|
Thanks for the work on the MMR calculator! I've noticed a couple of issues which will hopefully help you make it better.
On June 04 2012 03:05 skeldark wrote: This is one fucked up graph. Tells me that the offsets for bronze silver are still wrong. They are the hardest to get. i will use your data to improve the algorithem. so big jumps should never happen. really have to look into them Also i think notthat have new offsets for eu. will update them over the db update as soon as he is online. I do think the analysers have some problems, maybe it's just the lower leagues as you've said - I'm a Bronze player. I included three graphs to show you what I mean.
Here is Analyser A from a previous version of the MMR Calculator. Look how at the end, despite my winning streak, the MMR fluctuates wildly. I'm not sure how the algorithms work, but in my second-to-last game I played against a high Bronze opponent, and the last game with a very low Bronze opponent - so maybe that confused the analyser in a way.
Here is the same graph with Analyser A&B from the latest MMR Calculator:
Looks much more reasonable.
However, Analyser A&D+ gives this:
Now, my MMR has definitely been on the rise recently as I have been paired with better opponents, so I am pretty sure the graph is wrong.
Ramscanner need 32 bit java. So you have to de install your 64 java and keep your 32 version. I check the java version and give out error an message in next patch.
Any way to make it work with 64 bit? It seems like 64 bit is the standard now, and some other apps actually require 64 bit Java to work on a 64 bit system.
|
On June 04 2012 08:26 sp313 wrote:Thanks for the work on the MMR calculator! I've noticed a couple of issues which will hopefully help you make it better. Show nested quote +On June 04 2012 03:05 skeldark wrote: This is one fucked up graph. Tells me that the offsets for bronze silver are still wrong. They are the hardest to get. i will use your data to improve the algorithem. so big jumps should never happen. really have to look into them Also i think notthat have new offsets for eu. will update them over the db update as soon as he is online. I do think the analysers have some problems, maybe it's just the lower leagues as you've said - I'm a Bronze player. I included three graphs to show you what I mean. Here is Analyser A from a previous version of the MMR Calculator. Look how at the end, despite my winning streak, the MMR fluctuates wildly. I'm not sure how the algorithms work, but in my second-to-last game I played against a high Bronze opponent, and the last game with a very low Bronze opponent - so maybe that confused the analyser in a way. Here is the same graph with Analyser A&B from the latest MMR Calculator: Looks much more reasonable. However, Analyser A&D+ gives this: Now, my MMR has definitely been on the rise recently as I have been paired with better opponents, so I am pretty sure the graph is wrong. Show nested quote +Ramscanner need 32 bit java. So you have to de install your 64 java and keep your 32 version. I check the java version and give out error an message in next patch.
Any way to make it work with 64 bit? It seems like 64 bit is the standard now, and some other apps actually require 64 bit Java to work on a 64 bit system. Thats the reason i did let B in. D++ filter caped games that have semi wrong information. However if you play a lot caped games he dont have enough data to work with. D++ focus on your point change and need good rows ( but he can work with row data he dont show in the graph ).
B workes with the data of the opponent and he dont filter caped guys. What gives him more data = better rows = better guess (but possible wrong ones because of the caped games) B will in avg rate you higher than d because caped guys show an upper border of mmr.
In future i will bring them both together. They both have advantages and disadvantages. EDIT: checked all graph of all players of the new version. Yours is the only one with so different results. We have to check the bronze offsets...
About 64. The problem is i use a api for the memoryscan and this api dont support 64. To support it i have to write the hole memory read stuff on my own. Means in c not in java because java cant read memory. Its on the todo list but not at the top spots. About other application. Until today the rule:" for every thing that dont work with 32 on a 64 comp there are 10 things that dont work with 64, is still valid" I never install a 64 bit system. The only reason i can see to do so is to support more memory under windows. However its the new standard and i try to find a solution.
|
What do the different dots mean now. There's 3 different ones ..
|
Full blue - wins Full white - Game unusable for calculation or missing game(s) Blue with white center - losses
|
|
Changelog talks about 3.5 but download is for 3.4?
|
On June 05 2012 02:49 Garf wrote: Changelog talks about 3.5 but download is for 3.4? 3.5 in few minutes
|
On June 04 2012 09:29 skeldark wrote: About 64. The problem is i use a api for the memoryscan and this api dont support 64. To support it i have to write the hole memory read stuff on my own. Means in c not in java because java cant read memory. Its on the todo list but not at the top spots. About other application. Until today the rule:" for every thing that dont work with 32 on a 64 comp there are 10 things that dont work with 64, is still valid" I never install a 64 bit system. The only reason i can see to do so is to support more memory under windows. However its the new standard and i try to find a solution.
Cant you force Java to use the 32bit JDK if its installed? If you can ask for a new JVM for your plugin to run in, I can imagine you could give Java some paramaters like -java d32 or something like that. Different story of you have to use the JVM that Sc2gears is using.
|
On June 05 2012 03:53 Raven283 wrote:Show nested quote +On June 04 2012 09:29 skeldark wrote: About 64. The problem is i use a api for the memoryscan and this api dont support 64. To support it i have to write the hole memory read stuff on my own. Means in c not in java because java cant read memory. Its on the todo list but not at the top spots. About other application. Until today the rule:" for every thing that dont work with 32 on a 64 comp there are 10 things that dont work with 64, is still valid" I never install a 64 bit system. The only reason i can see to do so is to support more memory under windows. However its the new standard and i try to find a solution.
Cant you force Java to use the 32bit JDK if its installed? If you can ask for a new JVM for your plugin to run in, I can imagine you could give Java some paramaters like -java d32 or something like that. Different story of you have to use the JVM that Sc2gears is using. How should i not run in the vm of sc2gears if we want to interact with eachother?
you could try to start sc2gears as 32 bit but he installed a complex starting way. In the end its not a big deal to write a c prog that gives me the values but its years ago i programmed in c. if someone give me the source of a c prog that reads mem of a other program and output it in a file or over the system stream i would gladly use it. I think its not more than half a site code.
|
On June 05 2012 04:00 skeldark wrote: How should i not run in the vm of sc2gears if we want to interact with eachother?
you could try to start sc2gears as 32 bit but he installed a complex starting way. In the end its not a big deal to write a c prog that gives me the values but its years ago i programmed in c. if someone give me the source of a c prog that reads mem of a other program and output it in a file or over the system stream i would gladly use it. I think its not more than half a site code.
Didnt know if inter-vm communication was possible I have some Java experience but its not my main language
|
On June 05 2012 04:07 Raven283 wrote:Show nested quote +On June 05 2012 04:00 skeldark wrote: How should i not run in the vm of sc2gears if we want to interact with eachother?
you could try to start sc2gears as 32 bit but he installed a complex starting way. In the end its not a big deal to write a c prog that gives me the values but its years ago i programmed in c. if someone give me the source of a c prog that reads mem of a other program and output it in a file or over the system stream i would gladly use it. I think its not more than half a site code.
Didnt know if inter-vm communication was possible I have some Java experience but its not my main language inter vm communication is possible but both site must programmed for it. I dont see any problem of using java 32 bit btw. There is a lot of stuff that have problems with 64 but i never heard: "oh this dont work you need at 64 bit jvm..."
|
3.5 OUT! Important: 3.5 comes with a new Game-save-System. All your games are transformed automatic into the new format. But we found over the time a lot of special rules when data is not valid and some old version had bugs that could produce invalid data (mainly 3.0 ). So there can be bad data in your old games. You can delete all your games to have a clean start or manual delete broken games over the new Data-editor in the Config.
-Also i added an update function so you don't have to download and unpack future updates yourself!
Download: MMR-Calculator V.3.5
|
I don't know what's going on but ever since the V3.0 versions and on, I haven't been able to get this thing to work.
+ Show Spoiler + 16:14:29 MMR-Stats 3.5 16:14:29 --------INIT START-------- 16:14:29 OK Data-Settings loaded! 16:14:29 ERROR No Data Found! Time to ladder... 16:14:29 OK Database loaded! 16:14:29 DB Version:7.1 16:14:30 OK DB up to date! 16:14:30 OK Triggers ready! 16:14:30 OK Gameobserver: ready! 16:14:30 OK Encoding: UTF-8 is avalable on this System 16:14:31 OK: Program is up to date 16:14:31 --------INIT END-------- 16:22:46 -------------------------- 16:22:46 Game Start detected 16:22:52 OK Webgrabber: Pre-Player 16:22:52 OK Webgrabber: Pre-Opponent 16:22:52 PRE-Trigger done! 16:29:5 OK Webgrabber Player 16:29:5 OK Webgrabber Opponent 16:29:5 OK PRECHECK Pre-Player passed 16:29:5 DATA Alg.B: 2 : failed, No Bonus Win 16:29:5 ERROR AFTERCHECK 1: both alg. failed for opponent 16:29:5 OK AFTERCHECK Opponent passed 16:29:5 OK LASTCHECK game passed! 16:29:5 POST-Trigger done! 16:29:5 OK Good Game! 16:29:6 OK Uploader: game uploaded 16:29:6 Ladder Game Done! 16:29:6 --------------------------
I really want know how far I have until I get promoted because I have 750ish points in Gold league, but this plugin is letting me down.
|
On June 05 2012 05:33 SoymasterYOS wrote:I don't know what's going on but ever since the V3.0 versions and on, I haven't been able to get this thing to work. + Show Spoiler + 16:14:29 MMR-Stats 3.5 16:14:29 --------INIT START-------- 16:14:29 OK Data-Settings loaded! 16:14:29 ERROR No Data Found! Time to ladder... 16:14:29 OK Database loaded! 16:14:29 DB Version:7.1 16:14:30 OK DB up to date! 16:14:30 OK Triggers ready! 16:14:30 OK Gameobserver: ready! 16:14:30 OK Encoding: UTF-8 is avalable on this System 16:14:31 OK: Program is up to date 16:14:31 --------INIT END-------- 16:22:46 -------------------------- 16:22:46 Game Start detected 16:22:52 OK Webgrabber: Pre-Player 16:22:52 OK Webgrabber: Pre-Opponent 16:22:52 PRE-Trigger done! 16:29:5 OK Webgrabber Player 16:29:5 OK Webgrabber Opponent 16:29:5 OK PRECHECK Pre-Player passed 16:29:5 DATA Alg.B: 2 : failed, No Bonus Win 16:29:5 ERROR AFTERCHECK 1: both alg. failed for opponent 16:29:5 OK AFTERCHECK Opponent passed 16:29:5 OK LASTCHECK game passed! 16:29:5 POST-Trigger done! 16:29:5 OK Good Game! 16:29:6 OK Uploader: game uploaded 16:29:6 Ladder Game Done! 16:29:6 --------------------------
I really want know how far I have until I get promoted because I have 750ish points in Gold league, but this plugin is letting me down. I dont understand. Your log says everything is perfect. Where is the problem? he should show 1 dot for the opponent in no analyse. Cant say if the analysers will filter the game tho. THere is no data for you, because you lost and your opponent have no bonus pool. You can still get the data in this situation if you turn on memory scanner.
|
Could you add on the dot a tooltip saying date and hour of the game? Maybe even the name if you keep track of it. It may be useful to track down specific games.. sometimes I don't even know if my latests games have been recorded or were labeled as "bad", cause I can't remember at which "number" I was before starting to ladder.
Ah, another things I've noticed: sometimes I happen to play against opponents who don't belong to any division yet (first game of the season, or maybe first games ever). I usually get HUUUGE point gains against these (since they are probably considered way higher than they are by SC2). How does your software handle that? Should they just be discarded I suppose, or not?
|
On June 05 2012 07:01 Malhavoc wrote: Could you add on the dot a tooltip saying date and hour of the game? Maybe even the name if you keep track of it. It may be useful to track down specific games.. sometimes I don't even know if my latests games have been recorded or were labeled as "bad", cause I can't remember at which "number" I was before starting to ladder.
Ah, another things I've noticed: sometimes I happen to play against opponents who don't belong to any division yet (first game of the season, or maybe first games ever). I usually get HUUUGE point gains against these (since they are probably considered way higher than they are by SC2). How does your software handle that? Should they just be discarded I suppose, or not? Not a bad idea. I look into the graph api to change the tooltipgenerator. But to be honest im kind of worried that blizzard will shut this down. I know that they are aware of our calculation and i see that their goal is, to hide any information from the player. It would not surprise me, if they take measures to make any calculation impossible with next patch.
Other thing: this games are interesting. I throw the game if you get to much points because the f function have to big error margin than ( +21++) under that i take the data. If someone have 0 points or exactly 73 points i reject the data. If he is new ( few games played ) his sigma is still a factor and can screw up the data. I think a rule that a game against someone with less than 10 wins is bad data would can solve this.
|
Still getting some problems. No longer getting a 500 error from Blizz, so all stats should be coming through perfectly.
MMR Log: + Show Spoiler + 21:43:55 MMR-Stats 3.5 21:43:55 --------INIT START-------- 21:43:55 OK Data-Settings loaded! 21:43:55 ERROR No Data Found! Time to ladder... 21:43:55 OK Database loaded! 21:43:55 DB Version:7.1 21:43:57 OK DB up to date! 21:43:57 OK Triggers ready! 21:43:57 OK Gameobserver: ready! 21:43:57 OK Encoding: UTF-8 is avalable on this System 21:43:58 OK: Program is up to date 21:43:58 --------INIT END-------- 21:45:32 -------------------------- 21:45:32 Game Start detected
SC2 Gears Log: + Show Spoiler + Successfully copied last replay to: C:\SC2Replay Archive\TvZ Decadent vs FnaticVibb 2012-06-04 22-03-00.SC2Replay Successfully copied last replay to: C:\SC2Replay Archive\ZvZ FnaticVibb vs Raif 2012-06-04 23-10-04.SC2Replay Successfully copied last replay to: C:\SC2Replay Archive\ZvP FnaticVibb vs Darksoul 2012-06-04 23-24-14.SC2Replay
|
On June 05 2012 07:28 FnaticMatt wrote:Still getting some problems. No longer getting a 500 error from Blizz, so all stats should be coming through perfectly. MMR Log: + Show Spoiler + 21:43:55 MMR-Stats 3.5 21:43:55 --------INIT START-------- 21:43:55 OK Data-Settings loaded! 21:43:55 ERROR No Data Found! Time to ladder... 21:43:55 OK Database loaded! 21:43:55 DB Version:7.1 21:43:57 OK DB up to date! 21:43:57 OK Triggers ready! 21:43:57 OK Gameobserver: ready! 21:43:57 OK Encoding: UTF-8 is avalable on this System 21:43:58 OK: Program is up to date 21:43:58 --------INIT END-------- 21:45:32 -------------------------- 21:45:32 Game Start detected
SC2 Gears Log: + Show Spoiler + Successfully copied last replay to: C:\SC2Replay Archive\TvZ Decadent vs FnaticVibb 2012-06-04 22-03-00.SC2Replay Successfully copied last replay to: C:\SC2Replay Archive\ZvZ FnaticVibb vs Raif 2012-06-04 23-10-04.SC2Replay Successfully copied last replay to: C:\SC2Replay Archive\ZvP FnaticVibb vs Darksoul 2012-06-04 23-24-14.SC2Replay
This have nothing to do with the website or data at all. The trigger detected one start and than hang in there. The only thing i can think of is, that you started the new game before the pretrigger realise that the game is over. So for him its an endless game. And as long as the pretrigger blocks the posttrigger is not allowed to run. On the other hand where is the webgrabbing of the one pretrigger run than....
What you posted is all thats in the log file?
|
On June 05 2012 07:32 skeldark wrote:Show nested quote +On June 05 2012 07:28 FnaticMatt wrote:Still getting some problems. No longer getting a 500 error from Blizz, so all stats should be coming through perfectly. MMR Log: + Show Spoiler + 21:43:55 MMR-Stats 3.5 21:43:55 --------INIT START-------- 21:43:55 OK Data-Settings loaded! 21:43:55 ERROR No Data Found! Time to ladder... 21:43:55 OK Database loaded! 21:43:55 DB Version:7.1 21:43:57 OK DB up to date! 21:43:57 OK Triggers ready! 21:43:57 OK Gameobserver: ready! 21:43:57 OK Encoding: UTF-8 is avalable on this System 21:43:58 OK: Program is up to date 21:43:58 --------INIT END-------- 21:45:32 -------------------------- 21:45:32 Game Start detected
SC2 Gears Log: + Show Spoiler + Successfully copied last replay to: C:\SC2Replay Archive\TvZ Decadent vs FnaticVibb 2012-06-04 22-03-00.SC2Replay Successfully copied last replay to: C:\SC2Replay Archive\ZvZ FnaticVibb vs Raif 2012-06-04 23-10-04.SC2Replay Successfully copied last replay to: C:\SC2Replay Archive\ZvP FnaticVibb vs Darksoul 2012-06-04 23-24-14.SC2Replay
This have nothing to do with the website or data at all. The trigger detected one start and than hang in there. The only thing i can think of is, that you started the new game before the pretrigger realise that the game is over. So for him its an endless game. And as long as the pretrigger blocks the posttrigger is not allowed to run. On the other hand where is the webgrabbing of the one pretrigger run than.... What you posted is all thats in the log file?
Yep, that's everything in the log, also there was perhaps an hours wait between 1st and second game, as I went and did some work (just alt tabbed out).
|
On June 05 2012 07:34 FnaticMatt wrote:Show nested quote +On June 05 2012 07:32 skeldark wrote:On June 05 2012 07:28 FnaticMatt wrote:Still getting some problems. No longer getting a 500 error from Blizz, so all stats should be coming through perfectly. MMR Log: + Show Spoiler + 21:43:55 MMR-Stats 3.5 21:43:55 --------INIT START-------- 21:43:55 OK Data-Settings loaded! 21:43:55 ERROR No Data Found! Time to ladder... 21:43:55 OK Database loaded! 21:43:55 DB Version:7.1 21:43:57 OK DB up to date! 21:43:57 OK Triggers ready! 21:43:57 OK Gameobserver: ready! 21:43:57 OK Encoding: UTF-8 is avalable on this System 21:43:58 OK: Program is up to date 21:43:58 --------INIT END-------- 21:45:32 -------------------------- 21:45:32 Game Start detected
SC2 Gears Log: + Show Spoiler + Successfully copied last replay to: C:\SC2Replay Archive\TvZ Decadent vs FnaticVibb 2012-06-04 22-03-00.SC2Replay Successfully copied last replay to: C:\SC2Replay Archive\ZvZ FnaticVibb vs Raif 2012-06-04 23-10-04.SC2Replay Successfully copied last replay to: C:\SC2Replay Archive\ZvP FnaticVibb vs Darksoul 2012-06-04 23-24-14.SC2Replay
This have nothing to do with the website or data at all. The trigger detected one start and than hang in there. The only thing i can think of is, that you started the new game before the pretrigger realise that the game is over. So for him its an endless game. And as long as the pretrigger blocks the posttrigger is not allowed to run. On the other hand where is the webgrabbing of the one pretrigger run than.... What you posted is all thats in the log file? Yep, that's everything in the log, also there was perhaps an hours wait between 1st and second game, as I went and did some work (just alt tabbed out). is something in the gamedata? do you have ramscanner enabled or not?
|
|
|
|