|
On January 05 2014 14:58 Spirit09 wrote:Show nested quote +On January 05 2014 14:24 korona wrote:
There are so few teams with higher league MMR that you might have been playing more against teams that have lower MMR than your team. You lose more by losing those teams than gain by winning them. And there are lots of decayed teams grinding back up. Correct me if I'm wrong, but as long as mmr decay exists (i.e. as long as some significant number of players/teams experience it and than return to the ladder) does this not mean that even players/teams who play regularly will effectively experience a drop in MMR due to playing against others who have experience mmr decay directly? Does mmr not need to be zero sum to prevent this, and decay is not just not zero sum, but is also entirely one sided in that people only decay down, not up? I'm already imagined that having some proportion of the player base drop mmr over time will result in the entire player base dropping mmr over time due to the fact that we all play against each other (six degrees of separation style etc..). What you say might have happened to me would appear to be this occurring. This would mean Blizz would need to continuously lower league offsets by some proportion related to the number of games in which players experiencing mmr decay play.
Yes. If there was a player A who had MMR of 600 and then went inactive for more than 4 weeks, his MMR would be dropped to 300. He would now be considered as 300 MMR player. And if it would take him 100 matches to grind his way back up to 600 MMR (60 W - 40 L), he would affect all his 100 opponents (just as much wins and losses normally affect based on the current MMR differences). MMRs of 60 of his opponents would drop.
And yes. If Blizzard is going to fight the cascading effect only via adjusting thresholds, it will be continuous effort. And it does not fix the problem that wildly different skilled players are sharing similar MMRs due to the decay.
|
I am not sure if its been answered
Instead of SC2gears now Scelight is coming
Are you planning to switch MMR tool to Scelight?
Thanx
|
|
|
This plugin does not seem to work on my Mac 
I played 18 Games by now but there was not a single game with 100% data quality.
If I take a look into the log, it looks always the same for every game.
Here for example my last:
22:16:16 Post-trigger: Start 22:16:56 Post-trigger: Reading web profiles 22:17:00 Webgrabber: Player true 22:17:04 Webgrabber: Opponent true 22:17:04 Post-trigger: Done reading web profiles 22:17:04 Bonus calc (Nitai diam-): max end: 87, max fetch: 87 22:17:04 Bonus calc (ZioCroccante diam-): max end: 87, max fetch: 87 22:17:04 Calculating adjusted points for the players (either Alg. A or B needs to succeed for certain player) 22:17:04 DATA Alg. A : failed, No pre-game data (RAM scanner possibly off) or pre-game data bad: 22:17:04 DATA Alg.B: failed, No Bonus Win: http://eu.battle.net/sc2/en/profile/4169060/1/Nitai/22:17:04 ERROR AFTERCHECK Both alg. A and B failed: http://eu.battle.net/sc2/en/profile/4169060/1/Nitai/22:17:04 DATA Alg. A : failed, No pre-game data (RAM scanner possibly off) or pre-game data bad: 22:17:04 DATA Alg.B: failed, No Bonus Win: http://eu.battle.net/sc2/en/profile/2556827/1/ZioCroccante/22:17:04 ERROR AFTERCHECK Both alg. A and B failed: http://eu.battle.net/sc2/en/profile/2556827/1/ZioCroccante/22:17:04 ERROR LASTCHECK : Both players invalid 22:17:04 Gamecheck: false 22:17:04 POST-Trigger done 22:17:04 Data quality needs to be 100% to be able to calculate MMR 22:17:04 Data quality for calculating Player's MMR: 50% 22:17:04 Data quality for calculating Opponent's MMR: 50% 22:17:05 OK Uploader: game uploaded 22:17:05 Ladder Game Done!
Thats how every game looks like
The odd thing seems to be the line "No Bonus Win" because that is not true!
Before I started this game, I checked and had 10 bonus pool and I won that match and got all 10 points.
So, what am I doing wrong?
I really want to use this tool.
PS: the memory reader is not available for Mac OS, otherwise I guess, there would already be a 100% game.
|
On January 08 2014 06:32 nitai wrote:This plugin does not seem to work on my Mac I played 18 Games by now but there was not a single game with 100% data quality. If I take a look into the log, it looks always the same for every game. Here for example my last: Show nested quote +22:16:16 Post-trigger: Start 22:16:56 Post-trigger: Reading web profiles 22:17:00 Webgrabber: Player true 22:17:04 Webgrabber: Opponent true 22:17:04 Post-trigger: Done reading web profiles 22:17:04 Bonus calc (Nitai diam-): max end: 87, max fetch: 87 22:17:04 Bonus calc (ZioCroccante diam-): max end: 87, max fetch: 87 22:17:04 Calculating adjusted points for the players (either Alg. A or B needs to succeed for certain player) 22:17:04 DATA Alg. A : failed, No pre-game data (RAM scanner possibly off) or pre-game data bad: 22:17:04 DATA Alg.B: failed, No Bonus Win: http://eu.battle.net/sc2/en/profile/4169060/1/Nitai/22:17:04 ERROR AFTERCHECK Both alg. A and B failed: http://eu.battle.net/sc2/en/profile/4169060/1/Nitai/22:17:04 DATA Alg. A : failed, No pre-game data (RAM scanner possibly off) or pre-game data bad: 22:17:04 DATA Alg.B: failed, No Bonus Win: http://eu.battle.net/sc2/en/profile/2556827/1/ZioCroccante/22:17:04 ERROR AFTERCHECK Both alg. A and B failed: http://eu.battle.net/sc2/en/profile/2556827/1/ZioCroccante/22:17:04 ERROR LASTCHECK : Both players invalid 22:17:04 Gamecheck: false 22:17:04 POST-Trigger done 22:17:04 Data quality needs to be 100% to be able to calculate MMR 22:17:04 Data quality for calculating Player's MMR: 50% 22:17:04 Data quality for calculating Opponent's MMR: 50% 22:17:05 OK Uploader: game uploaded 22:17:05 Ladder Game Done! Thats how every game looks like The odd thing seems to be the line "No Bonus Win" because that is not true! Before I started this game, I checked and had 10 bonus pool and I won that match and got all 10 points. So, what am I doing wrong? I really want to use this tool. PS: the memory reader is not available for Mac OS, otherwise I guess, there would already be a 100% game.
The problem here is that game start detection is not supported for Mac. It uses Win registry to check the state of the game. --> No pre-game profiles are available when using Mac.
Pre-requisite to get a potential 'good game' when pre-game profiles are not available is that the opponent wins and does have bonus pool left after the match (the 'after-check' line you were wondering meant this).
Read more from the original post. Search for: "Question: What is a pre-requisite to be able get a 'good game' (a game from which user's MMR can be calculated). What is the difference if memory reader is on vs off?"
|
Ah, OK thanks... so as it look like in the QA, I need a game, where I lose and the opponent has bonus points.
There is no difference turning on the memory reader since I am using a MacOS and the memory reader only supports Windows.
Damn I must be unlucky... I shouldn't win that many games xD
|
On January 08 2014 15:43 nitai wrote:
Ah, OK thanks... so as it look like in the QA, I need a game, where I lose and the opponent has bonus points.
There is no difference turning on the memory reader since I am using a MacOS and the memory reader only supports Windows.
Damn I must be unlucky... I shouldn't win that many games xD
Yes. Memory reader would not make the difference as it would even be automatically disabled, if you would 'enable' it when using Mac. But that section also gave concrete info regarding your situation. It is shame that Blizzard hid the change points for lost matches one month before HotS was published. Before that you got good readings much much more often when having memory reader disabled than now.
|
It says the file is privite , how i DL it?
|
I swear this was made to crush my dreams...
"Wow! I feel so close to master league! Lets look at the graph....oh..ok "
Good tool though, really does help measure progress instead of just playing and hoping :D
|
On January 08 2014 22:43 NoBuildOrder wrote:
It says the file is privite , how i DL it?
Between Dec 18 and Jan 3 the MMR tool server was down and manual download of DB file was required. But now as the server is back up, the tool downloads the DB file automatically (unless blocked e.g. by firewall). As manual installation is now unnecessary the DB file mirrors have been set private (they were older versions anyway).
The tool itself (version 10.0) is available from the link in the original post http://skeletor.jimmeh.com/mmr/MMRPlugin.zip It is also available from the mirror: http://www.mediafire.com/download/jgemp4zvqpar25w/MMRPlugin_10_0.zip
On January 08 2014 22:52 NorthQuab wrote:I swear this was made to crush my dreams... "Wow! I feel so close to master league! Lets look at the graph....oh..ok " Good tool though, really does help measure progress instead of just playing and hoping :D
Yes. Blizzard's ladder ranks & points often deceive you, if you try to estimate your MMR only based on them. In the end of the ladder season many of the top positions in different divisions are often taken by people who have faced MMR decay and are actually playing with lower league MMR.
Just remember to turn the tool on (launch sc2gears) every time before you start your laddering session. Missing games cause that estimates are less accurate as the tool expects that all matches are recorded (next 'good game' will of course fix the value from that game forward).
|
Hi guys,
i kinda played some games and everytime i get the error below. I play on a MacBook Pro (no VM), SC2Gears + MMR Plugin installed and woerking. Any hints how to solve it and get my MMR? Is it even possible to get it on Mac (without Memory Reader enabled, as I understand it doesn't work on Mac)?
1:29:49 Post-trigger: Start
01:30:29 Post-trigger: Reading web profiles
01:30:32 Webgrabber: Player true
01:30:36 Webgrabber: Opponent true
01:30:36 Post-trigger: Done reading web profiles
01:30:36 Bonus calc (xxx diam-): max end: 99, max fetch: 99
01:30:36 Bonus calc (yyy diam-): max end: 99, max fetch: 99
01:30:36 Calculating adjusted points for the players (either Alg. A or B needs to succceed for certain player)
01:30:36 DATA Alg. A : failed, No pre-game data (RAM scanner possibly off) or pre-game data bad:
01:30:36 DATA Alg.B: failed, No Bonus Win: http://us.battle.net/sc2/en/profile/xxx
01:30:36 ERROR AFTERCHECK Both alg. A and B failed: http://us.battle.net/sc2/en/profile/xxx
01:30:36 DATA Alg. A : failed, No pre-game data (RAM scanner possibly off) or pre-game data bad:
01:30:36 DATA Alg.B: failed, Unknown change points for player: http://us.battle.net/sc2/en/profile/yyy
01:30:36 ERROR AFTERCHECK Both alg. A and B failed: http://us.battle.net/sc2/en/profile/yyy
01:30:36 ERROR LASTCHECK : Both players invalid
01:30:36 Gamecheck: false
01:30:36 POST-Trigger done
01:30:36 Data quality needs to be 100% to be able to calculate MMR
01:30:36 Data quality for calculating Player's MMR: 50%
01:30:36 Data quality for calculating Opponent's MMR: 50%
01:30:37 OK Uploader: game uploaded
01:30:37 Ladder Game Done!
|
On January 09 2014 08:45 Samovarul wrote:Hi guys, i kinda played some games and everytime i get the error below. I play on a MacBook Pro (no VM), SC2Gears + MMR Plugin installed and woerking. Any hints how to solve it and get my MMR? Is it even possible to get it on Mac (without Memory Reader enabled, as I understand it doesn't work on Mac)? + Show Spoiler +1:29:49 Post-trigger: Start 01:30:29 Post-trigger: Reading web profiles 01:30:32 Webgrabber: Player true 01:30:36 Webgrabber: Opponent true 01:30:36 Post-trigger: Done reading web profiles 01:30:36 Bonus calc (xxx diam-): max end: 99, max fetch: 99 01:30:36 Bonus calc (yyy diam-): max end: 99, max fetch: 99 01:30:36 Calculating adjusted points for the players (either Alg. A or B needs to succceed for certain player) 01:30:36 DATA Alg. A : failed, No pre-game data (RAM scanner possibly off) or pre-game data bad: 01:30:36 DATA Alg.B: failed, No Bonus Win: http://us.battle.net/sc2/en/profile/xxx01:30:36 ERROR AFTERCHECK Both alg. A and B failed: http://us.battle.net/sc2/en/profile/xxx01:30:36 DATA Alg. A : failed, No pre-game data (RAM scanner possibly off) or pre-game data bad: 01:30:36 DATA Alg.B: failed, Unknown change points for player: http://us.battle.net/sc2/en/profile/yyy01:30:36 ERROR AFTERCHECK Both alg. A and B failed: http://us.battle.net/sc2/en/profile/yyy01:30:36 ERROR LASTCHECK : Both players invalid 01:30:36 Gamecheck: false 01:30:36 POST-Trigger done 01:30:36 Data quality needs to be 100% to be able to calculate MMR 01:30:36 Data quality for calculating Player's MMR: 50% 01:30:36 Data quality for calculating Opponent's MMR: 50% 01:30:37 OK Uploader: game uploaded 01:30:37 Ladder Game Done!
You get the same answer as the previous poster (It is possible to get good data if the opponent wins & has more than 0 bonus pool in the end of the match):
On January 08 2014 07:15 korona wrote:
The problem here is that game start detection is not supported for Mac. It uses Win registry to check the state of the game. --> No pre-game profiles are available when using Mac.
Pre-requisite to get a potential 'good game' when pre-game profiles are not available is that the opponent wins and does have bonus pool left after the match (the 'after-check' line you were wondering meant this).
Read more from the original post. Search for: "Question: What is a pre-requisite to be able get a 'good game' (a game from which user's MMR can be calculated). What is the difference if memory reader is on vs off?"
And here is that Question & Answer from the original post:
Question: What is a pre-requisite to be able get a 'good game' (a game from which user's MMR can be calculated). What is the difference if memory reader is on vs off?
+ Show Spoiler +Answer: Player's MMR is calculated from opponent's data. If the web profiles are accessible *) then pre-requisite conditions are following:
Memscanner
Condition: On: Off:
Opponent loses and has 0 bonus pool left Yes No
Opponent loses and has > 0 bonus pool left Yes No
Opponent wins and has 0 bonus pool left Yes No
Opponent wins and has > 0 bonus pool left Yes Yes
The pre-requisite condition is always met for counting the opponent MMR as player's pre-game data is always fetched based on the profile URL (except not for Mac). Please note that even if the pre-requisites are met there are lots of other requirements that have to be met to be able to calculate the MMR. For example if your league is lower than master league & your MMR is in low end of that league then you are likely in so called 'capping zone' where your MMR usually cannot be calculated from your typical opponent's data (exception opponent's from other leagues). For those who have used the tool before WoL patch 2.0.4 the old pre-requisites were (after that Blizzard hid the true change points when a player loses and thus in those cases the change points can only be calculated if pre-game data is available): + Show Spoiler +
Memscanner
Condition: Scanner On: Off:
Opponent loses and has 0 bonus pool left Yes Yes
Opponent loses and has > 0 bonus pool left Yes Yes
Opponent wins and has 0 bonus pool left Yes No
Opponent wins and has > 0 bonus pool left Yes Yes
*) Also worth noting is that web profiles are not always updated promptly or there are other problems regarding the profiles. Thus if the tool detects there is something wrong in the fetched profile data it marks those games bad.
|
Ok, thanks for the answer. So, basically, my only chance - as you said - is that the opponent wins & has more than 0 bonus pool in the end of the match.
Joking: if I become invincible I will never find my MMR...  )
|
On January 09 2014 09:31 Samovarul wrote:Ok, thanks for the answer. So, basically, my only chance - as you said - is that the opponent wins & has more than 0 bonus pool in the end of the match. Joking: if I become invincible I will never find my MMR... )
Correct. You can read this to learn more about other requirements: http://www.teamliquid.net/forum/viewpost.php?post_id=19316583
And before you become 'invincible' make sure that you have got one 'good game' (game from which your MMR could be calculated) Then the tool will use value from that match to estimate your MMR for other matches (estimates are based on typical MMR change for a match).
|
![[image loading]](http://i.imgur.com/PsoKfKk.png)
gm eu = 2060 (last win = promote)
|
On January 10 2014 09:50 Val_ wrote:+ Show Spoiler +gm eu = 2060 (last win = promote)
EU GM border indeed seems to be about the same as last season (2070). You played your placement 45 minutes after the GM opened. Another user played his GM placement 15 min after GM opened and got about the same value. The initial value might have been little bit lower thought.
Cool thing is that we have first time KR GM who was promoted just 10 minutes after the KR GM opened (KR GMs for older seasons using the tool were promoted mid-season. Last season there was no KR GM users). His promotion value was ~1873. After checking many of his opponents from the following matches, it seems KR low border was indeed around 1860 to 1870. So I likely have had too high GM border for Korea for the previous seasons too (I have used EU value for them as there was no Korean data).
---
Edit: Based on the user graphs it indeed seems that there were no offset changes in beginning of the S17. This is only based on visual inspection of user profiles (especially checking those who had been promoted and had 'good games' against people of multiple leagues).
I also checked the MMR decay values from last season (final data from Dec 18th when the MMR tool server went down). There were 3 users that had been inactive for 4 weeks and had been decayed (there were many more, but I ignored all who did not have 'good games' on both sides of the inactivity period). But their decay value was between ~200 to 260. But as there were several decayed users who had been inactive for 3 weeks and had expected decay values (about half of the max decay of ~300), it is likely that the earlier 3 had actually been inactive for less than 4 weeks (e.g. played unranked and thus not faced max decay). Based on this Blizzard had not silently changed the decay value be worth of "few losses" like they claimed it to be in their recent article.
--
Edit2: Have had all time low amount of NA GMs using the tool. Only one has played after NA GM opening (data taken 6h 30min after the opening). And he has played only one match - his placement match (about 1h 10min after the NA GM opening). His placement MMR was about the same as last season's initial GM placement border. Also only 2 NA master users have faced GM:s.
On November 19 2013 05:10 korona wrote:Show nested quote +On November 12 2013 22:14 korona wrote:
But as reference (it is interesting to see if trend continues):
GM border for EU at start of S14 was 1950, it had increased to 2130 by start of S15.
GM border for NA at start of S14 was 1900 (or slightly lower), it had decreased to 1830 by start of S15.
S16 EU border was ~2070 when GM opened S16 NA border was ~1800 when GM opened
I will likely use following values for S17 (these will be later via publishing a new DB file (not today)):
S17 EU 2050
S17 NA 1780
S17 KR 1860
|
|
|
I am not sure what you mean? Based on the uploaded data you have had 11 good games (out of 98) this season. But 2 of them seem to suffer from data error (two upward spikes). First happened during the first days, when web servers were slow. Second last night.
To show good games on graph (color coding):
'Config' > 'Data' > 'Mark assumptions / bad games'. Remember to save the changes.
Also when hovering cursor over match dots it tells you the quality: good (calculated MMR), estimate (estimates based on good games), unknown (unknown opponent values), harcoded (only for archived games)
|
I just published a new DB file (ver. 17.2). You can get it by restarting sc2gears.
It changes the initial GM entry barriers for EU, NA and KR for this season that were mentioned here: http://www.teamliquid.net/forum/viewpost.php?post_id=20565355
It also sets more accurate bonus pool accumulation start times for each server (less data errors when rare occasion that match ends just when bonus pool tick happens). For other servers changes were small, but for China ~20 seconds. Especially the Chinese web server was extra sluggish in the start of the season, but now it seems to respond fast.
|
Thanks for a quick reply. I misinterpreted the graph. I'm still confused why none of the last games has a "good" prediction of opponents MMR.
Anyway, the last good game is slightly higher than the previous good games. How likely is it to be accurate, really? 9 out of last 10 "recent" good games are in similar range, it's just the latest one that doesn't seem to be "correct".
Cheers!
|
|
|
|
|
|
EPT
