unranked is too long for the field; none is just short enough so I went with it (N/A looked weird).
Not sure if I'm going to go with this setup; I can take out the conditional formatting for the ratings but I wanted something akin to WL's setup of "Not yet ranked with a rating of 420"; I'm still using the ratings to determine matchups.
No, there's no adjustment for first pick advantage, largely because I prefer simpler statistical models and in the long run things should even out. It's also a bit tricky to calculate the effects of first pick advantage across so many templates.
@Knyte, wouldn't first pick advantage only apply to games in which the picks overlapped?, as such it may take quite a few games to balance out.
Also, if you don't account for it, due to the fact that certain players will get the advantage more commonly than others, and also new players may have a minor advantage due to the fact that they haven't played long enough to have it even out, that certain players will inherently have an slightly inaccurate ranking?
That being said, wouldn't it be possible to figure out how the current ladder system uses it, and apply to those templates, and then try to apply them to the others, as free time arrives? (Please note: I really haven't tried to understand how the ladders currently work, I'm just assuming they consider first pick advantage)
@japan: Most templates you should end up overlapping some picks even if it's not in the first 3. Also consider cyclic order which a couple templates may have. I honestly don't think there's a bias to first pick. I see myself usually hoping for second pick more often than first unless there's a particularly OP territory on the board that's pickable.
Atm, the ladder accounts for it by a base rating boost/padding. I forget the exact number but it's pretty low.
I'm just suggesting Ideas, I also note that the difference would be extremely small, but may be critical if two players are equally good at a template, as they should have the same rating.
So I agree with Kenny; I also think it varies heavily by template.
First-pick adjustment seems to be something just pulled from chess (where there's a clear first-move advantage) but I haven't seen any good statistical measurements establishing anything of that sort in Warlight (especially when move order is cyclic here). I could start tracking first-pick "advantage" (P(team wins|player gets first pick) - P(team wins|team doesn't get first pick) and convert that into an Elo adjustment (which is pretty trivial, since Elo can be directly converted into probability), but that data would have to be collected from the league itself.
Given that the query game API is absolute shit in terms of how many games it can be run on (nearly 0 in this league as I've had to manually create them to bypass create game API restrictions), that's going to have to be saved for another day.
Anyone know a simple algorithm (ideally one I could express as a formula) I could use to determine the maximum possible number of games given an array of players and the remaining # of games they could play at one time?
The current method is a bit inaccurate (especially for small groups) and it bothers me.
Really, the main edge case is when a league has only one player who's got a limit of 2 or greater. Obviously, the max # of possible games is 0 but the current set up (divide by 2) thinks it's 1.
So long as x is less than n, and n isn't 0, the formula works fine. I think it's a case of handling those smaller values with extra checks.
With 2 players left the max can only be 1 game (2*1)/2 = 1
Also when doing checks I would pair up all those with a limit of 1 game first, then 2, then 3, etc. I think that gives you the best outcome, if you're not already :)
