EPTI've been trying to look up the post for a while now but it's buried pretty deep within one of the various sc2 forums.
Searchable Ladder, Updated Daily - Page 9
| Forum Index > SC2 General |
|
shinosai
United States1577 Posts
I've been trying to look up the post for a while now but it's buried pretty deep within one of the various sc2 forums. | ||
|
shinosai
United States1577 Posts
http://forums.battle.net/thread.html?topicId=23766800567&sid=5000&pageNo=2 Post 35 Sorry I misunderstood what the OP was referring to. I didn't mean to imply that one division is ranked better than the other, but simply explaining the basics of divisions. As far as comparison across divisions it's certainly something we've considered but there are issues, such that the rankings in one division don't directly translate to the other divisions. So you couldn't compare division 10 to division 48 and compare one player's points to another. | ||
|
btlyger
United States470 Posts
Great work man, hope to see this continue! | ||
|
shinosai
United States1577 Posts
Presuming 1500 is the average rating in division 1, but 2000 is the average rating in division 2, we would adjust everyone in division 2 downwards by 25% to achieve equality. That's the most balanced way I can think to do it with the presumption that rankings don't directly translate between each other. | ||
|
Gibybo
United States229 Posts
On April 12 2010 13:12 shinosai wrote: Proof here: http://forums.battle.net/thread.html?topicId=23766800567&sid=5000&pageNo=2 Post 35 Thanks for the link, this is interesting. I understand that rankings are not directly comparable (obviously since different divisions have different random levels of skill), but I don't understand why you can't compare the points. Everyone plays each other and the points awarded/lost are based on the hidden rating which afaik has nothing to do with division. I would appreciate any logical reasoning about why one division's points are more/less valuable than another division's points. At the very least they seem to be pretty close right now. On April 12 2010 13:28 shinosai wrote: Perhaps if you took the average of each division and adjusted ratings upwards or downwards based on the percentage of how far the division is from average, you'd have more accurate numbers. Just a guess, though. For example: Presuming 1500 is the average rating in division 1, but 2000 is the average rating in division 2, we would adjust everyone in division 2 downwards by 25% to achieve equality. That's the most balanced way I can think to do it with the presumption that rankings don't directly translate between each other. That's an interesting idea and certainly one way of doing it (and very easy for me to implement), but it makes some assumptions beyond what we know: - That divisions as a whole vary with a bias to how easy it is to get points in them. - That the average rating of a division is proportional to how easy it is to get points in that division. - That the effect of this bias applies equally to all players within it. I think it's relatively easy to show that at least one of those is wrong with a simple example: Consider a new division 45 with one strong player and 99 players that hardly play. This one strong player will have the same bonus points as the players in division one and will be playing against the players in division one (and all the other divisions) such that the points gained by the player in division 45 are equal the amount of points a player in division 1 loses (outside of bonus points) when they play against each other. The average of division 45 is still roughly 1000, and the average of division 1 is 1200. In effect, the player in division 45 gets a 20% bonus to his points when playing directly against a player in division 1, why? | ||
|
shinosai
United States1577 Posts
On April 12 2010 13:32 Gibybo wrote: Thanks for the link, this is interesting. I understand that rankings are not directly comparable (obviously since different divisions have different random levels of skill), but I don't understand why you can't compare the points. Everyone plays each other and the points awarded/lost are based on the hidden rating which afaik has nothing to do with division. I would appreciate any logical reasoning about why one division's points are more/less valuable than another division's points. At the very least they seem to be pretty close right now. Any logical reason would be pure speculation. | ||
|
Gibybo
United States229 Posts
On April 12 2010 13:34 shinosai wrote: Any logical reason would be pure speculation. We have quite a lot of data and a pretty decent understanding of the ladder system. We can do a lot more than speculate with that. On April 12 2010 13:16 btlyger wrote: Hey just wanted to say the new layout and coding is amaaazzing. I was easily able to find myself and see my rank and everything, and was easily and QUICKLY able to browse the rest of the ladder. Great work man, hope to see this continue! Thanks!  | ||
|
shinosai
United States1577 Posts
On April 12 2010 13:50 Gibybo wrote: We have quite a lot of data and a pretty decent understanding of the ladder system. We can do a lot more than speculate with that. Well to start with you'd have to start recording rating exchanges between various players and establish a point of comparison to see if in fact rating exchange is unbiased in terms of divisions. It's not unreasonable to ask this as Blizzard says that the points are not comparable - so we'd just need to prove it for ourselves. A logical reason (again, speculation) would be that one division's points are more or less valuable than another division's points because: Blizzard has coded the ladder so that each individual division exists in its own bubble, with points being given out based on a combination of rank within the division, the existing points within the division, and ELO (the only thing people in different divisions have in common). If we do establish that the rating exchange varies from division to division, however, we come into a pretty huge problem: If Blizzard can't compare points between divisions yet, then we'd have to figure out a formula ourselves. And even for a person gifted in mathematics, this would be quite difficult. Thus my advice on averaging the points is one for simplicities sake, rather than accuracy. edit: Just from a personal point of view (this isn't good evidence, just my subjective experience) I'd say the system definitely feels like it's based more on rankings than it is in points for comparison. I'm in the top 10, I regularly play (and beat) top 10 players. I don't regularly fight people that are near my "points". | ||
|
Gibybo
United States229 Posts
On April 12 2010 14:07 shinosai wrote: ell to start with you'd have to start recording rating exchanges between various players and establish a point of comparison to see if in fact rating exchange is unbiased in terms of divisions. I just started doing this by asking my opponents how many points they gained/lost after the game. It would be great if some other people did this too so we could get some more data I played one game against someone in another division, I got 11 points but he only lost 7. (div 28) I played another game against someone in my own division, I got 14 and he lost 14. (div 21) I played another game against someone in another division, I lost 6 he gained 8 (div 7) Of course bonus points are excluded. Just for reference, the average ratings of those divisions as of 4/11 10 AM: 28: 1181.825 21: 1131.927 7: 1205.135 Out of this very large sample size it does appear there might be a bias to different divisions. Now we just have to figure out what it is (and why..) | ||
|
nagash
Australia58 Posts
Edit: Ah I have the flu, my reading comprehension is not so great, sorry about that. | ||
|
Gibybo
United States229 Posts
On April 12 2010 21:11 nagash wrote: Keep in mind, when you win you gain extra points from your bonus pool. So if you weren't taking that into account that could be why you have different readings? On April 12 2010 14:57 Gibybo wrote: Of course bonus points are excluded. | ||
|
radiatoren
Denmark1907 Posts
On April 12 2010 14:57 Gibybo wrote: I just started doing this by asking my opponents how many points they gained/lost after the game. It would be great if some other people did this too so we could get some more data I played one game against someone in another division, I got 11 points but he only lost 7. (div 28) I played another game against someone in my own division, I got 14 and he lost 14. (div 21) I played another game against someone in another division, I lost 6 he gained 8 (div 7) Of course bonus points are excluded. Just for reference, the average ratings of those divisions as of 4/11 10 AM: 28: 1181.825 21: 1131.927 7: 1205.135 Out of this very large sample size it does appear there might be a bias to different divisions. Now we just have to figure out what it is (and why..) That is very interesting. However i think the most important part of comparing divisions are the number of points gained from bonus pools in each division. If one division has gotten 1200 bonus points and another has gained 2000 there will be more points gained outside of the games on the latter division. If i had to make a system, that would be the only problem with comparing divisions. Therefore it could be funny to have stats like how many people are in the divisions each hour/day (depending on how exactly that part of the system works) and how that would impact the points gained or lost. Also tracking of activity of each player in the divisions are needed to gain a deeper knowledge of this function in the ladder. Those stats would give an enormous advantage in developing the comparison of the divisions. | ||
|
Gibybo
United States229 Posts
On April 12 2010 21:28 radiatoren wrote:That is very interesting. However i think the most important part of comparing divisions are the number of points gained from bonus pools in each division. If one division has gotten 1200 bonus points and another has gained 2000 there will be more points gained outside of the games on the latter division. If i had to make a system, that would be the only problem with comparing divisions. Afaik, bonus points are awarded equally to all players (if someone joins later, they are given enough to make up for the ones they missed), so certain divisions shouldn't have more than others. I might be wrong of course but there was another thread on this forum about how it worked with more details. Therefore it could be funny to have stats like how many people are in the divisions each hour/day (depending on how exactly that part of the system works) and how that would impact the points gained or lost. Also tracking of activity of each player in the divisions are needed to gain a deeper knowledge of this function in the ladder. Those stats would give an enormous advantage in developing the comparison of the divisions. I'm not sure what you mean here, like how often they are online? Or how the divisions grow over time in terms of the number of people in them? | ||
|
Tamerlane
Canada424 Posts
On April 12 2010 13:12 shinosai wrote: Proof here: http://forums.battle.net/thread.html?topicId=23766800567&sid=5000&pageNo=2 Post 35 That was posted before the ladder reset, it was true for that period, but the calculation for bonus pool changed a lot since then. I found out that before the reset, you gained 1 bonus point (sent in the pool) for every player that was in the division (whether they were there when you got it or when they joined) and I can guarantee this is no longer the case after the reset. The only variable I can envisage with the new ladder system regarding the bonus pool, is how much time has elapsed starting at some point which I believe is fixed for everyone, that would be the time at which the ladder reset. As I am writing this, the reset has been done 17 days and 9 hours ago. I have used 45 bonus points when I was in the silver league, which amounts to approx. 5 days and 2 hours after the reset (for the last bonus point used) - that's an average of 2,71 hours/pt. I have used 115 points while I was (still am) in the gold division. I have been in that league for 12 days, 3 hours and 25 minutes (between switch and the time I used my last bonus point), that would make an average of 2.53 hours/pt. But since that was after a ladder promotion, we would have to calculate that on a period where the initial bonus pool was depleted (on top that I had not 1000 MMR after promotion). So, I depleted my initial bonus pool on 2010-04-05 at 01:22 and earned my last bonus point on 2010-04-12 at 06:52 (although it is entirely possible that I earned that last point in the previous 2 hours, that's an margin of error of ~0,05) - that's a length of 7 days, 5.5 hours (and for a total of 60pts). That makes an average of 2.89 hours/pt. On the other hand, I had entered a 2v2 division quite some time after the reset (april 3rd) with 77 points in the bonus pool, that's an average of 2,87 hours/pt, which is inside the margin of error for the value calculated since I am in the gold division. Bottom line is, I would be extremely surprised there would be another factor than time that affects the bonus pool. | ||
|
shinosai
United States1577 Posts
I agree with the post above me that says that time is the only thing that affects the bonus pool, as it goes in line with what previous blue posters have told us. | ||
|
Mt.DiabloakaUniverse
14 Posts
Now when I get mercd I know they have awesome stats and don't feel so bad!!! | ||
|
Autodidact
United States27 Posts
Thanks for the site!! | ||
|
fillyjonk
8 Posts
http://aoe3.game-server.cc/rating2/ladder/age3ySP1v1 It provides a very clear overview of player rankings and personal stats, in addition to having fun and handy features like the ability to view your rating history in graph form to illustrate your skill progression through total played time. But for starters maybe you could just steal the clean layout! Edit: Completely forgot to give my appreciation for starcraftrankings as well! The above post was not to say that starcraftrankings is not very much appreciated in its current form. (Irregardless of problems of direct rating comparisons and personal layout preferences!) Extremely nice feature having links to replays right at the player list, I must say! | ||
|
Jugan
United States1566 Posts
Thanks for all the hard work! | ||
|
Pervect
1280 Posts
| ||
| ||
