Ladder makes no sense to me. (Numbers inside)

On August 07 2010 12:33 richlol wrote:
The numbers just don't look right. Are you saying that the average rating for bronze players is 761.9?

On August 07 2010 12:33 richlol wrote:
The numbers just don't look right. Are you saying that the average rating for bronze players is 761.9?

On August 07 2010 12:35 catamorphist wrote:
Can you stop calling the number by your name a rating? It's just points that you get for playing and winning. It doesn't rate anyone, and it doesn't make much sense to compare people by it.

Anyway, this seems reasonable enough to me.

On August 07 2010 12:35 baeracaed wrote:
Well, one thing is that these stats won't be as accurate as they may at first seem, due to the fact that b.net promotes/demotes people between leagues in intervals. There was a recent quote form a Blizzard dev saying it's every 30 games, but I'm pretty sure it's less than that early on.

On August 07 2010 12:58 Chiburi wrote:
The rating/point system is not equal among the leagues. 800 points in bronze league is not the same as 800 points in diamond league. The system has some kind of conversion factor that it uses. When I got promoted from platinum to diamond, my points went from about 600 to about 300.

On August 07 2010 12:30 MichaelJLowell wrote:
Average Win Percentage: 61.13%, 63.41%, 63.61%, 73.19%, 65.80%
Players Above 80%: 0, 0, 0, 13, 2
Players Above 70%: 5, 5, 3, 37, 11

On August 07 2010 13:14 brad drac wrote:

I believe everyone should be ignoring all the numbers apart from these(rating is related to number of games played more than skill and is not comparable cross-league). This is indicative of a problem. I don't know what exactly it is or how to resolve it, but it does seem to be there.

On August 07 2010 12:30 MichaelJLowell wrote:
Noticed on Starcraft 2 Rankings (http://www.sc2ranks.com) that I'm currently in the Top 400 in the North American Diamond League. And quite frankly, I'm not that good. So I was curious: Where is everyone else?

Decided to cull some data. Scanned the Top 50 players in each league on the North American server using the above site. This data is current as of 9:00 P.M. EST on August 6, 2010 and I don't make any promise it's perfect:

Stats per league (from Bronze to Diamond, highest in bold, lowest underscored):

Average Rating : 761.9, 619.4, 588.5, 672.2, 798.6
Average Games: 69.8, 54.3, 58.6, 38.7, 165.8
Average Win Percentage: 61.13%, 63.41%, 63.61%, 73.19%, 65.80%
Players Above 80%: 0, 0, 0, 13, 2
Players Above 70%: 5, 5, 3, 37, 11
Players Above 60%: 28, 38, 48, 50, 43
Players Above 50%: 47, 50, 50, 50, 50

Other than the obvious bottleneck for Platinum League players trying to make their way into Diamond, there's really nothing I can figure out. Why is the Bronze League rated so high and why does it work down as it gets to Gold? Seriously, I'm confused. o.o

If anyone requests it, I'll upload the spreadsheet with my numbers.

On August 07 2010 13:24 Chiburi wrote:

Platinum has a high winrate beacuse their average number of games played is so low. Your first few games are free wins if you are skilled enough to be diamond/plat and most platinum guys just havent played enough games against people of their own skill level. I don't see much of a problem based on the data you highlighted. There are only a few people in the lower leagues with high win rates that probably bombed placements and havent played enough to get moved to where they belong.

"rating is related to number of games played more than skill and is not comparable cross-league"
This part I agree completely with, but win% is also related to games played since your first few games being easy wins can skew your win rate without many games played.

On August 07 2010 13:14 brad drac wrote:

I believe everyone should be ignoring all the numbers apart from these(rating is related to number of games played more than skill and is not comparable cross-league). This is indicative of a problem. I don't know what exactly it is or how to resolve it, but it does seem to be there.

On August 07 2010 13:44 Jermstuddog wrote:
I think one of the biggest issues is peoples understanding of what bronze/silver level play consists of.

For the record, my buddy plays in silver league, his current BO in PvT is:
Get 8 probes
build a pylon
2 gateways, and send zealots into the enemies base. Continue making nothing but zealots, eventually upgrade to 3 gateways.

He asked me for help with his strategy and after I suggested "make more probes" he said that isn't what he's looking for, he wants to be able to win with just zealots.

I was just dumbfounded... he actually wins with this from time to time...

If you are good enough to realize that you suck, you belong in platinum or diamond league. Anything below that has major issues with basic game mechanics.

On August 07 2010 13:33 paralleluniverse wrote:

You sound like someone who has no understanding of the ladder system.

Rating is not related more to the number of games played than skill. This was not true in WC3 and there's no reason why it would be true in SC2. Look at, for example, how few games Dayvie has played compared to those below him.

Win ratio is a combination of 2 factors: the effectiveness of the AMM to find an equal skilled player and personal skill.

Until you can find a statistically rigorous way to separate the contribution from each of the two factors, so that we can examine only the part of win ratio that is due to the personal skill factor, win ratio is not a very useful metric for personal skill and difficult to interpret in terms of personal skill.

On August 07 2010 13:55 brad drac wrote:

You seem to be ignoring the bonus pool. This makes it so that with two players of equal skill, assuming decent matchmaking, the one who's played more games will almost certainly have the better rating. Obviously at the very top level, where all the players probably use their bonus pool almost every day, rating is a decent, if vague, metric of skill. But as the previous poster said "The system only cares whether you win or lose," and I believe the consensus here is that win ratio is the sole determinant of promotion/demotion, which was the only thing I was pointing out a problem with.

On August 07 2010 13:55 brad drac wrote:[snip]Obviously at the very top level, where all the players probably use their bonus pool almost every day, rating is a decent, if vague, metric of skill. But as the previous poster said "The system only cares whether you win or lose," and I believe the consensus here is that win ratio is the sole determinant of promotion/demotion, which was the only thing I was pointing out a problem with.

On August 07 2010 15:12 RiceMenace wrote:
I think diamond is full and platinum players are simply unable to push up anymore.

If you look at this: http://www.sc2ranks.com/ranks/us/platinum/1

You can see CauthonLuck with 50-7 stuck at #1 plat with super high ELO. Can't be a coincidence.

On August 07 2010 16:15 lololol wrote:
These stats are useless.
Points are not comparable not only between leagues, but also between divisions in the same league.

On August 07 2010 16:19 paralleluniverse wrote:

Of course points are comparable between divisions.

Divisions are purely illusory, they don't affect in anyway how you are matched and how points are acquired, only that you are arbitrarily gated off with 100 players with whom you can only be ranked against.

On August 07 2010 17:35 Lunacy wrote:


Points are comparable between divisions, but we don't know what that conversion factor is. Also, as the general skill level of players on b.net changes the conversion factor is going to change as is the meaning of that score within a division.

But, ignoring bonus pool inflation, I would argue that for a player's division score to be accurate their score needs to have stabilized at some value. Meaning that it would be useful to have a graph of how someone's score is changing over time, if the slope of the line is greater than the bonus inflation their score is not high enough.

So, in a nut shell, mass gaming isn't required to have a high score, but you do need a significant number of games, especially at the lower divisions.

On August 07 2010 18:05 JAN0L wrote:
How is it posible that averange win% is above 60% for all leagues???
if some one looses somone wins so for all players averange must be 50%
or there are so many people that played 3/4 placements and never finished them

On August 07 2010 18:05 JAN0L wrote:
How is it posible that averange win% is above 60% for all leagues???
if some one looses somone wins so for all players averange must be 50%
or there are so many people that played 3/4 placements and never finished them

On August 07 2010 20:35 lololol wrote:
Points between divisions in the same league are also not comparable, a blizzard rep confirmed that early in the beta. It was probably in the beta forums, so it may have been deleted by now, since I couldn't find it with a quick google search.

Anyway, I've had several cases with disproportionate points won/lost during the beta. In once case I was shown as favored and lost 16 points for that game, but our point difference was 18 points after the game(so 50 or 66 point difference before the game depending on if he had bonus points left, since he had 0 after the game) and 12 points are lost for an even matchup(16-8 is 66.6% chance to win for me, which is ridiculous for such a small point difference.

Your points are also inflated over time due to the bonus pool, so a player that gets placed in a new division right now, will have to receive something like 400+ bonus points in order to be even to players in old divisions that have accumulated these bonus points over time.

If new players receive bigger and bigger bonus pools after their placement matches, regardless if they are placed in a new league with no players or a league from release, then it's possible that it has been changed, but if they don't, then it's pretty clear points between different divisions even in the same league are not comparable.

On August 07 2010 20:39 Dionyseus wrote:


They are comparable. The favored/unfavored system goes by your hidden rating, not your visual rating. The more games you play the closer the hidden rating will be to your visual rating.

On August 07 2010 21:31 gerundium wrote:


Almost true. If your skill level stagnates and you assume the player pool around you as a constant you indeed would end up in a situation where you could not alter your rank anymore (except for short term variance.).
But your assumption misses two crucial elements: Bonus pool and the player population. The player population is not constant, over the lifetime of a game the general playerbase skill improves and the bottom of the player pool will usually not stick around long (which is partially the reason for the higher average skill.), thus your ranking will not stagnate. Next to that you will never see a stagnation actually happen because bonus pool obscures it completely (everyone's ranking goes up due to bonus pool over time.)

On August 07 2010 17:51 paralleluniverse wrote:
You're confusing division with league.

On August 07 2010 20:35 lololol wrote:
+ Show Spoiler +
Your points are also inflated over time due to the bonus pool, so a player that gets placed in a new division right now, will have to receive something like 400+ bonus points in order to be even to players in old divisions that have accumulated these bonus points over time.

+ Show Spoiler +

On August 07 2010 12:30 MichaelJLowell wrote:
Average Rating : 761.9, 619.4, 588.5, 672.2, 798.6
Average Games: 69.8, 54.3, 58.6, 38.7, 165.8
Average Win Percentage: 61.13%, 63.41%, 63.61%, 73.19%, 65.80%

On August 07 2010 20:35 lololol wrote:
Points between divisions in the same league are also not comparable, a blizzard rep confirmed that early in the beta. It was probably in the beta forums, so it may have been deleted by now, since I couldn't find it with a quick google search.

Anyway, I've had several cases with disproportionate points won/lost during the beta. In once case I was shown as favored and lost 16 points for that game, but our point difference was 18 points after the game(so 50 or 66 point difference before the game depending on if he had bonus points left, since he had 0 after the game) and 12 points are lost for an even matchup(16-8 is 66.6% chance to win for me, which is ridiculous for such a small point difference.

Your points are also inflated over time due to the bonus pool, so a player that gets placed in a new division right now, will have to receive something like 400+ bonus points in order to be even to players in old divisions that have accumulated these bonus points over time.

If new players receive bigger and bigger bonus pools after their placement matches, regardless if they are placed in a new league with no players or a league from release, then it's possible that it has been changed, but if they don't, then it's pretty clear points between different divisions even in the same league are not comparable.

On August 07 2010 22:40 Nightmarjoo wrote:
My assumption is that points are comparable between divisions. But, I've seen weird/buggy point things too. I was 480 diamond yesterday, and played vs a guy who was 654 diamond (he was #3 in his div, I was like #60 in mine) or so, and I won 10 points for beating him. How on earth was I favoured against a guy in the same league as me who had more than 150 more points than me? of.

On August 07 2010 22:52 barrykp wrote:
I'm confused about the so-called 'points'. Some people in this thread referred to them as a rating, with someone saying it is an ELO rating (the numbers seem very low for this, especially given the presence of a bonus pool). Is there a source or a thread I haven't read that clarifies this, or at least that explains the significance of these points?

World of Warcraft arena uses the ELO system and it refers to it as a rating. If Starcraft 2 were using the same system why would it call it points? Of course we all mprobably have hidden ELOs that are used for matchmaking.

This whole system seems very convoluted. I'm not sure why they don't tell us how it works exactly.

On August 07 2010 23:21 ZapRoffo wrote:
Edit: I see a lot of people spreading the idea that bonus pool inflates the ladder, when it doesn't. I think this thread will help explain a lot of the questions people have about the ladder system:

On August 07 2010 23:38 carwashguy wrote:

Bonus pool points actually do cause inflation (how could they not?). The point is that, despite them, the leaderboard will still accurately rank players relative to each other--so long as they keep playing games. It's a positive way of preventing top players sitting on the top, never having to play another game.

Anyway, I quickly mapped out correlations for the top 25 players. Take it for what it's worth:

http://spreadsheets.google.com/pub?key=0AgsClBPVhGPpdDBiRmNuT1daZjRmZkE2V21mNW05Mnc&hl=en&single=true&gid=0&output=html

Win rate somewhat correlates with points, which is about right. There's no reason to believe that the leaderboard doesn't accurately rank people. Moreover, it shows that you don't get to the top simply by playing a whole lot of games. That seems to negatively correlate.

A larger sample size is in order. Anyone up to writing a SED script to parse this data?

On August 08 2010 01:15 ZapRoffo wrote:
... and non-inflationary (it will not grow over time).

On August 08 2010 02:21 carwashguy wrote:
I have read that other thread. First off, I'm talking about the ladder, not the hidden matchmaking rating. Other than league placement and matchmaking, the two ratings don't influence each other. Of course we have no reason to believe that inflation will occur in the hidden rating.

Strictly speaking, bonus pool points will cause lasting inflation: meaning points increase but skill-value doesn't. Say a player earns 200 points. In two months, he could have 400 points without having improved his skill one bit. This is due to the bonus pool constantly pumping new points into the system. Points will grow over time. Where we agree: the ranking shouldn't alter, as everyone's points will grow equally in relation to one another--assuming they continue playing.

If the bonus pool keeps pumping new points into the system, I can't see how you can claim that people's points will go "back down." Where would the points go?

On August 08 2010 02:40 ZapRoffo wrote:

it compares my rating to your hidden rating to evaluate my points: my 1000 vs. your 800, therefore I win less than default (12), +10 or so.

It compares your rating to my hidden rating to evaluate your points: your 1000 vs. my 800, therefore you lose more than default (12), -14 or so.

[...]

As long as people play frequently, this correction will continuously happen to make up for the trickle of bonus points.

On August 08 2010 02:40 ZapRoffo wrote:


The games are not zero-sum with regard to ladder rating.

If my rating is 1000 but my hidden rating is 800, and your rating is 1000 and hidden rating 800, and we play and I win, it compares my rating to your hidden rating to evaluate my points: my 1000 vs. your 800, therefore I win less than default (12), +10 or so.

It compares your rating to my hidden rating to evaluate your points: your 1000 vs. my 800, therefore you lose more than default (12), -14 or so.

If we play again with the opposite result, I will lose 14 and you will gain 10, resulting in a net loss of 4 for both of us because we are both overranked.

As long as people play frequently, this correction will continuously happen to make up for the trickle of bonus points. This can even be controlled (not sure if it is) by having the rate of bonus pool awards be based on amount of activity rather than time.

On August 07 2010 22:37 carwashguy wrote:

I'd expect to see a greater rating disparity among the edges of the logarithmic distribution of players. Your top 50 stats confirm this. In other words, the very worse players are so bad that everyone can beat them, so the top bronze players have it easier. On the other hand, the top Diamond players are so good that no one can beat them, and they have it easier. So it makes sense that we'd see higher points in bronze and diamond.

Top diamond players play a lot of games, which makes a lot of sense, but does anyone have any ideas as to why the top platinum players play the least?

On August 07 2010 22:37 carwashguy wrote:
does anyone have any ideas as to why the top platinum players play the least?

On August 08 2010 03:09 catamorphist wrote:


Is there any evidence that it works this way? I assumed it was non-zero-sum in the other direction, because whenever I win, I get around 20 (before bonus pool), and whenever I lose, I only lose 2 or 3. I've played about 30 games (http://us.battle.net/sc2/en/profile/281144/1/catamorphist/), and I didn't bomb my placements or anything, so I don't think my rating should be absurdly inaccurate.