General Notes: Data is for the periods of March 2015 to August 2015 Data contains roughly 92000 games Data contains ~1800 winners and ~2300 losers Games are ranked only Games less than 5 minutes and longer than 4 hours have been excluded from the dataset Map add more charts as I explore the data Games on Underdepths are included in totals, but I am not displaying it due to very few games played on that map Interesting Stuff: No faction benefits significantly more than any other faction in terms of Player One advantage overall However, there is a variety of differences in Player One WR per faction on a per map basis (which may be a sample size issue in some cases) When games feature those whose ratings are close (0-99 difference), the Player One advantage is more pronounced Even on K'thir Forest, where there is an experiment favoring of Player Two, Player One still wins more than 50% of the time Maps that tend to favor Player One less are generally disliked for being boring or have other features that players dislike (Tundra Lake, Brood Pits) Tundra Lake is the only map that has before and after data so far, but shows some promise in the change Other changed maps: Axillium, Brood Pits, Elemental Plateau, Ironfist Pass, K'thir Forest, Ruins of Valdac, Sundered Lands, Snarling Tangle
Super useful data. Thanks Sok. How many games in the sample? Suspicion confirmed that IS Mountains is a hellish map for FW.
Most of them are way longer than 4 hours, which leads me to believe they are bad data from games that didn't end properly.
There was Underdepths, which had less than 30 games, so I removed it from the above, but everything else all has at least 954 games (Forglar Swamp):
The data structure makes it hard to split games into factions (since each game has 2 factions), but every faction has at least 50-60 games on each map even for K'thir Hill (with Splits taking up the rest).
Hrm. That does seem kind of small particularly when you combine it with all the various faction-faction combinations. It makes me wonder if we really have large enough sample sizes to make significant conclusions about them, but I guess we have to work with the data we have rather than the data we would like to have. I guess that makes drilling even further down into seeing if a particular rune is a significant predictor for winning or not winning is probably not even possible.
It makes sense that first turn would factor more in games between players close in rank. The advantage doesn't seem to be that large, and it's probably skewed by rage quits on top of that.
Well... Thank you. Not disappointed, I can see why SL cries about ISisOP. But the difference in % is pretty close. Certainly not disappointed.
Why not slightly lessen starting nora for the 1st player. That way they can deploy 1st turn but unless the champ is dirt cheap, cannot deploy the second turn, giving a little edge to the second player to get his 2nd champ out first. But that said, the % of 1st turn winners is barely over 50 so it isn't that op tbh. Balancing games like this one is hard to do perfectly, but those numbers are pretty dang close to an even 50
They are pretty well in-line with the rest, but have small samples per map, so it's fairly meaningless. When I do the report for OVERALL win rates for factions/splits (rather than Player One WRs), I will do more with the splits as it'll be more relevant there.
There's definitely things to consider changing. But I didn't want to make any such changes without having the data. I believe my next steps will be to continue to tweak some of the more problematic maps, and then after we have stabilized the maps, then see what global changes could be made.