[G] Comprehensive SC2 League and Ladder Guide - Page 18

On May 12 2011 03:22 fishjie wrote:
Oh cool, so back when I was still in diamond, I was in the top 20% of north america? that's cool. Too bad i'm platinum now.

Question: I didn't see it in the article, but the player skill distribution graph looks like a bell curve/gaussian/normal distribution or whatever you call it. Is that correct? So majority of players are within one standard deviation of gold? My statistics is garbage, so sorry if I didn't use terms correctly.

On May 12 2011 03:55 Excalibur_Z wrote:


Yeah. It's not exactly like that, and supposedly there's a little bump in the curve in Diamond and Master and this is explained by more hardcore players who visit TL and watch pro streams, but I didn't know how to illustrate that in the curve so I just left it as is. It should still be pretty accurate. The curve is actually something I threw together in wolframalpha and I wanted the breakpoints to be as close to accurate as possible, so when I imported all the breakpoints into an image editor I just plotted where the 98% spot is, where the 80% spot is, etc. For the other leagues it should be pretty close to 100% accurate, the only difference would be a tiny bump toward the end of the curve.

On May 12 2011 05:55 aksfjh wrote:


Well, like all skill based systems, they try to attain the Gaussian distribution. In this case, we're actually seeing something like a log-normal distribution, where the mean, median, and mode are not located in the same spot. The overall nature of the system is the same, where a majority of the (active) population exists in a single "bulge," but the taper on one end happens over a longer segment.

How does this happen?
You see, in a Gaussian system, everything has a natural tendency to gravitate to a single point to varying degrees. In a ranking system, this is the root, starting point, or whatever you might call it. It's the rating that everybody starts with. Ideally, in all systems, everybody participates the same amount, which keeps the points distributed evenly throughout the players. However, we all know that many Silver and Bronze players don't play as often as Masters or Diamond players. This begins to skew the "bulge" in the bell towards those which play less, while decreasing the rate of the slope on the opposite side.

A good way to imagine this: 10 bad players get on, play and lose 1 game, and then quit. 5 other good players are playing during this same time, but don't quit. While one of the bad isn't on, they play one another and their points revolve around a mean value. In this scenario, the root score is 1000, so those 5 players have an average score of 1000, winning 5 points for a win, losing 5 for a loss. At the end of the day, those 5 players have a total of 50 points injected into their group from those 10 players that got on and lost to one of the 5 players, so their average becomes 1010. Now, there are 5 players whose average scores revolve around 1010, and 10 players whose average score revolves around 995. Do this over a long stretch of time and you begin to get a larger population whose average revolve around a number less than the root, and a smaller population whose average revolves around a number greater than the root. In essence, multiple Gaussian distributed populations.

On May 12 2011 09:00 IzieBoy wrote:
such beautiful graphs (what program did you use?)

On May 07 2011 04:51 Excalibur_Z wrote:
Okay, so the ladder adjustment hotfixes just went live on the remaining (non-NA) servers yesterday. This was the hotfix that modified the MMR breakpoints for each league. This is going to cause a shift in league populations over time, as more and more players play games post-hotfix. What is important to note about this hotfix is that it applies to each region differently.

What this should mean is that the populations for each server should start going back down to the expected percentiles (2/18/20/20/20/20), but because the MMR breakpoints are no longer going to be the same for each region, there may be actual skill discrepancies across regions for the same leagues.

To use some hypothetical numbers as an example (and I'm just using SC2Ranks values because they're more visible, active players will be a little different but this should illustrate what the effects will be), let's say that right now 55% of the population in LA is in Bronze while 26% of China is in Bronze, and the Bronze threshold is like 1000 MMR. LA has fewer players which means skill doesn't deviate as much from the starting point. What will happen is that after the hotfix, LA will get closer to 20% in Bronze, and China will also. However, China is already pretty close to 20% in Bronze so their MMR breakpoint for Bronze won't change very much. LA is really far from 20% so the MMR breakpoint will change to whatever matches up with the current 20th percentile, so maybe 300 MMR or something.

Not sure how this will impact servers like KR which is already regarded as far and away the most skilled server, but I can predict that we'll see more types of statements like "Korean Plat = NA Master" =)

On May 11 2011 14:56 Belisarius wrote:
This is a really great article.

Just out of interest, do we know how blizzard ended up handling the ladder reset with respect to MMR?

We obviously didn't see everyone earning 20+ for each win after the reset, which is what I'd expect if MMR carried over and points didn't, but at the same time it seemed to place everyone pretty accurately back where they were.

I apologise if this has been previously explained.

On May 15 2011 02:03 DexVitality wrote:
Cool with this I can really understand how the placement system works but my question is has blizzard made it easier to get into masters for arranged teams? I play in SEA and for randoms it is really really hard to get in but lately with AT, you can be in masters after 5-0 Placement or sometimes even 4-1... which I find may be the wrong way to go, people should start and work their way up, not start at the top right after 5 games.

On May 15 2011 18:48 Trowa127 wrote:
Good read, thanks a lot. I'm World rank 500 in platinum or something at the moment as I recently switched races and I started playing Diamonds about 30 games ago - I'm currently 16 wins 12 losses but still no promotion - I'm assuming it will happen soon.

On May 15 2011 19:10 Gulzt wrote:


Just 30 games this season?? Or since you're playing against diamond players?
I mean, with 30 games of which you lost almost half it's impossible to be 500th in the world in platinum, since that isn't even enough to empty your entire bonus pool.

Anyway, when I was finally promoted to diamond I had a 80% winratio for 20 games. Doesn't mean anything though, as it's impossible to know what my MMR rating was at that point.. I was always top8 platinum though. Since the hotfix it should be slightly easier though.