loutil wrote:betiko wrote:iAmCaffeine wrote:betiko wrote:iAmCaffeine wrote:I find Kaskavel's description easier to understand than betiko's..
actually my idea might not be exactly the same as kaskavel's... I'd like the win rate we currently have to be replaced by a formula that readjusts to 50% all game types that have more than one opponent based on each of their expected win rate. It's kind of hard to word is simply...
I'm honestly finding it all pretty confusing. Just as soon as I think I've grasped the concept I try and work it out in a practical manner and get stuck. If, like Kaskavel's suggestion, winning a 5 player game counted towards your percentage as though you'd won the same as 5 x 1v1s, it won't correspond with the amount of games won/played. Is that an issue? Maybe not.
well his concept basically would bring in too much weight in those multiplayer games on the game total, so I think mine would give a more accurate result. winning 5 1v1 games should still have more weight towards your win% than winning 1 5 player game. What I'd like to see, is that 5 player game win % readjusted to your expected win rate. Supposing each win/loss gives 20pts, you need 1/5 to break even on those 5 player games. so you need to win 20% of them. In a 1v1 you'd need to win 50%.
That means that you need 2,5 times less wins to break even on 5 player games, therefore your win rate in 5 player games should be multiplied by 2,5 to make sense if you aggregate everything the way it's currently done.
This intuitively makes more sense but I am struggling with how you would apply it using that formula. If I play 10 1v1 games and win 6 my win % is 60%. If I then play an 8 player standard game and win, what is my overall win % ?. Is my multiplier 4? Would that mean I now have 10 wins in 11 games for a win % of 91%? Or, are you saying I would have 10 wins in 14 games or 71% ? With the proposed formula from Kaskaval would my wins now be 17 - 4 or 81%?
I am not certain I understood your example but I will take it the way I understood.
so you would have played a total of 11 games and would have won 7, so with the actual count that would make 64% on your wall
with the count I'm suggesting:
6/10 games won in 1v1 with no mutiplier
1/1 game won in 8 player games. your expected win rate there is 12,5%, normally a x4 multiplier should apply, but given that you are already at 100% your 8 player games should remain with a 100% win rate.
I guess the multiplier should apply only up till 100% (let's face it, this example is based on way too little games)
with my suggestion, if you have played 7 games, well the win rate will always be calcuated on 7 games; multi single games don't hold more weight. The win rate would remain 64% in your example.
Let's take an example with a larger amount of games to make more sense:
player A has played 100 games and has won 35:
-50 games in 1v1/team/poly and has won 20 (40% win rate which is bellow the expected 50% win rate)
-25 games involving 2 opponents and has won 10 (40% "normal" win rate, which is above the 33% expected win rate in 3 player games; a 1,5 multiplier would apply to that win rate; this would be equivalent to 60% win rate)
-25 games involving 3 opponents and has won 5 (20% "normal" win rate, which is below the 25% expected win rate in 4 player games; a 2 multiplier would apply to this win rate, this would be equivalent to a 40% win rate)
player B has played 100 games and has won 40:
-100 games in 1v1/team/poly and has won 40 (40% win rate which is bellow the expected 50% win rate)
Player A for conquer club has a 35% win rate; which is below player B's 40%
With my calculation player A would have the following aggregated win rate: (40%*(50/100)+60%*(25/100)+40%*(25/100)=45%
so player A's win rate would become higher than player B; and assuming they've only faced players of the same level (giving them or taking them 20 points each) player A would have a higher score anyways. he would indeed have a better game record...
Posted: Tue May 12, 2015 1:33 pm
