|
On September 28 2010 00:14 Clearout wrote:Show nested quote +On September 28 2010 00:10 shannn wrote:On September 28 2010 00:07 Clearout wrote:On September 28 2010 00:01 xzidez wrote:The problem is that the system is inflated even without the bonuspool. So even with this system you wont be able to track your actual improvement  .. But its still less inflated Id guess. How is the system inflated without the bonuspool? :S Probably because your points are still inflating just not as much with the addition of bonus pool points. Was pretty sure it stayed the same. E.g. I win 12 points, someone loses 12 points. If you have a ridiculous win rate though, you will climb higher, but if you end at 50% W/L you shouldnt increase in points. Edit: Sorry for derailing. OT: Thanks for this! I'll make use of it to get a general idea of my "skill" 
No. In any ELO system, as more people join the lower ranks (and they do, in droves) the higher ranking scores inflate further. ELO point inflation has even become a controversy at the Grandmaster level of Professional Chess.
|
The bonus points are needed to compensate for those who have way to much time on their hands. Simplified example:
You play 1000 games with a 51% win percentage; at ten points a win, and minus ten points a loss, that's 200 points (no bonus)
You play 100 games and win 60%. Again, at +/- 10 points per win/loss, that's 200 points (no bonus).
In the current system, player 2 pulls ahead of player 1 (as he should) thanks to the bonus that accrues when he's not playing. But, under this standardized points system, the players appear equal, when they are not.
Win percentage doesn't work either; if you have a 60% win rate, but you just got promoted from platinum, you're not any better than a long-time diamond at 52%.
The system is flawed. I have no idea how to improve it.
|
United States12249 Posts
On September 28 2010 05:28 tGhOeOoDry wrote:
The bonus points are needed to compensate for those who have way to much time on their hands. Simplified example:
You play 1000 games with a 51% win percentage; at ten points a win, and minus ten points a loss, that's 200 points (no bonus)
You play 100 games and win 60%. Again, at +/- 10 points per win/loss, that's 200 points (no bonus).
In the current system, player 2 pulls ahead of player 1 (as he should) thanks to the bonus that accrues when he's not playing. But, under this standardized points system, the players appear equal, when they are not.
Win percentage doesn't work either; if you have a 60% win rate, but you just got promoted from platinum, you're not any better than a long-time diamond at 52%.
The system is flawed. I have no idea how to improve it.
The bonus pool accrues at a set rate for everyone, whether they're playing or not. It's not Rested XP from WoW.
|
On September 28 2010 04:50 gogogadgetflow wrote:Show nested quote +On September 28 2010 00:14 Clearout wrote:On September 28 2010 00:10 shannn wrote:On September 28 2010 00:07 Clearout wrote:On September 28 2010 00:01 xzidez wrote:The problem is that the system is inflated even without the bonuspool. So even with this system you wont be able to track your actual improvement .. But its still less inflated Id guess. How is the system inflated without the bonuspool? :S Probably because your points are still inflating just not as much with the addition of bonus pool points. Was pretty sure it stayed the same. E.g. I win 12 points, someone loses 12 points. If you have a ridiculous win rate though, you will climb higher, but if you end at 50% W/L you shouldnt increase in points. Edit: Sorry for derailing. OT: Thanks for this! I'll make use of it to get a general idea of my "skill" No. In any ELO system, as more people join the lower ranks (and they do, in droves) the higher ranking scores inflate further. ELO point inflation has even become a controversy at the Grandmaster level of Professional Chess.
However wouldn't this system eliminate a HUGE amount of inflation? Like much larger than ELO inflation? I mean the bonus pool accumulates at a constant rate here.
I like this idea. It's too confusing to keep track of how high people are right now. With the standardized rating, the point inflation would be a ton more stable.
Although I think I'm like 200 pts with this system so it must be lame 
|
United States12249 Posts
On September 28 2010 04:50 gogogadgetflow wrote:Show nested quote +On September 28 2010 00:14 Clearout wrote:On September 28 2010 00:10 shannn wrote:On September 28 2010 00:07 Clearout wrote:On September 28 2010 00:01 xzidez wrote:The problem is that the system is inflated even without the bonuspool. So even with this system you wont be able to track your actual improvement .. But its still less inflated Id guess. How is the system inflated without the bonuspool? :S Probably because your points are still inflating just not as much with the addition of bonus pool points. Was pretty sure it stayed the same. E.g. I win 12 points, someone loses 12 points. If you have a ridiculous win rate though, you will climb higher, but if you end at 50% W/L you shouldnt increase in points. Edit: Sorry for derailing. OT: Thanks for this! I'll make use of it to get a general idea of my "skill" No. In any ELO system, as more people join the lower ranks (and they do, in droves) the higher ranking scores inflate further. ELO point inflation has even become a controversy at the Grandmaster level of Professional Chess.
Except I think the Blizzard system for SC2 sidesteps this problem by dividing players into separate Leagues and resetting points every season.
|
United States4126 Posts
When a player gets promoted to a higher league, does their rating change by the same amount as someone else's? I remember I was promoted from a 800 plat player to a 500 diamond player.
|
sc2ranks has a convenient graph where you can see the rating history of players. In the last 35 days, the top ranks have gone up by about 700 points, while the bonus pool has only been 420 in the same time. So there are other sources of (significant) inflation besides the bonus pool.
Furthermore, as ZapRoffo's quoted post above explains, the bonus pool may not induce any inflation at all (though the theory on the Ladder Misconceptions FAQ is that the game compares current points minus bonus pool to MMR, while ZapRoffo's post assumes that the game compares current points directly to MMR, and the former would cause inflation equal to the bonus pool)
It's possible that the inflation we are seeing is primarily the system taking its sweet time to converge to players' true skill levels and bonus points have nothing to do with it, and in a few weeks we could see the rate of inflation drop below the bonus pool rate of 12/day.
|
On September 28 2010 05:33 Excalibur_Z wrote:Show nested quote +On September 28 2010 05:28 tGhOeOoDry wrote:
The bonus points are needed to compensate for those who have way to much time on their hands. Simplified example:
You play 1000 games with a 51% win percentage; at ten points a win, and minus ten points a loss, that's 200 points (no bonus)
You play 100 games and win 60%. Again, at +/- 10 points per win/loss, that's 200 points (no bonus).
In the current system, player 2 pulls ahead of player 1 (as he should) thanks to the bonus that accrues when he's not playing. But, under this standardized points system, the players appear equal, when they are not.
Win percentage doesn't work either; if you have a 60% win rate, but you just got promoted from platinum, you're not any better than a long-time diamond at 52%.
The system is flawed. I have no idea how to improve it. The bonus pool accrues at a set rate for everyone, whether they're playing or not. It's not Rested XP from WoW.
To clarify: If you win one game every day, you get 10 bonus points per win. If you win 10 times a day, you get 1 bonus point per win. Whether intended or not (I believe intended, but I have no source and therefore cannot say), the effect is that bonus points help people who are better but play less keep up in the rankings.
|
On September 28 2010 07:33 Garion wrote:
sc2ranks has a convenient graph where you can see the rating history of players. In the last 35 days, the top ranks have gone up by about 700 points, while the bonus pool has only been 420 in the same time. So there are other sources of (significant) inflation besides the bonus pool.
Furthermore, as ZapRoffo's quoted post above explains, the bonus pool may not induce any inflation at all (though the theory on the Ladder Misconceptions FAQ is that the game compares current points minus bonus pool to MMR, while ZapRoffo's post assumes that the game compares current points directly to MMR, and the former would cause inflation equal to the bonus pool)
It's possible that the inflation we are seeing is primarily the system taking its sweet time to converge to players' true skill levels and bonus points have nothing to do with it, and in a few weeks we could see the rate of inflation drop below the bonus pool rate of 12/day.
The TOP players have not had their win ratio stabilize at 50%. This means that the more they play, the higher their rating is going to get. They win more than they lose after all.
To clarify: If you win one game every day, you get 10 bonus points per win. If you win 10 times a day, you get 1 bonus point per win. Whether intended or not (I believe intended, but I have no source and therefore cannot say), the effect is that bonus points help people who are better but play less keep up in the rankings.
It doesn't matter how many games you play in a day. You get bonus pool points at a rate of about 1point every 2 hours. This is a set rate that doesn't change for anyone - everyone gets bonus pool points at this rate. Whenever you win, you get points for winning and also a bonus pool point per winning point until you run out of bonus pool.
EXAMPLE
You have a bonus pool of 100 saved up. You win an even match which normally gives you +10. You also get an additional 10 bonus points from this win putting your actual points gained at +20. The 10 bonus points you get are subtracted from your bonus pool points leaving you at 90 bonus pool.
You have a bonus pool of 4 saved up. You win an even match which normally gives you +10. You would normally get an additional 10 bonus points from this win, but you only have 4 so you get the maximum amount you can get which is 4. This puts your actual points gained at +14 and your bonus pool at 0.
|
On September 28 2010 04:00 funkybovinator wrote:http://www.teamliquid.net/forum/viewmessage.php?topic_id=118212Show nested quote +On May 11 2010 23:51 ZapRoffo wrote:
Some more information:
With a matchmaking rating system, the way points are assigned is as follows. There is a default point assignment (was +/-12 for wow, seems similar in sc2) for an "equal match result". The amount won or lost in any given match, though, is determined by comparing your displayed rating to your opponent's matchmaking rating. This is why many people are experiencing huge gains for wins and small losses. It's because they haven't played enough to raise their displayed rating to their matchmaking rating. They may be matched as an 1800 matchmaking rating, but are at 1300, so if they win against an equal opponent (1800 matchmaking), they get the points of a 1300 beating an 1800, which may be +20 or something. The opponent compares his displayed rating to your matchmaking rating to calculate his point change, if he's displayed 1600 and you are also 1800 matchmaking, he will lose -10 or so (slightly less than -12 default).
One huge misconception people I feel like people need to learn the truth about:
The bonus pool WILL NOT cause inflation of ratings in the long run as long as it only modifies your displayed rating and not your matchmaking rating, which appears to be the case. In the long run, displayed ratings converge to matchmaking rating, so if matchmaking rating is unaffected there is no long term effect.
An example:
I start with a big bonus pool and win up to 1600, and my matchmaking rating is 1700. Alice wins the same amount against similar quality opponents but with no bonus pool and goes to only 1350 or so, but also with 1700 matchmaking rating, because matchmaking is totally unaffected. Now in my games I will only be looking at winning +13 or so from my opponents who are 1700 matchmaking, while Alice is looking at something like +16 or +17 from her 1700 matchmaking opponents. I'm looking at -10 or -11 from losses, while she's looking at -8 or so from those same people. Eventually the result over a long enough period is we both end up at 1700 if no change in skill happens.
Even if I got enough of a bonus pool to get to 1900 or something, once that runs out I'm going to lose more for losses than I get for wins against people who are my skill level until I get to the appropriate level. The bonus pool just functions to get people's displayed rating jump started so if they took a break they can jump to their rating more quickly.
I don't think that's right anymore funky, by observing what's happened. I was assuming bonus pool points affect how much you win/lose from future matches because your rating is overvalued. Seems this is not the case, since the graphs on sc2ranks seem to show rather constant growth.
|
I don't think that's right anymore funky, by observing what's happened. I was assuming bonus pool points affect how much you win/lose from future matches because your rating is overvalued. Seems this is not the case, since the graphs on sc2ranks seem to show rather constant growth.
Except the growth is faster than the bonus pool. And if you look at the top ranked players, they are still getting more than 12 points per win, indicating that their MMRs are higher than their points, with or without adjustment for bonus pool.
There is at least 240 points per month inflation happening beyond the bonus pool, whether this is from new players joining, the system taking time to converge, or something else entirely.
If the bonus pool is adjusted for when awarding points, then we have:
360 points/month bonus pool inflation+
240 points/month unknown inflation.
If the bonus pool is not adjusted for (i.e. your original theory), then we have:
600 points/month unknown inflation.
I haven't seen any evidence yet suggesting that either one is not the case.
|
if you then add this number to 1500 you can see your real elo
|
yea i understand why blizzard implemented the bonus pool, it really does trick you into thinking you improved... and that feels good! but what i do to get a better idea of my actual improvement is just to pick the top 3 or so in my division and see how far i am... so if i'm 400 points behind them a week ago and still 400 points behind this week, then even though my rating has gotten higher, i haven't really gotten much better...
to take this idea further, it would be nice to create an index of a few top players in the world and use that average to be the standard, constantly updated of course...
|
On September 29 2010 22:54 Garion wrote:Show nested quote +
I don't think that's right anymore funky, by observing what's happened. I was assuming bonus pool points affect how much you win/lose from future matches because your rating is overvalued. Seems this is not the case, since the graphs on sc2ranks seem to show rather constant growth.
Except the growth is faster than the bonus pool. And if you look at the top ranked players, they are still getting more than 12 points per win, indicating that their MMRs are higher than their points, with or without adjustment for bonus pool. There is at least 240 points per month inflation happening beyond the bonus pool, whether this is from new players joining, the system taking time to converge, or something else entirely. If the bonus pool is adjusted for when awarding points, then we have: 360 points/month bonus pool inflation+ 240 points/month unknown inflation. If the bonus pool is not adjusted for (i.e. your original theory), then we have: 600 points/month unknown inflation. I haven't seen any evidence yet suggesting that either one is not the case.
This isn't really what I'd call inflation. The reason for this is because these players have not had their rating settle yet. The matchmaking system will try to match players higher and higher until they stabilize at about a 50% win ratio. However, since these players are at the top of the ladder, the system does not work that well for them since there are no higher players to match them up with. For the average player who's starting to even out into roughly a 50% win ratio, there should be no inflation in the standardized rating that does not reflect an increase in skill relative to their peers. Because this is relative to the populace, if more people join who are generally worse than this player, his rating will go up because his relative skill level will be higher.
if you then add this number to 1500 you can see your real elo
This is false. Your REAL rating (HIDDEN MMR rating) is HIDDEN. There is no known way to pinpoint your exact MMR.
|
so why not just use your world/region ranking as the most solid rating, since it normalizes bonus pool and at high levels of play, it can be assumed that bonus pool is always exhausted?
|
I would like a standardized rating because right now the points dont mean anything.
|
I just track my progress by looking at my global rank on SC2ranks.com
|
Global rank on SC2ranks is a pretty good indicator of where you're at... but it only stays relevant if your bonus pool has been exhausted. This standardized rating system will roughly work even if you have hundreds of bonus points left.
|
On September 28 2010 00:07 Clearout wrote:Show nested quote +On September 28 2010 00:01 xzidez wrote:The problem is that the system is inflated even without the bonuspool. So even with this system you wont be able to track your actual improvement .. But its still less inflated Id guess. How is the system inflated without the bonuspool? :S
the point is:
if 2 teams (or 1on1) play against each others
and both are equally matched
the winner gets 12 points
the loser loses 10 points
i dont know about patch 1.1++ but from early beta until 1.0 it run that way.
to prove or disprove my point, just watch the match history of your enemy and compare.
if they ALWAYS lose as mutch points as you win, there will be no direct inflation (bonus points excluded)
|
Just reducing the bonus pool won't help you pinpoint the MMR raiting, at all. I'm relatively low on pts, but compared to people that i've been near on blizzard's top 200 list, they'd be on average ~400 pts ahead of me, with similiar bonus pool left.
The points you gain/lose from the game are not only influenced by both player's MMR, but (i don't remember correctly) comparison of one player's MMR to other's ladder rating; same in reverse - both can appear 'favored' to each other, etc. MMR changes, however, only depending on opponent's MMR. There's a ton of confusion which results in "wtf 0_o" rating.
The only thing i more or less trust(since it's also not very stable due to weird bonus pool effects) is relative ranking worldwide on sc2ranks.com.
Still, it wasn't uncommon for me to keep winning +5 pts from players ~300-400 pts above me, and then lose ~20pts on a game against same kind of player. Makes me facepalm.
|
|
|
|
|
|
EPT
