Log in

No account? Create an account
24 October 2010 @ 06:13 am
WPC Poland Thread  
Been in Poland now for about a day (having arrived early this year to get more adjusted sleep-wise and hopefully better able to perform at my best - last time I did this was before Brazil, certainly my best WPC performance to date). I'm unfortunately running into a weird internet connectivity issue with my laptop (I get very few packets per second, ~10-20% ping efficiency) but Wei-Hwa's laptop works fine so it might be something specific to a mac os computer on this local network. Anyway, the short of it is I probably won't do liveblogging this year if I don't sort this out.

Competition has a lot of good classic variation rounds and some oddball rounds. Bonuses are also a little curious in some respects. In several rounds the bonus increases as more people finish (if n people finish, the first solver gets 5*n points, next gets 5*(n-1) points, ..., but this causes a few problems). If a solver does very well on a round that is otherwise hard to finish, they earn 5 (maybe 10) bonus out of 200 or so for the round which is tiny, when there also is no time bonus. On the other hand, if the round is very poorly timed and dozens finish, then making one mistake can have the effective cost of many points (on the order of the round value). Fortunately only a few rounds use this quirky system but I cannot see it's advantage over several other choices. And I can certainly see the odd position of say having 3 members of team Germany or Japan done when I'd say finished, giving those teams 10 more points relative to the US just because I claim my 5. So the incentives feel a bit off. But that's probably focusing too much on one specific detail than the overall set-up which is very nice.

Individual semi-finals will take the top 8 competitors, staggered start from scores, with 1st place choosing 7 types from a pool of puzzles for the group. Four fastest finishers then become the finalists with the fastest finisher choosing the 7 types from a pool of puzzles for the finalists. Puzzle selection power is huge, so I'm certainly aiming to qualify at the top (not that I wouldn't otherwise).

There will also be a team final with the top four teams. It will be sort of weakest link style but with a relay element. Each individual has a task but they must be done in a particular order. After the first finishes, he/she gets 1/9 of a team task and can start on it while the next member of the relay starts his individual task. The successive members bring 1/9, 2/9, and 5/9 of the team puzzle back. Once the full puzzle is there and solved, team podium order will be determined. This will be interesting (and I certainly feel I can predict 3 of the team finalists right now, although maybe not their order).

Anyway, we've already scouted out the hotel for the scavenger hunt round and I've finished solving all the practice problems I planned for the year, so I'm not sure what I'll do with myself before the competition starts in 2 days. I'm also a part of the evening entertainment - a Puzzle Academy with Vladimir Portugalov and myself. He'll talk about a memorable puzzle he wrote. I'll talk about a memorable puzzle I solved. I doubt the puzzles being discussed will be the same but hopefully it will be interesting.
(Anonymous) on October 25th, 2010 06:11 am (UTC)
Go Dr.Sudoku Go!
Go Dr.Sudoku Go! Win it! =)

ze top blurberryztbb on October 26th, 2010 06:10 am (UTC)
I still advocate a bonus system where any competitor can declare him- or herself done with a round at any time, and claim a bonus that is a function of the amount of time left in the round (i.e., the amount of time they are declining to use). Of course it has to be calibrated so that especially at the top end of the competitors, the bonus is smaller than what you ought to be able to get by continuing to solve puzzles, if you have puzzles left to solve and a reasonable amount of time left in which to solve them. None of this stuff where if you have a mistake you lose your bonus. A nice feature is that if someone finishes a puzzle with a couple of minutes left in a round and doesn't think they can solve another puzzle in the time remaining, they can claim a bonus point or two, and get a small advantage on someone who finishes the same number of puzzles but has to go right down to the wire to do so.
(Anonymous) on October 26th, 2010 09:00 am (UTC)
"Especially on the top end"? I think you forget the low end. Even on a WPC the range is very high, I would estimate a ratio like 3:1 or 4:1 in terms of "earned points per minute" between the best and the worst. You have to prevent the worst puzzlers from declaring "finished" right from the start. But that would mean, that for the best puzzlers the boni are only very marginally.
ze top blurberry: driftingztbb on October 26th, 2010 03:41 pm (UTC)
This is not hard to do if you want to, e.g. you can insist that the bonus earned on a round cannot exceed the number of points earned on that round, so that you have to do *something* before the bonus kicks in. (Probably not a bad idea even with the top end in mind, so that if there's a "1 tough puzzle" round, you can't gamble that you're better off giving up rather than trying to solve it.)
(Anonymous) on October 27th, 2010 10:22 am (UTC)
It should always be more rewarding trying to solve another puzzle, for all solvers regardless of personal skill level. I just can't see this happen with this kind of system.
MellowMelonMellowMelon [wordpress.com] on October 27th, 2010 10:17 pm (UTC)
I've always believed that competitions for something should be more about skill in whatever is being tested than learning how to optimize the system. A change like this would go far in the wrong direction.

In the decathlon Thomas Snyder ran, one of the top finishers pointed out that he would have scored some more points had he not tried to solve a very tricky puzzle and instead declared himself done immediately, due to a time bonus being awarded for a mostly correct set of answers. As said by him, it should never be advantageous to not solve a puzzle. Mathematically speaking, the only way to get rid of this issue is to cap the time bonus at the smallest value of a puzzle you didn't solve, but this would make mistakes just as ruinous if a large time bonus is on the line.

If your goal is to reduce the penalty from dumb mistakes, keep in mind to go for a time bonus is to say you don't need those last N minutes to even check your work. I see little wrong with making such a bold statement come with an appropriate risk. I say this as someone whose stupid mistake record is ridiculous (in a bad sense), so if I'm biased it's in the other direction.
ze top blurberry: driftingztbb on October 27th, 2010 11:08 pm (UTC)
Hi Palmer. I start with basically the same premise as you, but come to a different conclusion. Competitions should be designed so that the most skilled competitors tend to win, i.e., you want to reduce variance. Everyone makes dumb mistakes from time to time, and obviously they ought to be penalized whenever they occur, but _massive_ penalties for a dumb mistake never made sense to me. On some round, competitor A and B may both have roughly the same small probability of making a dumb mistake. If lightning happens to strike competitor A but not competitor B, then B should win the round, but should the win really be decisive? If mistakes were _common_ at the top end, then there would be an argument for it, but for guys at the top end I would tend to model dumb mistakes as random low-probability events whose influence on selecting the champion should exist but not be too large.

Of course the suggestion is not a panacea -- you have to think about the type of round you're putting together before you think about the bonus structure you want (if any). I'm thinking about the sort of round where there are, say, 20 puzzles each worth ~15 points, to be done in an hour. At the top end you're budgeting 3 minutes per puzzle, tops. If there's a bonus of 1 point per unused minute -- or even 2 points -- no one near the top is going to think about quitting on a puzzle in order to claim the bonus, unless there's just a minute or two left.

I don't think one should worry about the low end too much. In the middle of the pack (which included me, back in my WPC days) you probably still wouldn't judge it advantageous to quit too early on such a round. At the lower end, when you're looking at folks who are solving puzzles at half or 1/3 the rate of you and Thomas, does it matter if they game the system? It's just less fun for them, when the standings at that level don't really matter much. [But if you want to, it's easy to build in ways to prevent gaming the system too much at the low end, as I mentioned in an earlier post here.]
(Anonymous) on October 28th, 2010 05:19 am (UTC)
I agree with Palmer - with the type of bonus structure that that's proposed above, it becomes important for the competitor to carefully consider their bonus strategy. Some of the scoring becomes a function of your skill in correctly adapting to the rules of the bonus structure, rather than the skill of actually solving the puzzles. Yes, the most skilled competitors should end up at the top - but our goal should be for that to be the most skilled puzzlers, rather than the most skilled game theorists.

Any rule which potentially results in a player saying "I will lose points by continuing to solve puzzles" is a mistake, in my opinion. The 2010 decathlon had a fun bonus structure which encouraged people to try puzzles they might not have otherwise done (which I think is great) - we observe that this structure caused a minor dilemma for a top solver - in the context I don't think that was a problem, but I don't think that would be appropriate for the WPC.

I actually do think that a dumb mistake should cost you the bonus - a good solver can check their answers, the same as everyone else - it's part of the skill of the contest, and part of the skill of solving puzzles, and it's good that everyone shares the experience of being hurt by a dumb mistake. The bonuses here in Poland seem nicely weighted (tho they are irrelevant to me so I haven't given it much thought) - for individual rounds they are not so generous as to overwhelm the final results, but provide a few extra points here and there for confidently proving you are a faster solver.

As a lower end solver - does it matter if we game the system? Yes! It matters to US very strongly!! I've had several discussions along these lines with other solvers this week.
We don't want to be asked to make a choice between gaming the system or losing points because we enjoy solving the puzzles (see also: the stickers on the Sym-a-Pix round this year). You're precisely right to say this would be less fun for us at the lower end of the rankings - that would be a probable feature of the bonus system you propose.

Thanks for the discussion by the way, I'm fascinated!