WPC 2013: Day 2 Morning

See the first WPC 2013 post for an explanation of the post date. Also, all round scores can be referred to here.

The day 2 morning is done, and the scores for one of its rounds as well as the day 1 rounds are available. That led to lots of good news. First, what I considered an okay performance on rounds 4 and 5 turned out to be the high scores, and my round 7 score was only behind Thomas’s and Ulrich’s (comparing with both of them after the round probably skewed my perception a bit…). Second, the weighting system of the world rounds may have turned out to be a bit broken in a way that favored me. I figured with the system they used that top scores would be around 600. Instead, I have two scores over 800 on rounds where the scoring rules specify that only four people can score above 500. That seems a bit too lopsided. The around the world concept seems to work great in general, but I think the scoring rules could use another look. Third, my day 2 morning went so well that I have probably clinched first place in the preliminaries even if I skip the afternoon entirely.

Day 2 began with Black and White matrix. This was an exact copy of round 4 in this online test released by the Hungarian authors: http://logicmastersindia.com/lmitests/?test=M201309P2. Again, we get 60 minutes. The only question was whether this will be much harder, to prevent lots of people finishing, or a better and cleaner puzzle that will give lots of high scores. It turned out to be the latter. I rocketed through it, basically never getting stuck anywhere, and finished at maybe 15-16 minutes out of 60. I then spent about 3 minutes triple checking everything.

Despite this, when I turned in after 18:5x, I was still nervous as hell about errors. A clean round would mean a whopping 1010. A mistake knocking me into partial bonus would shave off over a hundred points, and if I really botched anything and lost all time bonus I’d be down to 550 or less. I was a bundle of anxiety after that round to the point of annoying Thomas a bit, which was entirely my fault. But in the end, I did get my 1010, which topped the round. The US team did really well in general, with Thomas and Will close behind my own time.

Next was the Assorted round, a long 90 minute round with 10 types and three instances of each. This round was incredible; the Hungarian authors are excellent. Each type had the first puzzle 2-themed, the next puzzle 3-themed, and the third puzzle 4-themed. By “themed”, I mean every puzzle had some kind of numeric clue and they only ever used 2s, 3s, and 4s. No cheater clues anywhere. Quite remarkable, the only flaw being that in order to get this perfect setup they had to sacrifice logical solutions in several of the puzzles.

The round performance was less hot. I was going really nicely for about 75 minutes until I had just 3 of the 30 left to do, with 15 minutes left to clear them out. Finishing seemed within reach. Unfortunately, I bounced around these three and solved none of them, whereas others were getting much closer to finishing (total 900) at the end of time. The tweak-heavy nature of the round didn’t help, since I wasn’t even sure what the best approach to the puzzles I had left were. So my 735 (assuming no errors) isn’t going to be quite as good as my other round performances.

Next was the Diagonal Dissection round. This is a bunch of dissection puzzles: just divide a polyomino into the given number of congruent polyominoes. Except here, polyominoes are considered connected by diagonal adjacency, so the search space increases tremendously. This round was a huge worry for me, and I constantly pestered my team to make practice material since you can’t get much out of stuff you made yourself. My pestering succeeded (probably since I provided so many other puzzles to them) and the puzzles they made helped me out a lot. Thanks to Will, Wei-Hwa, and Jonathan. Even so, I was feeling like the round was a lot of guessing and luck. So I completely surprised myself by finishing 6 minutes early (out of 30). I don’t know if that was the fastest or not, but the scores I hear reported so far show not many people were near finishing, so this looks like another outstanding round for me.

The final day 2 individual rounds are a visual puzzles round and a somewhat short zodiac round with 12 likely animal-shaped puzzles of various kinds. The visual puzzles round was always going to be something I’d lose a bit of ground on, and the zodiac round was probably going to go as well as any assorted round. But since both are low-valued (300 and 450 max respectively), there’s not much ground to gain, so I believe a first place finish for the preliminary rounds is secure, and likely a convincing one. Since I’ve already bagged a lucky playoff win in 2011, I won’t be too upset with myself if I screw up the third day, so I feel like I’ve already accomplished everything I wanted to do this year, despite the championship being barely half over. I’m really not sure where this level of performance came from, but I can’t exactly complain…

After those 2 individual rounds are the two day 2 team rounds. I believe the team has a solid lead now, but we’ll make sure to defend it this afternoon.


2 Responses to “WPC 2013: Day 2 Morning”

  1. Tom Collyer Says:

    I’d be interested to hear more thoughts from you on the around the world scoring. The trouble with people having to skip rounds because in terms of scoring you are comparing apples with oranges.

    The wrong thing to do, of course, is to give the round winner the same score, and I think benchmarking things at 5th did provide the stability that the round really needed to try and make fair scoring comparisons.

    That things spiked off at the top end was something that Zoltan and I had noted was possible during our discussions prior to the event, but it is impossible to precisely gauge what to make of this without seeing the raw data. The fact that there were spikes certainly suggest to me that you were a long way ahead of the field for those 800’s, and it is only fair that this should be reflected in your scores. It’s worth noting that were also some very high scores in other rounds too.

    Perhaps this is just an artefact of using a rather basic linear normalisations, and I’d like to go back and look at other models with the raw data (e.g. the logarithmic one used on croco and the CTC) and see what differences there are.

    • MellowMelon Says:

      I think I’d have to see the raw scores too before I could say much more. If those 800s do in fact result from the margin between 1st and 5th being larger than the margin between 5th and the median, maybe there’s some justification. One variation I know I’d like to see is what happens if you pin to 5th only instead of 5th and the median, though I confess I have no idea which way things will swing if you do that.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )


Connecting to %s

%d bloggers like this: