Win Percentage Seems Cool, but is Actually Pretty Useless

[Itrade]

Win percentage is a nifty feature, but it's rather useless if you think about it for a while.

A player who plays only 1v1 games with a 25% win percentage is losing to a lot more players than someone with the same win percentage who only plays only 8-player games. This is because at a 25% win percentage a exclusive 1v1er loses to three players for every one he beats (losing three games for each one he wins), whereas someone with a 25% percentage who plays 8-player games loses to three players for every seven players he beats (but also losing three games for each one he wins).

So win percentage can't be used to gauge skill since a win percentage like 30% might be abysmal for, say, a triples or quadruples player, but is excellent for an seven or eight player games player.

That would still be okay, though, since you could just look through their game history and see what kind of game they like to play and use that to see whether a percentage is good or not, but there's a problem. If a player likes variety it's very hard to tell whether his percentage is good or bad. A five-player game guy with fifty games played and a win percentage of 10% could make himself look much better by playing fifty 1v1 games or triples games and winning half of them, thus booting his win percentage up to 35%. It would thus cause him to be underestimated in future 1v1 games while being overestimated in future 5-player games.

So what I'm saying is win percentage seems cool, but is actually pretty useless most of the time. A better indicator would be count how many players he's beaten compared to how many times he's lost. In team games, teams would count as one player each.

Um, this should be the end of the post but I don't have a proper conclusion in mind and I have something more to say that should've gone in the middle that I'll just chuck in here. Assuming everyone is of the same skill and there are no deadbeats, you should have an 1/X chance of winning of all your X-player games (Unless they are team games, in which case X is the number of teams). Any number above 1/X means you're good at what you do, any number under means you're not so good at what you do.

[bbqpenguin]

this is why we have that nifty points system...

[AAFitz]

Its not useless, you just have to view the info with the proper context, as you do with any statistic. One of the biggest factors beyond skill and quantity of games, is types of games that a player plays. When new scoreboards are available, we will have a new age of competitiveness in many different categories, that will give a lot of players different ways to find niches in the game, and excel, and keep track of it.

The game is for fun, the score, win stat, and feedback are ultimately all for fun, so precice levels of skill, winning ability, etc. are all for fun. Viewed in this context, they are far from useless. They simply arent a precise measure of anything. Its just an average based on a players total experience in CC, for that player to use as motivation, recognition, or whatever else he or she chooses. It is therefore very useful, if viewed properly.

[Itrade]

That's why I said "pretty useless" instead of "totally useless". I find win percentages to still be a lot of fun, especially since I have a 31 percent win percentage even though I play mostly eight and six-player games.

[Teutonics]

Well, if you don't like the deceptiveness of the win %, then the 'average points won per game' would be a better stat for you. It gives a comparison of how well people are doing that is not affected by the different types of games they play.

For example, if Jack only plays in 2-player games and wins one out of every two games against an equal opponent (same score) then his average point per game would be zero (breaking even), while his win percentage would be 50%. Whereas, if John only plays in 8-player games and wins 1 out of 8 times against equal opponents then he would also have a point average of zero and a 13% win rate.

Both guys are doing equally well in their own games and their average points per game (both zero) reflects their equality. It doesn't mislead people as their 50% to 13% win rates do.

-------

When I look up at the scoreboard I see sjnap (currently 2nd) kicking ass at well over 5 points per game, but Warsteiner taking the #1 spot back through the sheer number of games he plays. His average is 2 points per game.

Then there's the tight race between poo-maker (3916) and Blitzaholic (3901), which is somewhat misleading as well because poo-maker makes an average of 4 points per game, whereas Blitzaholic makes 1 point per game.

[DiM]

Well, if you don't like the deceptiveness of the win %, then the 'average points won per game' would be a better stat for you. It gives a comparison of how well people are doing that is not affected by the different types of games they play.

avg points per game would be almost as useless as the win%.

a major that plays other majors in 1v1 games will average 20 points per win.

a major playing 4p games vs cooks will also average 20 games per win.

[gimil]

A possible winning formula would be point won per player. So in an 8 player game you win 70 points this would average at 10 points taken per player.

[DiM]

A possible winning formula would be point won per player. So in an 8 player game you win 70 points this would average at 10 points taken per player.

this would theoretically give a clear indication of what types of players you encounter. low ranks high ranks etc.

but in practice it won't work:

for example if i play 10 8p games and average 5points per player per win because i play low ranks
but i also play 70 1v1 games vs higher ranks and average 25 points per player per win.
on the total i will get an average of 15 points per player per win. which in fact won't tell me much.


point is no stat like this will be of any use unless games are broken into separate groups. put all 1v1 games in 1 group with the coresponding stats (win% points per player, etc) then a group for triples or doubles or 6-8p games and so on. and in the end you find yourself with 50+ groups.

[Teutonics]

Well, if you don't like the deceptiveness of the win %, then the 'average points won per game' would be a better stat for you. It gives a comparison of how well people are doing that is not affected by the different types of games they play.

avg points per game would be almost as useless as the win%.

a major that plays other majors in 1v1 games will average 20 points per win.

a major playing 4p games vs cooks will also average 20 games per win.

I don't understand your point. If you are saying that it's more difficult for a major to beat 3 cooks in a 4p game than it is for him to beat another major in a 2p game, and should be awarded more points (instead of getting 20 for both wins) then you've got an issue with the current point system. Avg points per game is just a reflection of the current point system.

[owenshooter]

it seems like you are trying to justify your low win percentage...-0

[zell565]

I think they should just get rid of all ranks and percentages.

Please don't look at my win percentage...Please don't look at my win percentage...

[Teutonics]

A possible winning formula would be point won per player. So in an 8 player game you win 70 points this would average at 10 points taken per player.

I think you and I are saying the same thing, but differently.

I believe you made a math error. If you were playing equally ranked opponents, you would win 140 points from those 7 players, or 20 points per player (not 70 & 10). You would also win 20 points per player if you were playing in 2p games or 5p games or whatever.

[DiM]

Well, if you don't like the deceptiveness of the win %, then the 'average points won per game' would be a better stat for you. It gives a comparison of how well people are doing that is not affected by the different types of games they play.

avg points per game would be almost as useless as the win%.

a major that plays other majors in 1v1 games will average 20 points per win.

a major playing 4p games vs cooks will also average 20 games per win.

I don't understand your point. If you are saying that it's more difficult for a major to beat 3 cooks in a 4p game than it is for him to beat another major in a 2p game, and should be awarded more points (instead of getting 20 for both wins) then you've got an issue with the current point system. Avg points per game is just a reflection of the current point system.

no i'm not saying that. i'm just saying avg points per game is as useless as the win% as it won't really give you any info on how good a player is.

[Teutonics]

I disagree. We're all trying to raise our total points (well, most of us..) and points won per game is a clear indication of how rapidly we are achieving that goal. If you have two majors with equal points, but one has played 5 times as many games as the other, there is a difference in their skill level.

And average points won per game makes that distinction clear.

[DiM]

I disagree. We're all trying to raise our total points (well, most of us..) and points won per game is a clear indication of how rapidly we are achieving that goal. If you have two majors with equal points, but one has played 5 times as many games as the other, there is a difference in their skill level.

And average points won per game makes that distinction clear.

nope it doesn't at all.

2 majors
1 plays only triples vs cooks and averages 5 points per game.
1 plays only 1v1 vs cooks and averages 5 points per game.
which is more skilled??

or another example:
1 plays 1v1 vs other majors and gets 20 points per win but since he also loses games his average is 12 points per game
1 plays only 3p crossword freestyle games vs new recruits and gets 20 points per win but since he sometimes loses averages 15 points per game.
who's the skilled guy and who's the crappy one? is the second better because he has a bigger average?? i doubt it because if you analyze his games you'll see he's nothing but an abuser, a specialist that preys on noobs.

[kerntheconkerer]

True, but it still is cool to have a high win %

[MajorRT]

Actually (Points-1000)/games would be more accurate, since this takes into account starting point value of 1000.

[Itrade]

1 plays only 3p crossword freestyle games vs new recruits and gets 20 points per win but since he sometimes loses averages 15 points per game.
who's the skilled guy and who's the crappy one? is the second better because he has a bigger average?? i doubt it because if you analyze his games you'll see he's nothing but an abuser, a specialist that preys on noobs.

Replace 3-player crossword freestyle games vs new recruits with 8-player AoM foggy freestyle games vs anyone who joins and you have me. Now I feel all guilty and stuff.

I do think average points won per game is a better indicator of skill than win percentage; however average points won per game is still less than ideal.

The current system of point distribution at the end of a game works very well since it takes into account the skill of of the players you are playing against and how many of them there are. I have no idea what point I was trying to make with this but I'll leave it in anyway.

[DiM]

Actually (Points-1000)/games would be more accurate, since this takes into account starting point value of 1000.

that wouldn't be any good either.
generally only the last few hundred games matter when it comes to score.

for example i have 2483 points after 624 games. with the formula you presented that would put me at 2.37 points per game.
but the funny thing is that i've been hovering around this number of points for the last 300 or so games. why? because this is probably my level, i have plateaued and can't advance much further.
so 300 games ago i would have had ~5 points per game. does this mean my skill has dropped in the meantime??

probably when i reach 1000 games i'll still be around 2400-2800 points. that would put me at ~1.4-1.8 points per game. an even lower value. will that mean that my skill dropped even more??

or let's say a new recruit wins his first game and gains 20 points. that puts him at 20 points per game.
now let's look at blitzaholic, one of the best players around. with your formula he averages at 0.72 points per game. does that mean the new recruit is 30 times better than blitz??
that would also make me with my 2.37 points per game. 3 times better than blitz. i own you blitz hear me, i own you.

[MajorRT]

Only problem with the current system(or advantage to prem. players) is that the more games played, if you're above average, the higher your score will go. On the top 10 list, actually mhennigan is the most dangerous, at #9....most points per game , even after taking away the initial 1000. WE NEED A PLAYOFF TOURNAMENT!

[Itrade]

That's not a problem, that's a feature.

DiM, I've got a post above yours where I say something or other but you were writing your post when I cranked it out and you might've missed it. I was wondering what your opinions are so if you could kindly scroll back up and address it that would be much appreciated.

[Teutonics]

Actually (Points-1000)/games would be more accurate, since this takes into account starting point value of 1000.

that wouldn't be any good either.
generally only the last few hundred games matter when it comes to score.

for example i have 2483 points after 624 games. with the formula you presented that would put me at 2.37 points per game.
but the funny thing is that i've been hovering around this number of points for the last 300 or so games. why? because this is probably my level, i have plateaued and can't advance much further.
so 300 games ago i would have had ~5 points per game. does this mean my skill has dropped in the meantime??

probably when i reach 1000 games i'll still be around 2400-2800 points. that would put me at ~1.4-1.8 points per game. an even lower value. will that mean that my skill dropped even more??

or let's say a new recruit wins his first game and gains 20 points. that puts him at 20 points per game.
now let's look at blitzaholic, one of the best players around. with your formula he averages at 0.72 points per game. does that mean the new recruit is 30 times better than blitz??
that would also make me with my 2.37 points per game. 3 times better than blitz. i own you blitz hear me, i own you.

MajorRT presented the best formula so far.

And DiM brought up a valid reservation, if we only consider a player's average point win per game, without taking into consideration how many games the player has completed. As DiM stated, a good player that is new to CC will initially have a high average point per game (PPG), which will begin to fall as he plays more games and advances through the ranks.

So, what we need to do is look at the PPG in a way that reflects this consideration. One way that comes to my mind is to break the PPG into tiers based on the number of games played. For example:

Tier A - up to 50 games
Tier B - 51 to 100
Tier C - 101 to 200
and so on...

and the PPG could be represented as something like A2.4

[Itrade]

I personally like amount of players beaten divided by amount of games lost as a way to tell how good a player is. It doesn't take into account the skill level of the other players, though.

[greenoaks]

the system we have now takes into account the skill level of the other players. lets leave it the way it is.

[MajorRT]

I agree the current system takes skill into account. Eventually if you play many games you will plateau, and this will be a true reading of your(CC)skill, much like chess ELO ratings.You'll stay at that level unless you improve strategies,or deteriorate, like playing drunk. BUT until you play lots of games, some formula taking into account points won per game is very important to see eventually where you will end up.Again, on the leader board ,watch out for mhennigan who gets many ppg.