EPTMatch Making Rating Tool - Page 129
| Forum Index > SC2 General |
|
synd
Bulgaria586 Posts
| ||
|
KingAlphard
Italy1705 Posts
| ||
|
MMR-Stats
26 Posts
This Program dont work with 2.0.4 yet. Please wait for next update. | ||
|
Braunz
Canada13 Posts
| ||
|
synd
Bulgaria586 Posts
And yeah, I understand that the plugin (MMR stats) isn't yet working and the servers are closed. So I ask can they be opened so we can provide data fast and the shit gets fixed? ^^ | ||
|
korona
1098 Posts
On February 23 2013 10:49 synd wrote: The program (SC2Gears) does work with the replays and the parser is updated. Dunno if it's 100% working but most of the stuff is correct. And yeah, I understand that the plugin (MMR stats) isn't yet working and the servers are closed. So I ask can they be opened so we can provide data fast and the shit gets fixed? ^^ On February 23 2013 02:23 MMR-Stats wrote: This Program dont work with 2.0.4 yet. Please wait for next update. Like it says MMR-tool does not yet support v 2.0.4 matches. Thus all data is flagged as bad data as if the data would be used for MMR calculation it would give false results. There is still no timetable yet when new version of MMR-tool is released. There were several changes Blizzard made in v. 2.0.4 that have to be implemented and tested before releasing new version of MMR-tool & flagging the future data good. Here are few but not all: 1) unranked / ranked support. --> all unranked matches played by the user have to be ignored as they do not affect ranked MMR. Blizzard does not separate unranked & ranked matches in match histories. There are ways to detect if a match was unranked or not, but no way to be 100% sure (will likely have false positives/negatives), unless there is a field in replay file that tells if the match was unranked or not & sc2gears supports reading it (current version does not have support). 2) Bonus pool mechanic changes regarding lost matches. Blizzard hides the true change points in match histories. These can be often calculated (display bug will cause problems) if the memory reader is on. If the memory reader is off & and opponent loses, there is no way to calculate the change. --> less cases where true change can be calculated. 3) Display bug regarding bonus pool at least on EU & NA servers. Possibly also on other servers but have not checked it. This cannot be fully fixed before Blizzard fixes their web profiles. A dirty fix can be made (flagging all data bad where a player has 0 bonuspool according to the web profile). But after Blizzard fixes their profiles the new data recorded would be invalid --> When Blizzard fixes the display bug the database should be locked fast & new version of MMR-tool released & database unlocked & users manually flagging all games played between Blizzard's fix & release of new MMR-tool version as bad. To keep you updated I have written several posts during the last week where I describe many of the changes needed. To get more info please read them. When a new version of the tool is released, we will notify you in this thread, in its original post and the MMR tool will tell there is a new version available. EDIT: Database closed in this case means data is flagged as bad. Data is still collected, but calculations cannot be made based on it before support for 2.0.4 has been implemented (only estimates based on previous good data or good data after 2.0.4 is supported). And 2.0.4 data collected with the old version cannot be fixed retroactively. Thus we will not unlock the database like synd has requested several times before we will release the new version of MMR-tool. | ||
|
Lysergic
United States355 Posts
Could you edit the title of this post to include better search keywords, such as MMR-Stats or MMR plugin? It can be hard to find this thread. | ||
|
korona
1098 Posts
On February 24 2013 23:31 lysergic wrote: I'm sure you know this already, but Versus A.I. games are scanned by MMR-stats now that they're AutoMM. Games against AI show up on my graph and affect my MMR. Thank you for letting us know! The current development version of unranked detection should handle this case also. | ||
|
MMR-Stats
26 Posts
Support for SC2 WoL v. 2.0.4 Check details from the original post: http://www.teamliquid.net/forum/viewmessage.php?topic_id=334561#1 Please note that we decided to publish this version without memory scanner to get it out faster. This means you will get less 'good games' but now the new SC2 patch is supported. InfCereal has already prepared a new version of memory scanner which is already functional, but more testing is needed. We hope to be able to publish new memory scanner during this week. | ||
|
Rainman5419
United States92 Posts
| ||
|
Aimconquest
15 Posts
| ||
|
korona
1098 Posts
On February 26 2013 04:07 Aimconquest wrote: Quick question: Have the adjusted points cutoffs changed? I seem to losing 14 a match instead of the 12 I would normally with the amount of points I have? I am not sure what you mean. The adjusted points are calculated the same way as always: ladder points - used bonus pool. Of course regarding ladder point changes now when you lose, deductions are done primarily from the bonus pool and secondarily from ladder points. Blizzard hides the points taken from the bonus pool in match history. If memory scanner is not on (it is disabled for everyone in 8.2) it is not possible to calculate how many points opponents lost (combined from bonus pool + ladder points) as we don't have prematch data. For the player himself it is possible as the tool knows his active profile (does not need mem scanner for that). Due to bonus pool display bug after user has spent all his bonus pool visible in his web profile, calculations are not possible (in-game client show the correct larger bonus pool). | ||
|
Braunz
Canada13 Posts
| ||
|
korona
1098 Posts
On February 26 2013 06:50 Braunz wrote: this thing sucks, ive done 30+ games, not a single one was good, no graph, time to uninstall it :D Yes. Due to bonus pool display bug on Battle.net profiles lots of data has to be invalidated. Also there are now less cases when a good game can be achieved when memory reader is off. We hope to be able to publish it during this week so do not uninstall yet  (chances of getting good game will dramatically rise when memory reader is available and on). | ||
|
breath
United States199 Posts
| ||
|
InfCereal
Canada1759 Posts
On February 26 2013 07:14 breath wrote: If you don't mind me asking how are you figuring out if the game is unranked or ranked ? We store the win/loss before the game, then check it after the game. If it didn't change, it's unranked. | ||
|
Braunz
Canada13 Posts
On February 26 2013 07:07 korona wrote: Yes. Due to bonus pool display bug on Battle.net profiles lots of data has to be invalidated. Also there are now less cases when a good game can be achieved when memory reader is off. We hope to be able to publish it during this week so do not uninstall yet (chances of getting good game will dramatically rise when memory reader is available and on).Thats good news, but doesnt that increase the risk of getting caught by blizzard and banned? | ||
|
korona
1098 Posts
On February 26 2013 07:14 breath wrote: If you don't mind me asking how are you figuring out if the game is unranked or ranked ? Blizzard does not separate ranked, unranked or vs Ai games on match histories. Also a field containing this info has not yet been found from replay files (if there is one, if found the detection will be 100%). Thus the detection happens simply by comparing win & loss counts. This is not a bullet proof solution as there may be web profile syncing issues which might lead to false positives/negatives in rare cases. Also if the web profiles are down we cannot use this method. For player himself we always fetch pre-match records so we can compare. If unranked match is found its data is dropped. If the memory reader is off we don't have pre-match data for the opponent. In these cases we set his data invalid if his point change is 0 on his match history. If memory reader is on, we can compare opponent's win & loss counts (+ possible checks regarding remaining bonus pool changes). But whether the opponent plays ranked or unranked, it always affects the player's MMR. Thus we only set opponent's data invalid if unranked match is found or suspected (player's MMR is calculated from opponent's data. If opponent is playing unranked we cannot calculate it). At the moment we don't consider extending memory reader to read it from SC2 memory. | ||
|
breath
United States199 Posts
 Keep up the good work guys, love this tool. | ||
Excalibur_Z
United States12237 Posts
On February 26 2013 07:36 korona wrote: Blizzard does not separate ranked, unranked or vs Ai games on match histories. Also a field containing this info has not yet been found from replay files (if there is one, if found the detection will be 100%). Thus the detection happens simply by comparing win & loss counts. This is not a bullet proof solution as there may be web profile syncing issues which might lead to false positives/negatives in rare cases. Also if the web profiles are down we cannot use this method. For player himself we always fetch pre-match records so we can compare. If unranked match is found its data is dropped. If the memory reader is off we don't have pre-match data for the opponent. In these cases we set his data invalid if his point change is 0 on his match history. If memory reader is on, we can compare opponent's win & loss counts (+ possible checks regarding remaining bonus pool changes). But whether the opponent plays ranked or unranked, it always affects the player's MMR. Thus we only set opponent's data invalid if unranked match is found or suspected (player's MMR is calculated from opponent's data. If opponent is playing unranked we cannot calculate it). At the moment we don't consider extending memory reader to read it from SC2 memory. In SC2gears, Ranked matches were listed as "Game type: AutoMM", is this the case for Unranked matches too so there's no way to differentiate? | ||
| ||
