|
On January 17 2014 05:35 THF wrote:
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.
Values for your opponents cannot be calculated as you are capped. Your MMR needs to rise quite considerably before it is possible. You can read more (think yourself as 'opponent' when reading the examples): http://www.teamliquid.net/forum/viewpost.php?post_id=19316583
On January 17 2014 05:35 THF wrote:
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".
Yes the value for the last 'good game' (opponent: sepisepi) is very likely incorrect (whether it is because data errors or slightly incorrect cap detection border). You can either set it 'bad' immediately via config or wait for the next 'good game' and if it is in line with the older ones, then set the 'spike game' 'bad'.
Also the match played on 2014-01-05 against 'YoungWolf' has incorrect value (clear spike almost two leagues upwards, if you don't see the spike, you likely have 'autocorrect' enabled from the low toolbar, I usually prefer to keep it disabled). You can set it bad via config.
To set games bad:
'Config' > 'Data' > 'Refresh' > Choose match from the list > 'mark as bad game'
|
2014-01-19: MMR tool may stop working partially or entirely when SC2 v. 2.1 is published
When SC2 patch 2.1 is published early next week following things will potentially happen: 1) Depending how large the changes are to the replay format, the tool may stop working (if sc2gears is unable to parse basic data from the replays), 2) Memory reader will stop working until new memory addresses are solved and published, 3) Blizzard will likely change league offsets / thresholds. This means that MMR graph for games played after the patch will become erratic + MMR values & league entry barriers will be incorrect until new offsets are calculated. Offset calculations require lots of collected match data (it takes potentially 2 to 3 weeks to collect enough data). When new offsets are published, old calculations will be automatically fixed, 4) There may be other potential changes that may cause that the tool stops working (e.g. a new version of the tool will be needed, if web profiles are changed)
|
MMR stats seems to be really bugged.
![[image loading]](http://i.imgur.com/VrQhXRL.jpg)
For starters, see that last rising bar, where I won like 10 games in a row? That was at the start of this season already, yet it marks the 17th season as áfter that jump.
Also during my 50 day inactivity, I apparently didn't lose MMR, or at least only 10 MMR?
According to my MMR, I have never been below masters league.
It would be true since my last 8 seasons were all masters league, but this time i ended up in platinum :').
So after 5 games or so I got promoted to diamond, and right now im hovering at around top diamond, but no masters in sight.
Yet my MMR says I have 1550 MMR, which would be mid masters?
I have memory reading on and always have sc2 mmrstats on.
my account: http://eu.battle.net/sc2/en/profile/2240242/1/Kaluro/
|
On January 21 2014 09:15 kaluro wrote:
MMR stats seems to be really bugged.
Neither of the things you reported were caused by bugs. Please read further.
On January 21 2014 09:15 kaluro wrote:+ Show Spoiler +For starters, see that last rising bar, where I won like 10 games in a row? That was at the start of this season already, yet it marks the 17th season as áfter that jump.
I don't have your records from the beginning of the season as MMR tool server was down (server got up later same day as new season started). You likely played before that and had an old Bnet DB file for S16 (Manual update for the new DB file for S17 was required as it could not be downloaded from the server as server was down).
When a new season starts, it takes some time for Blizzard to update the web profiles & the API (both return data for the old season at first). Also after they had updated the web profiles and the API, the API still randomly gave either S16 or S17 data for some time. If you had old DB and web API said that the season was 16, it would have thought it was correct and marked that the match was played during S16. If you had old DB and the API returned S17, then it would have marked S17, but set the data invalid. If you had new DB and API said S16, it would have set the season number invalid. If you had new DB and API said S17, it would have set season as S17.
The incorrect season markers are technically not caused by a bug, but intended. Description from MMR tool v. 10.0 change notes regarding markers says:
"Please note that different markers usually take only one attribute from each match into consideration. Thus if there is some other problems with the match data, incorrect markers can be shown."
Season markers mostly look only the season numbers stored regarding each match. Thus if that data is invalid, it can lead to incorrect season markers.
I don't store a list of exact season start times for each server. It would be the only way to get exact season markers. During WoL finding out those times would have been extremely laborious as Blizzard switched the seasons for each server manually (I usually knew it only for 1 or 2 servers and do not have that data anymore). After HotS they automated the season change process and I could deduct the start times based on bonus pool accumulation start times. But I still find it unnecessary to use exact times for season marker determination. It would even require that a new version of the entire tool is published when each season starts, unless a long list of season start times would be distributed e.g. via Bnet DB files (I am not fond of that idea).
The current solution is good enough. Also one can hope that Blizzard will not mess up their SC2 API updates when future seasons start (they have messed up the updates for two seasons in a row).
On January 21 2014 09:15 kaluro wrote:Also during my 50 day inactivity, I apparently didn't lose MMR, or at least only 10 MMR? According to my MMR, I have never been below masters league. It would be true since my last 8 seasons were all masters league, but this time i ended up in platinum :'). So after 5 games or so I got promoted to diamond, and right now im hovering at around top diamond, but no masters in sight. Yet my MMR says I have 1550 MMR, which would be mid masters? I have memory reading on and always have sc2 mmrstats on. my account: http://eu.battle.net/sc2/en/profile/2240242/1/Kaluro/
Not a bug. You don't have any good games after your inactivity period. Thus it estimates your current MMR based on a MMR value calculated before your inactivity period. The next good game will fix your current MMR value (if you manage to get a 'good game' before SC2 patch 2.1. When 2.1 is published Blizzard will likely change league offsets. Thus values from 'good games' played after that are not compatible with current values. In the future I will likely need to prepare a new version of the tool that will archive all matches played before v.2.1).
You can read how 'good games' affect estimate games in your graph from here: http://www.teamliquid.net/forum/viewpost.php?post_id=20506999
You can read when you are likely to get 'good games' e.g. from here: http://www.teamliquid.net/forum/viewmessage.php?topic_id=334561¤tpage=162#3226
It is also possible to color code good games on graph (Note that this requires you have match dots enabled from the bottom toolbar) if you enable following setting: 'Config' > 'Data' > 'Mark assumptions / bad games'. Remember to save the changes.
|
On January 21 2014 10:53 korona wrote:
It is also possible to color code good games on graph (Note that this requires you have match dots enabled from the bottom toolbar) if you enable following setting: 'Config' > 'Data' > 'Mark assumptions / bad games'. Remember to save the changes.
Ahh thanks! You were right  , sorry for the 'bug;' assumption.
I enabled dots and bad game checks now.
Turns out that my last 65 (!) games have been bad ones.
|
On January 21 2014 18:44 kaluro wrote:Show nested quote +On January 21 2014 10:53 korona wrote:
It is also possible to color code good games on graph (Note that this requires you have match dots enabled from the bottom toolbar) if you enable following setting: 'Config' > 'Data' > 'Mark assumptions / bad games'. Remember to save the changes.
Ahh thanks! You were right , sorry for the 'bug;' assumption. I enabled dots and bad game checks now. Turns out that my last 65 (!) games have been bad ones.
Yes, bad luck. Thought 15 of those were played when you presumably had the old DB file and the web profiles offered old data (just after season change). If it would have been a normal situation, you would have likely got 'good games' before you were promoted to diamond.
But if you want to manually count the estimates for your MMR after your inactivity period then please deduct 310 326 from the shown values (-310 for max decay and -16 for the last game (loss) before inactivity (this of course is already counted in the graph values)). Based on the last value in your MMR graph picture, it would suggest your MMR is somewhere around ~1220.
|
Can it be that MMR tracker isnt working due the patch? just wondering due i played a game but its not added in the list,
|
On January 22 2014 06:33 Dthehunter wrote:
Can it be that MMR tracker isnt working due the patch? just wondering due i played a game but its not added in the list,
On January 19 2014 14:59 MMR-Stats wrote:
2014-01-19: MMR tool may stop working partially or entirely when SC2 v. 2.1 is published
When SC2 patch 2.1 is published early next week following things will potentially happen: 1) Depending how large the changes are to the replay format, the tool may stop working (if sc2gears is unable to parse basic data from the replays), 2) Memory reader will stop working until new memory addresses are solved and published, 3) Blizzard will likely change league offsets / thresholds. This means that MMR graph for games played after the patch will become erratic + MMR values & league entry barriers will be incorrect until new offsets are calculated. Offset calculations require lots of collected match data (it takes potentially 2 to 3 weeks to collect enough data). When new offsets are published, old calculations will be automatically fixed, 4) There may be other potential changes that may cause that the tool stops working (e.g. a new version of the tool will be needed, if web profiles are changed)
|
On January 22 2014 06:33 Dthehunter wrote:
Can it be that MMR tracker isnt working due the patch? just wondering due i played a game but its not added in the list,
i saw the post like litterly after scrolling over it for 18times so ignore me please^^
|
On January 19 2014 14:59 MMR-Stats wrote:
2014-01-19: MMR tool may stop working partially or entirely when SC2 v. 2.1 is published
When SC2 patch 2.1 is published early next week following things will potentially happen: 1) Depending how large the changes are to the replay format, the tool may stop working (if sc2gears is unable to parse basic data from the replays), 2) Memory reader will stop working until new memory addresses are solved and published, 3) Blizzard will likely change league offsets / thresholds. This means that MMR graph for games played after the patch will become erratic + MMR values & league entry barriers will be incorrect until new offsets are calculated. Offset calculations require lots of collected match data (it takes potentially 2 to 3 weeks to collect enough data). When new offsets are published, old calculations will be automatically fixed, 4) There may be other potential changes that may cause that the tool stops working (e.g. a new version of the tool will be needed, if web profiles are changed)
2014-01-21: MMR tool cannot record new matches after v. 2.1 until it is able to parse new replays
At least two of the expected things happened when SC2 was patched to v. 2.1:
1) Sc2gears cannot parse data from new replay files. This also renders MMR tool inoperable until either sc2gears is updated (unknown if it will be updated as its successor Scelight has been published) or another replay file parser is ported for the MMR tool.
2) Memory addresses were changed and thus the memory reader cannot read the player names & ID:s. These can be solved, but it does not fix the main problem that the replay parsing does not work.
Edit: Also the league offset/threshold changes that Blizzard was talking about may have been small. There are no reports about mass promotions after people have played v. 2.1 games.
|
2014-01-21: SC2 version 2.1 sample replays are needed!
Sc2gears support for v. 2.1 replays will be looked at in near future (no timetable yet). To ease up the process please post links to some fresh 1v1 ranked replays from v. 2.1 to this thread or PM them to this account. The replays should preferably be from longer matches where lots of different types of units & unit abilities are used.
This message will be updated when enough replays have been gathered.
I believe there now is now enough sample replays. Thank you for those who shared replays in this thread & via PM:s. Amongst them were some replays that nicely complemented ones found from other sources.
|
|
|
|
|
2014-01-22: Please update Sc2gears
Sc2gears was updated earlier today by Dakota_Fanning to support v. 2.1 replays. This update is required by the MMR tool. To update Sc2gears go to its 'Help' menu and choose 'Check Updates (you may have enabled auto-updates so it already may have been updated)
New memory addresses for the memory reader have not yet been published. Also offset calculations require several weeks of worth collected data. Please use MMR tool even if the calculation results for new matches may be more or less incorrect for few weeks.
|
On January 23 2014 01:25 MMR-Stats wrote:
2014-01-22: Please update Sc2gears
Sc2gears was updated earlier today by Dakota_Fanning to support v. 2.1 replays. This update is required by the MMR tool. To update Sc2gears go to its 'Help' menu and choose 'Check Updates (you may have enabled auto-updates so it already may have been updated)
New memory addresses for the memory reader have not yet been published. Also offset calculations require several weeks of worth collected data. Please use MMR tool even if the calculation results for new matches may be more or less incorrect for few weeks.
New DB file was published (v 17.3). Please restart Sc2gears to download it. It contains new memory reader addresses.
Also note that after this update the memory reader does not work for KR, SEA and CN until they have been updated to SC2 v. 2.1.
--
Could someone confirm that the reader works after this update (solving the addresses was done different way than normally and the addresses were not tested much).
The memory reader results appear into the 'Game Log' tab of the 'Log' little after each ladder game has started. All data (name1, id1, name2, id2) should appear.
Example text how it should appear in game log:
21:06:09 Ramscanner: name1 : PlayerA
21:06:09 Ramscanner: id1 : 1-1234567
21:06:09 Ramscanner: name2 : PlayerB
21:06:09 Ramscanner: id2 : 1-2345678
--
Edit 1: Based on the uploaded data memory reader seems functioning ok
Edit 2: Some early statistics from those users who have played after the patch with the updated Sc2gears: 206 different characters accounts have played since 2.1 was published on their server. Around 5 have been promoted. All of them would likely have been promoted with old offsets too (they were just under the promotion threshold and were promoted by winning a match).
Post-patch graphs looked visually quite normal. Except one master player had faced a gold player and got a 'good game' that had totally incorrect value. Based on it he would have been in gold range based on old thresholds (he had several good reading before the patch that indicated he was close to bottom border of master league). Could not deduct the reason for it with a quick glimpse... Except that one abnormal value it almost looked like there would have been no chances (only quickly browsed the graphs and looked the promoted ones little longer)... Of course much much more data is needed before the situation can be analyzed.
Edit 3 / 2014-01-25: the abnormal value was because of missing games. That player had tanked his MMR himself.
|
Works for me
|
Hi, first of all thanks for the tool, I find it very useful to have an objective look at your win/losses.
I would like to know:
1) if it's possible to add the games which are missing. I have the replays, I will just need sc2gear and mmrstats to read those
2) if it's possible to just adjust the graph manually, like adding 4 missing games, and points as "player assumption"
Thanks.
|
I would like to say that both korona and Blizzard are representing extreme examples of customer support.
korona demonstrates the very best, with frequent, prompt and accurate communication on the subjects people are concerned with. Blizzard demonstrates the very worst, with infrequent, late and misleading communication usually focusing on things few people actually care about.
Just think how amazing Blizzard would be if they sacked all their customer relations managers and hired korona to lead the team instead!
|
On January 23 2014 22:13 KerriganSon wrote:
Hi, first of all thanks for the tool, I find it very useful to have an objective look at your win/losses.
I would like to know:
1) if it's possible to add the games which are missing. I have the replays, I will just need sc2gear and mmrstats to read those
2) if it's possible to just adjust the graph manually, like adding 4 missing games, and points as "player assumption"
Thanks.
If you're talking about "possible with the actual version", then no for both.
If you're talking about "possible to implement":
I'd guess no for 1), because it reads actual data of both players. However, if you'd let it read let's say a year old replay with current player info from their battle.net account, you'd most probably get really weird results. Even though both players may have been on similar MMR when the game was played, they may be very far from each other now and the tool wouldn't have a chance to know.
For 2) I'd guess it's possible, but not really useful. The graph will update on any good game anyways, so it's better to wait for one than to put in fake points that the tool would have to correct based on newer good games also. Overall I'd say: Not a good idea.
|
|
|
|
|
|
|
|
EPT
