whatever... i'm sitting at 59% right now... usually hover around 63%-57%... my best ever was hovering around 69% for about a solid month of steady games... i play mostly doubles, and always check a teams win percentage. the higher the percentage, the better the game, excluding unfair drops, etc... so, i think it is a fairly good indicator, but only if the person you are playing or the team you are playing, mostly plays the same type of games that you do...-0
p.s.-DiM... *just shaking my head*, you let me down, bro... sigh...
Thorthoth,"Cloaking one's C&A fetish with moral authority and righteous indignation
makes it ever so much more erotically thrilling"
There are a lot of elements of luck in the initial few turns in 1 v 1, in particular the drop and the first move. So while 50% for a 1v1 player may be comparable on paper to a 25% win rate on 4 player (standard), it takes a lot more skill to get from 50% to 60% on 1 v1 that it does to get from 25% to 35% on quads... b/c on 1 v1 there's only one guy out there who can screw up and hand you the game. In Quads, etc., especially esc., your got x N-1 people to hand you the game. All this is just conjecture based on personal observation, of course.
Itrade wrote: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.
sounds great but............I play almost all 8 player games for the last 100 or so and have kept my 40 percent. So I beat tons more players than I loose to. I also loose alot of points when I do. But usually all I have to do is beat one player at the right time and then it is down hill from there. Or make a good block and harvest after, or prey upon a lower ranks mistake. It is a more risky style but I love the wildcards who join in public games and all the different way games go compared to equally ranked players. Anyway, it would be cool to have a number in the profile like players beat/times beat so in 8 player you get 7 points when you win and one when you loose. So it may look like 657/121 or whatever. But many tourneys have that. Anyway it won't be much proof anyway cause I would have more kills but against lover ranks where someone else my score can have a lower kill rate but be better than me cause he/she beats higher ranks. Well, whatever
Owenshooter described how he uses the win %'s to size up his competition in doubles games. I'd like to be able to do the same in the various games I play. I want to know when to stay away from a high ppg guy like mhennigan, who would be much more likely to beat me than the other guys at his rank (even though *those* guys would all make short work of me...)
In essence, when sizing up my competition I want to know if a guy has plateaued at his current rank. As it is now, I make a quick judgment of that based on their ppg, but as DiM stated the ppg reflects the lifetime average as opposed to a player's current average and can be misleading.
So... what about this: Make the ppg a rolling average of a player's last 100 games, which wouldn't kick in until a player has completed 100 games.
Well, for those of you interested, I posted up an explanation of how you really need to model the win percentage in this game... it's not as simple as saying, well, the median player has a 1/n chance of winning...
Very nice! How about some graph of points per game and eventual score....assuming ppg starts at 1000, and is the derivative of the function of your changing score? Maybe then those who have a low amount of games can extrapolate to find their eventual score and compare to the big players, and sigh.....
Obviously you two don't understand the significance of this model... with it you could take the top 1 percential (known roughly by score) and their long run win rate, and back into A REASONABLE APROXIMATION OF HOW MUCH OF THE GAME IS DUE TO CHANCE FOR THE SETTINGS THOSE PLAYERS USE MOST.
And it also shows you that the skill level between two 1v1 players that are within 1-2 percentage points can be quite great, while at the upper tail of the distribution, a small skill edge translates into a huge difference in win rate.
If you don't care about understanding what's going on with the win rates, why read this thread???
Because, my mathematical madman, it' s clear that win percentage is not a great indicator or one's skill, but more a function of what type of games one plays. WHAT is the best way to determine who is or will be the highest? When one plays many games,you plateau, as many players attest to (except if you change strategies). BUT until then, I submit that average points per game ,minus your initial 1000 points IS the best way to gauge potential. This will level off, like a velocity of a long throw of a ball....until when the ball stops, it reaches the furthest point(your max score).
Why not use your graphic acumen to plot some curves for scores start to finish, based on these?
if you go back in the thread you will see that we were advocating ppg over win %. I ignore win % and instead calculate ppg & compare it to number of games played in order to guestimate whether my opponents have plateaued out.
It sounds to me like you didn't read the entire thread & just skipped to the end.
The problem with the ppg stat (and maybe this was said already... I appologize if it was but I did skim the posts in both threads) is that over time you are ppg will drop down as your game total (no limit) vastly exceeds the maximum points available per gaim (theoretically 8x100=800, but I've yet to find that game ever - and obviously leaving out "battle royale"). So yes, every ball will have 0 first movement at some point, but if you want to think of it like gravity... at the point the "ball" has stopped in this context, it has acquired substantial mass of its own.
That said, there are some other considerations: because of the point weighting system, you have ways to artifically inflate the ppg yield, like playing noobs instead of cadets. If you don't think this happens, go back and check your first 5 games... there's a good chance you had a few games against majors or colonels, but from private/cadet all the way up to seargeant you probably had 1 in 10 games agianst a decent ranking player.
Also, you have team composition to content with, so figuring ppg without taking into account the teammate's relative score is about as useful as looking at win percentage without checking into the type of games the person is playing.
Lastly, you have breakpoints due to the way the cc community is organized. For example, due to the inability to get good opponents, I probably averaged around <10 ppg playing 1v1 from the level of private to seargeant b/c even winning 60%, there was always often a net loss of points. Once I could get into seargenat's hall games, that number shot up substantially b/c I was playing opponents with approixately the same score.... there was no "change in strategy" as you put it.
Sorry, your argument didn't convince me because if you beat a weaker opponent, you get less points ; for instance if you defeat the last place(?by design) player, who has 1 point, you will get <1 point. The system takes this already into account ; I agree that once you play more games and gradually go to your true point level, it matters less. Actually once you plateau, your ppg total (int he most recent games) is zero! As you mentioned in your first thread, it's calculus! The ppg is either a derivative or a second derivative, of your score function....which in the end will flatten out somewhere.
I have played over 2000 games, and still have a continuing trend upwards, in points. Plateau-ing is not the only eventual possibility. There is always room to improve. If your score has plateaued, in your mind, then you just need to learn more and get better.
Yes, if a once-plateaued player learns new strategies, becomes able to play higher ranked opponents, chooses different partners, etc., then he can once again start climbing the ranks. Without a rolling average of the player's ppg from last 100 games or so, it would be difficult to distinguish this player from another that has plateaued out at roughly the same rank and number of games.
Yet, for the purposes of making rough guesstimates of whether a player's current rank reflects his likelihood to win some random game, I feel quite safe in assuming that the guy with a high ppg/ low game count will be just as deadly, if not more so, than someone with a slightly higher rank, but much lower ppg.
There should be some way that we could plot at least a rough chart utilizing the ppg to help distinguish the still climbing players from those that have leveled off since a rolling average (the best method) is unavailable.
While I think ppg would be very useful in distinguishing the difference between two players ranked around 1000-2000 (that being one who's only played a few games being on his way up the ladder vs one who's played 1000 establishing himself as exactly that good and no better), I'm not sure how useful it would be at either the top or bottom of the ranks. If a player has below 1000, chances are it's for a reason. Either they haven't played many games and haven't figured out how to win or they've played a bunch of games and haven't figured out how to win. While there may be more hope for the former, that likely doesn't have much effect on what kind of game they're going to give you right now.
Same goes for players above 2000. Comparing ppg is a double edged sword. Certainly those who've made it to page one without playing a ton of games can take pride in that, but there's something to be said for being able to hold your rank. To be able to maintain that score over time. That said, it's no harder for a player who was 1800 after 1000 games to go on a bit of a winning streak and make 2100 than it is for a player who's only played 50.
Perhaps a more interesting and useful stat would be average rank as calculated by number of days or games that rank has been held.
I agree that win % is a silly stat (considering how it depends on the type of game you prefer). But I think PPG could be a silly as well for reasons mentioned above. I'd be up for improving win % to weigh in the number of opponents.
Itrade wrote: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.
Not bad, but how about something simpler: instead of wins divided by games we use wins divided by # of opponents?
lackattack wrote:I agree that win % is a silly stat (considering how it depends on the type of game you prefer). But I think PPG could be a silly as well for reasons mentioned above. I'd be up for improving win % to weigh in the number of opponents.
Itrade wrote: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.
Not bad, but how about something simpler: instead of wins divided by games we use wins divided by # of opponents?
lackattack wrote:I agree that win % is a silly stat (considering how it depends on the type of game you prefer). But I think PPG could be a silly as well for reasons mentioned above. I'd be up for improving win % to weigh in the number of opponents.
What about a rolling average PPG of last 100 games?
