Outwitters WORLD CUP Winner - Alvendor (SWEDEN)

[awpertunity]

(09-27-2012 06:54 PM)Samura| Wrote:  

(08-30-2012 10:11 AM)awpertunity Wrote:  The PF awarded will be as follows:
SMALL MAPS (1 wit space each)
+4 points for 12 turns or less
+3 points for 28 turns or less
+2 points for 48 turns or less
+1 point for 49 turns or more

LARGE MAPS (2 wit spaces each)
+4 points for 14 turns or less
+3 points for 32 turns or less
+2 points for 56 turns or less
+1 point for 57 turns or more

Have some suggestions regarding the allocation of points based on turns, for future reference. My initial thought was that smaller maps don't necessarily mean quicker games, as personally I think larger maps offer much more varied attacking options as it's more difficult to fully defend attacks from different sides. Plus it's easier to catch your opponent by surprise. I feel that some small maps like Sharkfood Island seem to lead to quicker games, but Peekaboo is a map which allows more defensive play.

Now that we're almost done with the group stages and therefore have some data, I've done some analysis based on the games that have been completed (about 15-16 games for each map):

Turns to win:

Map : Mean / Median

Small maps: 25.4 / 23.0
Large maps: 25.7 / 21.5

Glitch: 23.8 /18.0
Peekaboo: 30.5 / 30.0
Sharkfood: 21.8 / 20.0

Foundry: 22.7 / 17.0
Sweetie Plains: 33.1 / 28.0
Long Nine: 21.3 / 17.0

Perhaps the sample size isn't very large, but the findings would seem to confirm my suspicions and make logical sense. Namely that there's no material difference in the turns to win between small and large maps. Rather it would appear that most maps are equal, EXCEPT for Sweetie Plains and Peekaboo where games are noticeably longer, therefore if anything we should allocate points on the basis of the 4 other maps vs Sweetie Plains + Peekaboo. We could of course do further analysis about standard deviations for turns to win etc, to determine how to allocate points say for each fixed increment of standard deviation vs the mean.

Not sure if anyone has been affected by this in the current tournament, but these are little improvements we can make for the next tournament.

Yes this is definitely a valid point. Ideally we would have a different scale for every single map based on a lot of data. I definitely thought about it while I was posting that (namely I assumed that Long 9 would have relatively short games even though it is a "large" map, and as your results show my suspicions were right as well), but ended up just keeping it simple. It is definitely late to retroactively do anything about it now but we can definitely keep this line of thought in mind for future tournaments.