EPTMatch Making Rating Tool - Page 173
| Forum Index > SC2 General |
|
korona
1098 Posts
| ||
|
Kratos13
United States12 Posts
| ||
|
korona
1098 Posts
On November 22 2013 14:53 Kratos13 wrote: I can't get my sc2gears to run :/ I tried plugging in the MMR thing and everything but everytime I try to open sc2gears it wont open. My OS is Windows 7 So you have some kind of sc2gears setup problem? Sc2gears requires Java 7. Make sure you have installed it. Java is available from http://www.java.com/en/download/index.jsp Make sure you have write rights in all sc2gears folders. If you don't have (e.g. you installed it under 'program files'), then you need to run it in administration mode. More instructions how to setup Sc2gears can be found from its website: https://sites.google.com/site/sc2gears/ Also remember that the MMR plugin requires that the auto-save feature of Sc2gears has been enabled & configured correctly. | ||
|
jsp
United States2 Posts
| ||
|
korona
1098 Posts
On November 22 2013 20:05 jsp wrote: The version number isn't bumped in the 10.0 release. It still shows 9.6 It is changed to 10.0 as supposed to. Are you sure you have updated the tool? Also quite sure tell is also the new filters at the bottom bar (these were not available in 9.6). Such as different markers & sub-graph selection. These are described more thoroughly in original post. | ||
|
jsp
United States2 Posts
On November 22 2013 20:09 korona wrote: It is changed to 10.0 as supposed to. Are you sure you have updated the tool? Also quite sure tell is also the new filters at the bottom bar (these were not available in 9.6). Such as different markers & sub-graph selection. These are described more thoroughly in original post. I was checking the xml file to make sure I updated to the latest version. The release date is 11/3, but the version is still 9.6 | ||
|
korona
1098 Posts
On November 23 2013 16:09 jsp wrote: I was checking the xml file to make sure I updated to the latest version. The release date is 11/3, but the version is still 9.6 Oh, I seem to have only updated the date on the XML file. I wonder if I have accidentally pressed undo before saving that file. But good news is that version number in the XML file does not matter as it is not used for anything except what is shown in sc2gears plugin manager. You can always check what the internal version of the tool is by opening the 'log' from the top toolbar. In the beginning of the 'main log' the internal version number is shown. Edit: It seems Blizzard has changed the web profiles little bit. They have at least re-added season numbers back. This is good as API was not often in sync with web profiles in the start of the season. The MMR tool parser seems unaffected by the changes (it primarily uses season number from the web page. If it not available, then it tries to read it from the API). Shadowed of sc2ranks had noticed the changes first: http://www.teamliquid.net/forum/viewpost.php?post_id=20281154 | ||
|
shenlong
219 Posts
| ||
|
korona
1098 Posts
On November 24 2013 10:05 shenlong wrote: hey guy i'm having a problem, it keep saying that "Needs a game with better data 0/9" i don't see a error message, do i have to play more games for this to start working? pls help Correct. MMR cannot be reliably calculated from all matches. Read more info e.g. from http://www.teamliquid.net/forum/viewmessage.php?topic_id=334561¤tpage=162#3226 Play more games and it time you will get a 'good game' from which your MMR is determined. | ||
|
donjoe
2 Posts
| ||
|
THF
United Kingdom20 Posts
On November 24 2013 23:04 donjoe wrote: I've downloaded SC2Gears v14.3.2 and the MMRPlugin v9.6, put the latter in the Plugins folder, have it showing up as compatible, but when I try to Enable it I just get 3 new empty windows (for plugin config and help and stuff) and the whole SC2Gears interface remains pretty much frozen: I can't close the new windows (though I can move them around) or the main program window and all I'm left with is to terminate the process forcefully. This is on Linux. What do I do to diagnose/fix this? Install sc2gears under wine. You'll need windows java too. | ||
|
donjoe
2 Posts
Anyway, I don't see what I'm supposed to gain that way. SC2Gears is a Java program and Linux has its own Java JRE (and I do have it installed) - that's what I use to launch SC2Gears and that part works. My problem is with enabling the MMR plugin, not with starting SC2Gears itself. | ||
|
korona
1098 Posts
On November 26 2013 05:42 donjoe wrote: Tried it, but I can't even get it to start like that - it says it can't create the Java Virtual Machine because it doesn't have enough memory (LOL, 2 GiB of RAM should be enough). Anyway, I don't see what I'm supposed to gain that way. SC2Gears is a Java program and Linux has its own Java JRE (and I do have it installed) - that's what I use to launch SC2Gears and that part works. My problem is with enabling the MMR plugin, not with starting SC2Gears itself. Actually you gain from running it under wine. Then game launch detection & memory reader will work. If you run it directly under Linux, the game launch detection does not work (it uses windows registry). And memory reader (only windows support) is only launched if game launch is detected & user has enabled it. But as I do not have Linux setup available, I cannot test what the problem is. First make sure you have write rights under all sc2gears directories when running it. Second make sure you have the latest version of the MMR tool (version 10.0, it is incorrectly marked as 9.6 in sc2gears plugin manager. The actual 9.6 cannot anymore fetch DB file from MMR tool server, because certain part of the code stopped working after Oracle released recent update to JVM. That code was replaced in 10.0). Third if it is possible to open 'View system messages' from the 'Help' menu of sc2gears check if there are MMR tool related exceptions (If the MMR tool crashes the reason is likely listed there). | ||
|
shenlong
219 Posts
On November 24 2013 16:58 korona wrote: Correct. MMR cannot be reliably calculated from all matches. Read more info e.g. from http://www.teamliquid.net/forum/viewmessage.php?topic_id=334561¤tpage=162#3226 Play more games and it time you will get a 'good game' from which your MMR is determined. Thank you so much for the feedback Korona!  You were right, it works perfectly fine now! | ||
|
fried_rice
198 Posts
Are these league boundaries accurate at all for the current season? Or I have to be consistently above 1360 to be promoted? Also what is the "autocorrect" function for? I don't quite get what it's supposed to do. Other than that, thanks for your awesome work korona. | ||
|
corydoras
161 Posts
About two weeks my MMR according to the tool reached 500+. After a couple of days, the same measurements (i.e. the same dots on the chart) show values that are smaller by about 50. I do not have any screenshots to compare the charts, unfortunately. This is not a result of MMR decay, because 1) this season I did not have breaks longer than 3-4 days b) the trend of my MMR is actually upward. Is this a result of getting more "good games" so the results plotted on the chart are actually being recalculated retroactively? Or am I missing something? | ||
|
korona
1098 Posts
On November 26 2013 18:11 fried_rice wrote: So the current mininum-MMR for Master League shows up as 1360 (off of "good" games), I did break this number on a couple of occasions and just now I beat an opponent of 1560 MMR, which once again put me above 1360, yet still no promotion. Are these league boundaries accurate at all for the current season? Or I have to be consistently above 1360 to be promoted? Also what is the "autocorrect" function for? I don't quite get what it's supposed to do. Other than that, thanks for your awesome work korona. Calculations from 'good games' have a certain error margin that is visible in the 'Analyse log'. Also results from losses often give slightly higher values than results from wins (the actual value is often likely somewhere between). Also even if it seems there was no offset changes between last & this season I have not yet confirmed it via calculations. The most user graphs seemed fine thought also regarding promotions (--> at least no larger changes). If you hover near the promotion line (several good games), you are likely very close to be promoted. If 'autocorrect' is on and there is some changes in the graph that are much larger than the error margin, the tool can automatically 'fix' it in some cases (considers some of the games bad). It does not change the data though. If you disable it, those jumps will once again be shown in the graph. I personally prefer to keep it off, as I am then able to notice values that are due potential data errors. On November 26 2013 18:35 corydoras wrote: I'm a silver on EU trying to reach gold in these troubled ladder times About two weeks my MMR according to the tool reached 500+. After a couple of days, the same measurements (i.e. the same dots on the chart) show values that are smaller by about 50. I do not have any screenshots to compare the charts, unfortunately. This is not a result of MMR decay, because 1) this season I did not have breaks longer than 3-4 days b) the trend of my MMR is actually upward. Is this a result of getting more "good games" so the results plotted on the chart are actually being recalculated retroactively? Or am I missing something?Your MMR cannot be calculated reliably from all games. When you play a new 'good game' (a game from which the MMR can be calculated), it adjusts all preceding estimate games in the graph. For example if uppercase X and Y are 'good games', then the affected estimate games are shown in lowercase x & y (Note that 'Y' is the last 'good game' you have played):
Also note that if you have enabled fine-tune from the settings (it is enabled by default), it fine-tunes the values in range of the error margins. To see which games are 'good games' you can either hover your mouse over the 'match dots' or switch following setting on: 'Config' > 'Data' > 'Mark assumptions / bad games'. MMR values for the estimate games are only estimates based on the 'good games' (typical MMR change for a match which is +16 for a win and -16 for a loss is applied to them). The estimates are often quite accurate too, even if lots of matches might have passed since last 'good game', but can also be somewhat inaccurate (e.g. like the 50 points transition you noticed when new 'good game' was played). And if you have missing games (not recorded) in between of estimate games, the estimate error increases. | ||
|
corydoras
161 Posts
On November 26 2013 19:05 korona wrote: Your MMR cannot be calculated reliably from all games. When you play a new 'good game' (a game from which the MMR can be calculated), it adjusts all preceding estimate games in the graph. For example if uppercase X and Y are 'good games', then the affected estimate games are shown in lowercase x & y (Note that 'Y' is the last 'good game' you have played):
Thanks! It turns out I underestimated your tool. Now I know it's even better I thought good games were affecting only future estimates. Eh, and I thought I was so close to the magic number between silver and gold. My wife is going to kill me if I keep laddering every evening... | ||
|
Mahavishnu
Canada396 Posts
| ||
|
korona
1098 Posts
On November 28 2013 08:06 Mahavishnu wrote: Does the mmr estimation often need 50 games? Or is this rare. May there be a configuration error? jw Sometimes if you are in such MMR range (low range of your league), where most of your opponents are capped (and you are consistent so your MMR does not drop enough to face uncapped opponents from lower league nor increase enough to face uncapped opponents from your league). I think I found your profile. Based on it you also have memory reader also enabled. You just need to play more and hope to get uncapped opponent. But when you move to such range that you get a 'good game', you are likely to get several of them. Read more: http://www.teamliquid.net/forum/viewmessage.php?topic_id=334561¤tpage=162#3226 | ||
| ||
