have been drawn together, with U.C.L. playing New College next. James Gratrex of the King's College Cambridge team has suggested this may be due to U.C.L. beating both Bangor and Manchester in the QFs, and the producers not wanting another rematch so soon.

I am inclined to agree, and suspect that, had James and co won last night, they'd have been placed against U.C.L. in the semis, and New College with Manchester.

So, on Monday, U.C.L. play New College, followed by Manchester vs Bangor the week after. And then, the final!

So, who's my tip for the final? To be honest, I cannot call it. It's been a closely fought series, with no real standout candidates for the title. Whoever wins will fully deserve the title.

21 comments:

  1. Out of interest, and because my ego could do with a boost, would it be possible to prepare stats on top number of starters per person? I know I got 25 over 6 matches which is maybe two or three shy of what I was hoping for, and wondered how it compared to the top players so far: Tomlinson, Cappleman, Dennis and Gilbert/ Brown spring to mind.

    ReplyDelete
    Replies
    1. Well, James, I've been keeping tallies over the course of the series. I don't have the full stats handy now, but I will post an overall list of individual starter stats, amongst other things, at the end of the series.

      Delete
    2. I'm being very cheeky in asking! Sorry for the trouble caused - I was thinking of finding out at some free time in June, maybe myself.

      Delete
    3. No offense taken James. I was going to do some kind of 'end of series review' anyway.

      Delete
  2. James is correct - When we arrived on the last day to film, we were explicitly told that we'd been pitted against Manchester specifically to avoid a rematch between UCL and either team. They claimed that they would normally pit the unbeaten teams against each other but that they also endeavour to prevent any rematches wherever possible (hence their being rather rare). New College seemed pissed off by this.

    Obviously, I know what happens. But I think that looking at the known facts, you have to fancy UCL for the title - they've beaten two out of the other three semi-finalists already, and have been more consistently impressive than New College as well as doing so without relying mainly on the contributions of a single team member for starters.

    Whatever happens, Bangor is already the most successful Welsh University on the program (at least in the Paxman era), being the only institution from Wales to make the semi finals (having also done so in 1999 as Paxman pointed out). From what I can tell no other Welsh uni has even made the quarters!

    ReplyDelete
    Replies
    1. More notably, Adam, no Welsh uni has won UC yet! Maybe...

      Delete
    2. If UCL are victorious next week, it would seem logical to just hand them the series trophy! (having beaten every other team...)

      Delete
  3. I've had a while to think about it of course - even so I was expecting UCL v. Bangor and New College v. Manchester, since UCL had played Bangor earlier, and surely they'd want to keep the unbeaten teams apart until the final?

    Not surprised, then, then New College are a bit annoyed at the draw, but then they'd probably have had to beat UCL in the final so why not get it out of the way?

    My own stats table confirms what we knew already, that UCL are far and away the best team so far, with Manchester second followed by New College and Bangor. It will be interesting to see how much New College were let down by tough Quarter-Final question sets or whether they did drop off in the quarters. Either way I'd call a UCL win against Manchester in the final. And hope I am wrong. I promise that I don't know what happens!

    ReplyDelete
    Replies
    1. A purely random though that's not really team-specific/relevant.

      If the at the gong of the final match, there was a tie and thus, a sudden-death situation..what would be the nature of the tie-breaker question? In terms of field of knowledge etc...
      quite exciting to think about!

      Delete
  4. It's the next starter, isn't it (or what would have been)? So it could be about anything that the questions are normally about.

    ReplyDelete
    Replies
    1. Yeah, difficult to see how it could be anything other than the next starter. That may mean favouring one side or the other but is the fairest way in the long run.

      Delete
  5. Hi everyone! I'm going to be appearing on the next series of University Challenge and have been reading this blog with glee imagining what you might have to say about our matches. It's been very entertaining, so thanks for that!

    But I'm posting here because I've come up with a fairly simple formula for assessing individual teams and want to know what you guys think. Here it is:

    sum(CS*BC)/MP

    where CS = combined score, BC = bonus conversion, and MP = matches played. Here's my rationale:

    1. Using the combined score takes into account the quality of the losing team as well as the winning team. Also, I've done some rudimentary statistics and discovered that of the statistics available after round one (winning score, margin, and combined score) only the combined score has a significant relationship with future success.
    2. Bonus conversion is an independent measure of a team's quality, so it makes sense to award them points in proportion to how well they did on bonuses.
    3. I want it to be applicable in all situations, so it needs to be standardised according to number of matches played.

    Now, the real weakness here is that I'm not accounting for speed on the buzzers. This means that teams who get few starters but are very good on bonuses will have an inflated score, as can be seen from my ranking of the semi-finalists:

    1. Manchester - 194
    2. UCL - 187
    3. Bangor - 153
    4. New College - 127

    Obviously UCL should be above Manchester, and possibly New above Bangor, though my gut instinct is that New's quarter final performances are more indicative than their R1/R2 performances.

    Does anyone have any suggestions for how I could integrate buzzer speed into the formula? I was thinking including a term for proportion of starters answered correctly, or something, but I'm not sure how. Otherwise I think it's not bad for something I came up with in ten minutes - apart from the top two slots, which are very close together anyway, I think I more or less agree with that ranking of the semi-finalists.

    ReplyDelete
    Replies
    1. Okay, having said all that, I went off and refined it, then applied it to every team in the competition. Here are the results:

      1. Durham (209)
      2. *UCL (207)
      3. *Manchester (187)
      4. Homerton College (170)
      5. *ICL (169)
      6. *New College (163)
      7. *Pembroke College (162)
      8. *St. George's (161)
      9. Jesus College, Oxford (160)
      10. Lancaster (159)
      11. *Bangor (157)
      12. Lincoln College (153)
      13. Sidney Sussex (152)
      14. York (149)
      =15. Warwick (143)
      =15. Magdalen College (143)
      17. *King's College (139)
      18. Exeter (135)
      19. Bristol (124)
      20. Aberdeen (117)
      21. Queen Mary (111)
      22. Strathclyde (109)
      =23. Jesus College, Cambridge (108)
      =23. Bath (108)
      25. Trinity Laban (100)
      26. St Andrews (86)
      27. Liverpool (85)
      28. Wadham College (81)

      (Yes, I am in fact procrastinating on my dissertation. Hush.)

      Anyway, several caveats: obviously teams with fewer matches to their names are going to have less reliable scores. Furthermore, I've marked the quarter finalists with an asterisk; they're likely to have lower scores on average due to the harder questions, which explains why Durham tops the list. Sidney Sussex (13) is the highest ranking team to go out in round 1. The quarter finalists are bunched very close together, for the most part, which is interesting. (Bangor, your score is being pulled down by the low-scoring first match - obviously you're better than Imperial. But I can't work luck into the formula, so this is the best I can do. :P) I'm not sure what's going on with your score, King's - I suspect it was the fact that your matches tended to be lower scoring on average than the others'. And finally, the highest score achieved for a single match was UCL's victory over Manchester, which saw them gain 232 Highly Arbitrary Units.

      Delete
    2. Oops, I just noticed I made two mistakes, which move Durham down to 2nd (200) and Warwick up to 12 (155).

      Now I swear I'm not going to post again until someone replies. Sorry about the flurry of comments!

      Delete
    3. Intriguing. I think the thing lacking here is a measure of wins. Lancaster, who won once and lost twice, ought to be pushed down the list because of that, ditto Pembroke. My own measure was attempting to compare the team's performances against the average for a winner or loser. Anyway, by most measures UCL turn out to be the standout team of the season.

      Delete
    4. Hi opaltiger, and welcome!

      Thank you very much for those stats of yours, and I very much look forward to seeing you on screen next series!

      Delete
    5. What university do you go to by the way opaltiger?

      Delete
    6. You might be right. I was quite surprised by how highly Pembroke scored, but like New College, that was largely on the strength of their very strong first/second round performances. I think this reflects another difficulty: the quarter finals onwards are definitely harder, and it's difficult to adjust for that. The problem with trying to introduce a measure of wins is, of course, that teams play such a variable number of matches and can lose under such different circumstances.

      I go to Trinity College, Cambridge.

      Delete
  6. I understand why you're weighing combined score, but the problem with that is that teams that get utterly thrashed in the first round get a much higher ranking than teams that nearly won. It balances out over the series as a whole but when there's only one match been played the thrashed teams have a huge advantage over teams that put up a better performance - for this reason Strathclyde (the lowest scoring team of all), Jesus Cambridge and particularly Exeter are much higher ranked than they deserve to be. Exeter (who scored only 85 points) in particular seem over-rated in this analysis, coming higher than Bristol, who actually won a match!

    Conversely, it seems to me that teams like Trinity Laban (who put up a half-decent show against York) should definitely be ranked higher than they are.

    ReplyDelete
    Replies
    1. While you're right in principle, to a certain extent I think I just have to bite the bullet - it's a pretty useless exercise in general, trying to rank teams that have only played one match, I think.

      But based on the stats, the results you point out aren't actually that strange. Strathclyde actually had pretty good bonus conversion (40%) compared to, say, Wadham (25%). Exeter, meanwhile, had a higher bonus conversion than UCL! Being thrashed usually means you're being beaten to the buzzer consistently, which is reflected in the score, but teams can make up for that with a decent bonus conversion. This also explains why Durham manage to do so well: they have the highest single bonus conversion of the series, with 72% in their first round match. It might be that the score underrates teams who are good on the buzzers but not so good on bonuses (Trinity Laban got 40% of the starters, for example, but only 29% of bonuses), but I'm willing to accept that - after all, bonus conversion is the only independent measure of a team's ability.

      Delete
    2. Hmm. I see how it works, but I'm not sure if it's desirale for a ranking system to produce results which feel so intuitively "wrong"... then again, Wales, we're told, are the 49th best Football Team in the world. And the FIFA rankings actually matter!

      There is one stastic though which measures everything: how many points a team scored, how good their opponents were, buzzer speed, bonus conversion, breadth of knowledge - and that's whether they win their games! On that basis I think you have to rank those teams that did win above those who didn't, even if we might not expect Britsol to beat, say, Lincoln College, had they met each other.

      Of course this is all idle speculation. As our own performance proves, a lacklustre first round doesn't mean a team aren't actually a good one, had they been given the opportunity to show it. If UC was a league instead of a playoff the standings would doubtless be very different.

      Delete