http://imgur.com/CVvjH
Match Making Rating Tool - Page 98
Forum Index > SC2 General |
shadogi
United States194 Posts
http://imgur.com/CVvjH | ||
skeldark
Germany2223 Posts
On October 24 2012 06:45 shadogi wrote: Why did I lose MMR for winning? I've also noticed that sometimes when I win against someone with a similar MMR, I will get 15-20 points, but other times it will give me 8 or less. What's up? http://imgur.com/CVvjH Because the program is not accurate to the last point! Every game have a errormargin. Often its +-12 points sometimes way more. I try to correct it as far as i can with the given information. Technical: Case A: Last game is a good game: Game 1 Was to high. Game 2 can not be corrected higher than 1 because Game 3 would tell otherwise. In this case you can see the fall after a win as correction of an overall to high value. Case B: Last game was a bad game. The correction function will wait for the next good game to correct the fall after a win. If next good game allow it, he will raise the marked game and all games between this one and the good game above the game before the Fall. If next good game dont allow it, he will not correct it because in this case a correction would create follow up mistakes. See Case A. For you its a mistake in the data, for me its a success that you complain about +-10 points ^^ See it this way:10 points wrong at 1500 points is more than 99% accurate And even here its possible the correction algorithm is still waiting for data and will correct it with next good game. | ||
korona
1098 Posts
EU server was supposed to be locked on Wednesday, October 24 at 23:59 p.m. CEST: http://eu.battle.net/sc2/en/blog/6298734/2012_Season_4_Lock_Incoming_-18_10_2012 US server was supposed to be locked at same time: Wednesday, October 24 at 23:59 p.m. CEST: http://us.battle.net/sc2/en/blog/7634960/2012_Season_4_Lock_Incoming_-10_18_2012 I monitored what happened to bonus pools during the night (EU time). EU bonus pools indeed stopped growing both in-game & website during the night. Max bonus pool for master+ seems to be 552 and for diamond- 320. But US bonus pools did not stop growing at that time (kept rising both on client & website). But 30 minutes ago 11:59 p.m PDT a news article was posted that US server would now be locked: http://us.battle.net/sc2/en/blog/7681241/2012_Season_4_Now_Locked-10_24_2012 If the US server was locked just 30 minutes ago (writing time, not posting) the max pool for master+ seems to be 561 and for diamond- 325. It has to be later confirmed if the bonus pools indeed stopped growing both on client & website on that time. Edit: Now 3 hours has passed since the 'locked' post and bonus pools have not increased both on client & website. --> I guess NA had clean lock too (except the originally announced time was incorrect (not 11:59 p.m. CEST but 11:59 p.m. PDT). | ||
skeldark
Germany2223 Posts
On October 25 2012 16:52 korona wrote: Ladders seem to be now locked at least on some servers. EU server was supposed to be locked on Wednesday, October 24 at 23:59 p.m. CEST: http://eu.battle.net/sc2/en/blog/6298734/2012_Season_4_Lock_Incoming_-18_10_2012 US server was supposed to be locked at same time: Wednesday, October 24 at 23:59 p.m. CEST: http://us.battle.net/sc2/en/blog/7634960/2012_Season_4_Lock_Incoming_-10_18_2012 I monitored what happened to bonus pools during the night (EU time). EU bonus pools indeed stopped growing both in-game & website during the night. Max bonus pool for master+ seems to be 552 and for diamond- 320. But US bonus pools did not stop growing at that time (kept rising both on client & website). But 30 minutes ago 11:59 p.m PDT a news article was posted that US server would now be locked: http://us.battle.net/sc2/en/blog/7681241/2012_Season_4_Now_Locked-10_24_2012 If the US server was locked just 30 minutes ago (writing time, not posting) the max pool for master+ seems to be 561 and for diamond- 325. It has to be later confirmed if the bonus pools indeed stopped growing both on client & website on that time. Edit: Now 3 hours has passed since the 'locked' post and bonus pools have not increased both on client & website. --> I guess NA had clean lock too (except the originally announced time was incorrect (not 11:59 p.m. CEST but 11:59 p.m. PDT). THx EU 552 US 561 I check values for other servers and update the datafile. If something is wrong again this season i will lock the client until next season starts | ||
skeldark
Germany2223 Posts
| ||
skeldark
Germany2223 Posts
Sea 551 CN 551 NA 561 EU 552 New Bnetdatabse File in 5 min. EDIT: New Bnetfile up. Restart the client to download it. Report possible mistakes in new collected data here. PS: sorry for tripe post something is wrong with the update... | ||
Nildawenn
United States9 Posts
hopefully this helps in some way. ![]() | ||
skeldark
Germany2223 Posts
Will calculate later per hand and check some graph if its only you who have trouble Mark the last game of the one high as bad (61 i think, hard to see on pic) | ||
Nildawenn
United States9 Posts
| ||
ThreeHeartRun
United States6 Posts
| ||
skeldark
Germany2223 Posts
On October 27 2012 15:28 ThreeHeartRun wrote: May I ask why the MMR tool assumes that you're playing opponents 10 mmr above you if there isn't data suggesting otherwise? Wouldn't it be more accurate to take the average of surrounding data (in the case of past games) or projecting the "slope" of opponents you've been facing (in the case of no more recent games)? Maybe I'm missing something with how matchmaking varies? To make it clear its an lie. If i put in an value that would make sense, you would think thats his real mmr. If i show you wrong data to make it look better, where is the diffrence between me and blizzard? I have to put a value in there or i get problems with the graph. Beside that the missing data is biased. Opponents who are weaker than you have a way higher chance to produce bad data. So the average opponent mmr out of the good games is to high. | ||
RobbyG
United Kingdom10 Posts
| ||
InfCereal
Canada1759 Posts
On October 28 2012 05:02 RobbyG wrote: I would like to ask you about the league boundaries; how accurate are they?, cause right now im 150 pts above my line, and hope for next season, but at the start of the last season I was 80 pts above it and did not promote me. I dont wanna criticise, just want to find out. The lines are currently inaccurate for a certain range. (I'm not sure what that range is). Skeldark's been busy lately, and has the lines sitting at "zero". They're still a pretty good indicator of where you stand, and should be fixed eventually. | ||
skeldark
Germany2223 Posts
On October 28 2012 05:02 RobbyG wrote: I would like to ask you about the league boundaries; how accurate are they?, cause right now im 150 pts above my line, and hope for next season, but at the start of the last season I was 80 pts above it and did not promote me. I dont wanna criticise, just want to find out. depends on the line. master is accurate. Diamand ok. Rest i cant really tell. Dont forget blizzard dont promote if you get over the line, They promote if you are stable over the line. What is stable for blizzard? no one knows... Promotion / demotion follow a lot of rules and are inaccurate them-self. So its hard to be accurate on something inaccurate. On October 28 2012 05:08 InfCereal wrote: The lines are currently inaccurate for a certain range. (I'm not sure what that range is). Skeldark's been busy lately, and has the lines sitting at "zero". They're still a pretty good indicator of where you stand, and should be fixed eventually. To hunt down an inaccurate line is kind of against the whole idea of this program. I dont do all the work to be accurate just to guess when an inaccurate icon change. So it has low priority on my todo list. I kind of hoped people will go for the numbers i show and forget the whole fake "league icon" thing. Looks like this will not happen ![]() EDIT:is the data of the lock ok for your InfCereal? Can it be that this is the first league lock without big bugs? | ||
leopardb
Germany92 Posts
![]() I've been switching to Zerg lately and through learning the mechanics and all, lost quite a lot, excuses, excuses, blabla. ![]() I just wanted to show my MMR graph since it looks weird that every 2-3 games i'm pitted against somebody way higher than i am, and belonging to my old league so to speak : i should have been already demoted to silver but, because of the ladder lock, stayed in gold. Now the high opponents begin to appear exactly after the ladder lock (i looked at the dates, and the first "high" dot belong to 26-10-2012, the other ones are before...). Hope it helps improving your plugin ! Thx again for all :D EDIT: updated the image to make my point clearer (hopefully) | ||
RobbyG
United Kingdom10 Posts
To hunt down an inaccurate line is kind of against the whole idea of this program. I dont do all the work to be accurate just to guess when an inaccurate icon change. So it has low priority on my todo list. I kind of hoped people will go for the numbers i show and forget the whole fake "league icon" thing. Looks like this will not happen ![]() EDIT:is the data of the lock ok for your InfCereal? Can it be that this is the first league lock without big bugs? I was using the tool only for comparing my rating with my previous ratings anyway, not even comparing it with other people's mmr, because I think it is good only for relative mmr; I was just curious, cause my goal is to go up one league at the begining of next season so if something has anything to do with finding out if I will, I thought I might ask about that something. L.E. I had a huge drop and rise of my graph over 2 games, like 500 mmr difference, I tried marking the games as bad as you suggested, but it just made it worse, I don't really care about that gap cause I've already had a few and it comes back to normal after a few games, just wondering if the mark the game bad thing should work? | ||
skeldark
Germany2223 Posts
you opponent is not really there. looks like not detected cap for me @RobbyG select "mark assumptions" in the settings. Than mark the Good Game in the wrong data field as bad. i think something went wrong with the lock again ![]() Anyway: only few days to the new season. Hope they are lazy again and dont change the offsets. I think about an change in opponent MMR. The fake +10 value is confusing for most. I try to find a away to display better that i dont have data for the opponent in this games. | ||
korona
1098 Posts
On October 30 2012 00:29 skeldark wrote: @leopardb you opponent is not really there. looks like not detected cap for me @RobbyG select "mark assumptions" in the settings. Than mark the Good Game in the wrong data field as bad. i think something went wrong with the lock again ![]() If you mean leopardb:s example, then that has been happening to me on my both accounts the whole season - not only during the lock. When opponent MMR could have been calculated it has almost always been hovering around 100 MMR points over league border regardless of my own MMR (sometimes differs a lot). I was going to post about this after the season. But with a quick check the lock period data seemed to be correct (at least all the adjusted points were calculated correctly that I checked). On October 30 2012 00:29 skeldark wrote: Anyway: only few days to the new season. Hope they are lazy again and dont change the offsets. I think about an change in opponent MMR. The fake +10 value is confusing for most. I try to find a away to display better that i dont have data for the opponent in this games. Or you could just force "mark assumptions" on as a default. Then it would be easy to deduce that they are dummy values. Edit: Checked personal data from games during the lock period (54 NA games, 2 EU games) and all data seemed correct. Did not check opponent data. | ||
skeldark
Germany2223 Posts
On October 30 2012 05:21 korona wrote: If you mean leopardb:s example, then that has been happening to me on my both accounts the whole season - not only during the lock. When opponent MMR could have been calculated it has almost always been hovering around 100 MMR points over league border regardless of my own MMR (sometimes differs a lot). I was going to post about this after the season. But with a quick check the lock period data seemed to be correct (at least all the adjusted points were calculated correctly that I checked). Or you could just force "mark assumptions" on as a default. Then it would be easy to deduce that they are dummy values. Edit: Checked personal data from games during the lock period (54 NA games, 2 EU games) and all data seemed correct. Did not check opponent data. Nice thanks. Was afraid something is messed up again. I will have a look if the cap-finder for the opponent have an bug. Forcing "mark assumption" looks a little bit ugly. If i dont show the dots the graph can get crazy on some filter settings. Its way easyer to work with complete data than with holes. Perhaps i just force assumptions for opponents and let the setting decide for the player. If they change the offsets for next season i have to write an new archive updater to save the old data correct ![]() EDIT: No its not a caped problem. Other way round! The point is that the once 100mmr below you are caped, so you only see the once that are way stronger than you. If i could calculate all the opponent data the graph would be all over the place. So the problem is not that he shows them to high, the problem is that he can only show the very high once. | ||
kromlic
2 Posts
+ Show Spoiler + 17:22:32 MMR-Stats 7.0 17:22:32 --------INIT START-------- 17:22:32 OK Encoding: OK 17:22:32 OK Loading DB 17:22:53 FATALERROR Versionchecker: failed 17:22:53 OK DB up to date! 17:22:53 OK Settings loaded! 17:22:53 OK Timeclient: Systemtime offset: 3 sec 17:22:53 OK Gamedata: 16 Games loaded. 17:22:53 Testing MemoryScanner ...OK 17:22:53 OK Triggers ready! 17:22:53 OK Gameobserver: ready! 17:23:14 FATALERROR Versionchecker: failed 17:23:14 OK: Program is up to date 17:23:14 --------INIT END-------- | ||
| ||