|
Note: I am NOT saying that how much one plays is irrelevant to one's improvement. I AM pointing out that a large number of players play a lot and do not improve, or do not improve quickly, and thus there are very significant other factors involved that are outside the scope of this article. That's all I'm saying, and please do not read more into my comments than that.
I have removed some analysis in the article with which a number of people have disagreed. My point in this analysis was to call attention to the fact that many players in lower leagues seem to have been playing large numbers of games since release and not getting very far, but people were reading it incorrectly to suggest that I am completely discounting the value of practice, so I've removed it, and am happy to let the numbers speak for themselves.
I think this says it best:
On December 09 2011 13:29 teamsolid wrote:Show nested quote +On December 09 2011 12:53 Lysenko wrote: That I might agree with, in that playing a lot of games could be necessary but not sufficient for skill improvement. Yea, I think this is the take away message.
I personally tend to play a lot, and am in gold league. I have the Kerrigan portrait for 1000 wins as Zerg, and I often hear comments from other players like "You must be really good if you've played that many games!" or conversely "You must be awful to be in gold league after playing that many games!" I was curious to see where I fell in the population in terms of activity.
Also, in the course of ladder-related discussions on TL, I found myself speculating about the possibility that movement in the MMR system might be dominated by the play of a tiny minority of extremely active players, while the majority of players remain stagnant from season to season because they don't play.
Since it can take a large number of games (as many as 30 or 40) to get promoted, due to the moving average smoothing used in league promotions, players who placed in, say, Diamond in an early season might log on, play a small number of games (perhaps one placement), and then stop playing, thus overpopulating higher leagues with players who really don't belong there but aren't playing enough to move to the correct place in the rankings.
So, I collected some information from sc2ranks.com on the distribution of activity by league and percentile. (Percentages below are "top x%," not "xth percentile," so 1% corresponds to 99th percentile.) While wins are not a perfect proxy for total games played, the vast majority of players are pretty close to 50/50 win percentages, even in master league, and it's the only data we have for lower leagues.
I didn't include grand master players because I felt that the activity requirements for staying in grand master league would skew the data in such a way as to make it uninformative.
This data includes North America only.
Number of wins for the most active players, by league
0.5% 1% 2% 10% 20% 30% 40% 50%
Master 364 303 232 145 95 66 45 29 Diamond 277 232 190 100 62 39 24 14 Platinum 214 176 142 69 39 22 13 7 Gold 195 157 122 56 30 17 9 5 Silver 181 146 114 51 28 16 9 5 Bronze 286 158 112 50 28 18 11 7
A few observations:
* The most active Master league players play no more than twice as much as the most active lower-league players, on average, and considering that that's the only league with a significant percentage of players over 50% win ratio, that may be an overestimate for their activity. The median Master league player plays about four to six times as much as the median Bronze through Platinum player, but the median Diamond player only plays twice as much, and in lower leagues there's a small difference.
* The most active 0.5% of players play 10 times as much as the median player in master league, and 30 to 40 times as much as the median player in platinum through bronze leagues.
* The most active 10% of players play 7-10 times as much as the median player in every league. This probably means that you're more likely to meet someone in the top 10% of activity in any given game than someone in the bottom 90%.
* Only the most active 30% of players have enough games to approach the roughly 30 game threshold that is typical for promotion or demotion according to Blizzard's comments on how that evaluation happens. (The 30 game period is probably related to the length of the MMR moving average that Excalibur_Z talks about in his league and ladder post.)
* Because the bottom 70% of players are probably not playing enough to significantly move their MMR in the season, it's quite likely that it's not possible to generalize about a given league's play quality from a sampling of actual games on the ladder. Or, to put it another way, looking at a particular player's skill level isn't likely to be predictive of the quality of the bulk of ladder games being played in that league. (In fact, it's quite possible that the vast majority of players in each league are significantly worse than the players who are actually playing the games in those leagues, and defining the league boundaries with their MMRs.)
* The top percentiles of activity in bronze league are far more active than higher leagues, most likely because of the win trading going on for portraits.
Follow-up
It would be interesting to sort some recent replays by the activity of each opponent and see if a given player wins more vs. the less-active opponents. This would provide a way to measure whether less-active players' stale MMRs are a significant problem in the ladder right now.
Edit: One additional factoid
These are the percentages of players by league who have 0 or 1 wins, which would include everyone who played just their one placement match for the season and then stopped:
Master: 15.1% Diamond: 18.9% Platinum: 23.2% Gold: 25.1% Silver: 26.6% Bronze: 26.2%
Edit 2: How was the data collected?
On December 09 2011 07:18 Lysenko wrote:Show nested quote +On December 09 2011 07:15 snively wrote:interesting read how did you get your data? you said you got it from sc2ranks.com but im just curious to know where on the site. The data is North America only. You can sort all players in a league by wins. I started with the total number of listed players, calculated the ranking for each percentile, located that ranking in the list, and read off the win number. Note that data on lower league players are collected less often by sc2ranks, which might tend to underestimate how much lower league players play.
Edit 3: A few observations on the impact of mass gaming on improvement
A few people in the thread have taken issue with my assertion that playing more doesn't by itself lead to improvement because Diamond and Master players play more than the other four leagues. To this I have a couple of specific answers:
* Bronze through Platinum have very similar distributions and numbers of players. If playing a lot of games directly led to improvement, Platinum players should play a lot more than lower league players, but the difference across Bronze, Silver, Gold, and Platinum is more like 20% at the most active end and nothing at all at the low end (not counting the couple of data points in Bronze that probably are affected by portrait farmers.)
* Diamond and Master players do play more. However, large numbers of the lower league players play just as much as the Diamond and Master league players, yet don't wind up there. The top quarter of Bronze through Platinum players play just as much as the top half of Master players in activity, yet they're not Master league. Most of these players have been playing since close to release, so they've had plenty of time to get there. Furthermore, those top quarter of Bronze through Platinum outnumber the top half of Master league by about 10:1.
* To the extent that Diamond and Master league players do play more, that's as likely to be a result of being engaged in the game and wanting to play more as it is that their success is a result of the extra play. That said, if extra play does benefit Diamond and Master league players, the effect applies disproportionately to those two leagues.
* I'm not saying that playing more doesn't help, but it's not the dominant mechanism for people getting better, otherwise the leagues would tend to be sorted by activity far more than they are.
|
I'd actually like to know how much bronze/silver actually play due to the portrait farmer inflation and I don't think i this is true to the extent you think it is a lot diamonds play loads of customs thus maintain their skill
|
interesting read
how did you get your data? you said you got it from sc2ranks.com but im just curious to know where on the site.
|
I think it's well understood in the TL community that while playing a lot of games can be one component of improving one's play
This is well-understood? I thought doing anything in excess or without moderation doesn't necessarily make you better but can deter your play.
* Master league players play no more than twice as much as lower-league players, on average, and considering that that's the only league with a significant percentage of players over 50% win ratio, that may be an overestimate for their activity.
Casual players are typically lower-ranked than Master leagues. So while the general population ratio of play is more, the amount of people who hardly play can have a drastic effect.
* The top percentiles in bronze league are far more active than higher leagues, most likely because of the win trading going on for portraits.
Win traders are below, not the top percentiles I believe.
This confirms the opinion often expressed by the best players that mass gaming doesn't directly lead to improvement
Mass gaming is relative to the person and thus, so is how to improve and how many games one should play if they want to seek improvement.
So...
|
On December 09 2011 07:15 snively wrote:interesting read how did you get your data? you said you got it from sc2ranks.com but im just curious to know where on the site.
The data is North America only. You can sort all players in a league by wins. I started with the total number of listed players, calculated the ranking for each percentile, located that ranking in the list, and read off the win number.
Note that data on lower league players are collected less often by sc2ranks, which might tend to underestimate how much lower league players play.
|
100 or more is a lot of games? i played 1000 this season :x
|
Russian Federation269 Posts
|
I'm kind of curious about why bronze seems to break the trend of games played going lower and lower as the leagues get lower. I have a suspicion that portrait farmers could be the cause of this trend, would love it if there was some way to look into it.
|
On December 09 2011 07:18 Torte de Lini wrote: This is well-understood? I thought doing anything in excess or without moderation doesn't necessarily make you better but can deter your play.
Most top players I've heard comment on this say that playing a lot is valuable IF you're otherwise working on your skills in a meaningful way. This makes playing a lot a component of a larger strategy to improve one's skill, but it doesn't preclude that doing so in a vacuum might not be counterproductive.
Casual players are typically lower-ranked than Master leagues. So while the general population ratio of play is more, the amount of people who hardly play can have a drastic effect.
Well, all players are typically ranked lower than Master leagues. However, these numbers are a pretty good indication that there are a lot of casual players (casual judging by activity) in Master league as well as all the other leagues.
Edit: Half the players in Master league have played fewer than 60 games this season. I'd call that pretty casual.
Win traders are below, not the top percentiles I believe.
Win traders are invariably the top percentiles of activity because they need to complete lots of games in order to get their portraits. They're not the top percentiles of rank or MMR, but I wasn't looking at that here.
Mass gaming is relative to the person and thus, so is how to improve and how many games one should play if they want to seek improvement.
So...
All different ways of saying the same thing.
|
Hmm I've always felt that a lot of master league players just play 1 game and never touch it again. It's not that they're clogging up the system or anything, but it's frustrating that people like that are in master league while it's so hard for me to get promoted from master to diamond x.x
|
On December 09 2011 07:24 KimJongChill wrote: Hmm I've always felt that a lot of master league players just play 1 game and never touch it again. It's not that they're clogging up the system or anything, but it's frustrating that people like that are in master league while it's so hard for me to get promoted from master to diamond x.x
What was interesting to me about this data is that it suggests that playing one game and walking away becomes more common the lower you go in leagues. I did not expect that.
|
This is a great post that is well thought out and supported with good data so thanks to the OP for posting it.
|
On December 09 2011 07:25 Lysenko wrote:Show nested quote +On December 09 2011 07:24 KimJongChill wrote: Hmm I've always felt that a lot of master league players just play 1 game and never touch it again. It's not that they're clogging up the system or anything, but it's frustrating that people like that are in master league while it's so hard for me to get promoted from master to diamond x.x What was interesting to me about this data is that it suggests that playing one game and walking away becomes more common the lower you go in leagues. I did not expect that. Well, i think they simply don't care about ladder and just mass customs It is the same since 2008 in Wc3, once you won 10 games straight up, you're matched against good opponents. (my point is that few people try in a competitive game once they felt they are getting crushed)
|
NVM, I missinterpreted the analysis
|
On December 09 2011 07:38 carloselcoco wrote:Show nested quote +* The most active 0.5% of players play 10 times as much as the median player in master league, and 30 to 40 times as much as the median player in platinum through bronze leagues. Stopped reading there.... You clearly did not look at your own graph... And he stated that this was because of win traders aka portrait farmers.
|
On December 09 2011 07:40 Erasme wrote: And he stated that this was because of win traders aka portrait farmers.
He misread something, and edited his post. The top 0.5% of Bronze players (with 289+ games) play about 30 times as much as the median Bronze player (with 7 games), which is just what I said.
|
What exactly do you mean by "win activity" in the chart? Is that just the average number of wins the respective percentile players have in their league?
|
On December 09 2011 07:42 Junichi wrote: What exactly do you mean by "win activity" in the second graphic? Is that just the average number of wins the respective percentile players have in their league?
Yes. Unfortunately, Blizzard doesn't make available the data for total games played. However, the vast majority of individual players have close to a 50% win/loss ratio, with only the bottom of Bronze and the top of Master league deviating from that, so number of wins is a reasonable proxy for activity when you average many players, particularly since most players tend to play others who are close to them in ladder ranking.
I labeled the chart "win activity" because losses are not included, and thus a player who plays 100 games and loses all of them wouldn't show up on the chart as active.
|
I don't know how you can post a graph clearly showing a strong statistical correlation between league and games played and then come to this conclusion
This confirms the opinion often expressed by the best players that mass gaming doesn't directly lead to improvement, in itself.
What data would have NOT confirmed your hypothesis? Someone do an R squared on this and prove the obvious; practice makes you better
edit: here, because I know people dont like to read
Uploaded with ImageShack.us
|
United States431 Posts
you don't have 1 data point on your table that backs up your opinion or the supposed opinions of the best players
|
|
|
|