|
On January 24 2014 03:40 Denoginizzer wrote:Well i got the new db 17.3 and version 14.3.3 of sc2gears. i see the trigger running after a game, he is counting 0/0 games and my log looks like this: (i played 4 games tonight) + Show Spoiler +third game: + Show Spoiler +Mainlog: + Show Spoiler +18:00:00 MMR-Stats 10.0 18:00:00 --------INIT START-------- 18:00:00 OK Encoding: OK 18:00:00 OK Sc2Gears auto-save is enabled. 18:00:00 Loading HOTS DB file... 18:00:00 OK Loading DB version 17.3 done 18:00:01 OK DB up to date! 18:00:01 Loading character profiles: 18:00:01 Profile: http://eu.battle.net/sc2/en/profile/3235069/1/Denoginizzer/ 18:00:01 Profile: http://us.battle.net/sc2/en/profile/4241209/1/Denogginizer/ 18:00:01 OK Settings loaded! 18:00:01 OK Timeclient: Systemtime offset: 0 sec 18:00:01 OK Gamedata: 2 Games loaded. 18:00:01 Loading Archive... 18:00:01 No Data Found! 18:00:01 Memory scanner is enabled 18:00:01 OK Triggers ready! 18:00:03 OK: Program is up to date 18:00:03 --------INIT END-------- keep up the good work ! i hope you can use this info
I have data for 7 uploaded games from your NA account (3 good games out of 7). So the reason is not that it does not record the matches. Could it be that you don't have write rights under the sc2gears directories (for example under 'program files' there are often limited write rights) and the MMR tool cannot save those matches into a save file. In this case you can running the MMR tool in admin mode or move sc2gears folder to somewhere else.
You can check if the save file is there. It is called 'games.i.data'. It is located in following directory:
[PATH_TO]\Sc2gears\User content\Plugin file cache\mmr.plugin.MMRPlugin\
Also are you sure you are looking at the your NA account graph as you also have setup an EU account that has no games played (your EU account is before your NA account in the list)?
|
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
Via MMR tool it is not possible. It might be possible to trick sc2gears to read the replay as a new one (game end detection is based on triggering sc2gears auto-save), but the tool would then read current data for both players that would lead to incorrect results. It would place the record of that match in the end of your match list (not to the correct position). Also it would upload the incorrect data to MMR tool server.
Actually I have to prevent this in some future version of the MMR tool. Just earlier today was looking that somebody had managed to record a game played with old version of SC2, even if the new version had already been published for that server (thankfully at least it had been set 'bad' as the data from the replay did not match the current web profiles).
It would be possible to add such feature. It would first need some kind of UI popup where you could choose the replay file from the disk, then it would read basic info from the replay (date, server, length, players, races, result, map and mark everything else unknown & game as 'bad' so it would be used as an estimate match), then shuffle it based on the date to correct position in the match list and then recalculate the graph. Does not sound that complex, but still it would require a lot of work (implementation, testing & publishing). Maybe someday (not near future) I find time / motivation to implement this. But there are lots of features in my wish list that I rank higher than this, but have not found spare time to add them. Keeping the tool running is quite a time sink itself (season end & start formalities, going through data for potential changes by Blizzard, checking collected data from time to time for potential new problems, offset calculations if needed, responding to user problems, etc).
Also if there would be such feature many users would abuse it. They would add estimate games to boost their graph to showoff it to their friends or followers. This would also mean you could not trust the graphs what people show e.g.in this thread & I would every time need to load the data from the server when checking someone's problems (which of course requires more time & effort).
On January 23 2014 22:13 KerriganSon wrote:
2) if it's possible to just adjust the graph manually, like adding 4 missing games, and points as "player assumption"
It is possible to manually add games as the save files are human readable (for example I have added all personal missing games manually). But to do it in the 'correct way' is not that straightforward. If I would give instructions then many would try it and fail in some respect. And those errors would potentially cause problems in the future (would potentially be pain to troubleshoot).
But for now as there is no supported feature for this, it might be better to just wait for the next 'good game' to fix the current MMR estimate.
|
Hi and thanks for the answer. I guess then I will just wait the next "good" game.
If the above is the preferred way, what is the purpose of "player guess" and "player assumption" icons in the legend at the bottom of the mmr window? Were those put there for future expansions? Those icons are the ones who triggered my request.
|
On January 24 2014 05:48 KerriganSon wrote:
Hi and thanks for the answer. I guess then I will just wait the next "good" game.
If the above is the preferred way, what is the purpose of "player guess" and "player assumption" icons in the legend at the bottom of the mmr window? Were those put there for future expansions? Those icons are the ones who triggered my request.
You can turn on the color coding (also hovering your cursor over the match dot tells the quality):
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.
Edit: Will standardize terms in some future version: Player assumption --> Player estimate. Will also hide "Player guess" legend that is meant to pre-results graph from the actual graph & hide 'Player win', 'Player lose' and 'Player estimate' legends from the pre-results graph.
|
On January 24 2014 05:48 KerriganSon wrote:
Hi and thanks for the answer. I guess then I will just wait the next "good" game.
If the above is the preferred way, what is the purpose of "player guess" and "player assumption" icons in the legend at the bottom of the mmr window? Were those put there for future expansions? Those icons are the ones who triggered my request.
If you never had a good game, you'd get "player guess" dots first. As soon as a good game shows up, the games before can be calculated (assumed) backwards. They will become "player assumptions" like all games that are not good games. If you win, but it's not a good game, the tool assumes that you gain 16 MMR. If you lose and it's not a good game, it assumes that you lose 16 MMR.
|
Update regarding league offsets for SC2 v 2.1:
I ran calculations for the 3381 recorded matches by 486 different accounts for SC2 v. 2.1. (quite low amount of matches. I wonder if people have not updated their Sc2gears).
As there is so little data at this point, the offset calculations returned results pretty much only for two offsets: for diamond-master offset and for master-gm offset. At the moment it seems that these offsets are either the same as before or there has been very small increase. For example old diamond-master offset was 250. The first results suggest that the current offset might be between 250 to 270. Same deal with GM. Old offset was 400. New values suggest 400 to 420. Originally when I calculated the old values in June, I fine-tuned them based on user graphs to be little bit smaller than the values suggested.
It takes time to collect enough data for other leagues. But if these early results are to believed, there might not have been any changes - at least for diamond-master and master-gm offsets. Based on these results I would use old value for diamond-master offset and potentially slightly increase master-gm offset (if user graphs would look ok after the increase).
Also the user graphs look normal (are not erratic) regarding post-2.1 data and promotions happen in correct places.
2 hours ago Blizzard gave an update where they suggest that they would be making league boundary changes in near future: http://us.battle.net/sc2/en/forum/topic/11423622282#1
In it they claim that they already made some boundary changes with patch 2.1. I wonder if the changes were actually something else than league offset / threshold changes...
But when they do concrete changes, I hope they settle into those values. MMR tool uses static offsets & if the offsets change all the time, it would produce incorrect results (keeping up with the changes would be painful).
|
On January 24 2014 14:47 korona wrote:
In it they claim that they already made some boundary changes with patch 2.1. I wonder if the changes were actually something else than league offset / threshold changes...
In the blog where they announced the change they said:
"Accordingly, around the time we release the 2.1 patch, we are changing how league boundaries are determined so that we can more closely match the desired percentage of players in each league."
Not "changing league boundaries" but "changing how league boundaries are determined". When I read that I wondered if they were going to switch to dynamic league boundaries, but as ususal Blizzards statements are hard to interpret.
|
On January 24 2014 15:05 Mendelfist wrote:Show nested quote +On January 24 2014 14:47 korona wrote:
In it they claim that they already made some boundary changes with patch 2.1. I wonder if the changes were actually something else than league offset / threshold changes... In the blog where they announced the change they said: "Accordingly, around the time we release the 2.1 patch, we are changing how league boundaries are determined so that we can more closely match the desired percentage of players in each league." Not "changing league boundaries" but "changing how league boundaries are determined". When I read that I wondered if they were going to switch to dynamic league boundaries, but as ususal Blizzards statements are hard to interpret.
Yes. I have been afraid for some time that they would switch to dynamic boundaries. Hopefully they do not as it might cause serious problems regarding the MMR tool.
After looking visually at the user graphs and transitions from 2.0.11 data to 2.1 data it seems the offsets are the same, but I have likely had slightly too low master-gm offset (testing the higher value returned positive results regarding user graphs. Have to test this with last seasons data later).
Edit: old master-gm offset values seem to be giving better results regarding last season's data.
Edit 2: Will have to wait and see until much more data is collected to get calculation results for the other offsets...
|
I wonder what is meant by dynamic offsets (I know this is just speculation at the moment but still...).
Would this merely by offsets that decrease at probably some constant level over time to keep pace with MMR decay? Or would this be something more complex that would move boundaries to ensure a specific % of players, or of active players (however determined) are in each league?
If the latter I wonder how they'd do that. Could the results of other peoples matches push you into a promotion or demotion while you are offline? How would they implement that, would the change come into effect only next time you played a game? That would cause problems for people sitting on the boundary unless they had some hysteresis effect.
On the other hand if it were the former then... thats just stupid. I'm still struggling to see how they convince themselves this ladder is effective. Imagine if this was how football leagues worked, where most teams were concentrated at the bottom and factors other than match results affected their rank, such as random additions and subtractions of points. Then every now and then divisions are all changed and suddenly you're in a different one. I'm pretty sure no one would accept such a system as anything other than a bad joke... but Blizzard don't seem to get it.  
|
On January 25 2014 04:20 Spirit09 wrote:
Would this merely by offsets that decrease at probably some constant level over time to keep pace with MMR decay? Or would this be something more complex that would move boundaries to ensure a specific % of players, or of active players (however determined) are in each league?
The latter, but I don't even know if dynamic thresholds is a good idea, or how it could be implemented. They would have to change the way they measure inactivity for example. They are currently relying on remaining bonus pool, but that doesn't work well during the season, just at the end of it.
If the latter I wonder how they'd do that. Could the results of other peoples matches push you into a promotion or demotion while you are offline? How would they implement that, would the change come into effect only next time you played a game? That would cause problems for people sitting on the boundary unless they had some hysteresis effect.
Noone gets promoted or demoted when offline. If the dynamic changes of league thresholds are small and slow it wouldn't necessarily be very obvious, but the more I think about it the less likely it seems. I think they have just made ordinary manual league offset changes.
That being said: That korona hasn't seen any changes is really strange.
|
On January 25 2014 06:06 Mendelfist wrote:
That being said: That korona hasn't seen any changes is really strange.
I wonder about this. korona's data so far is only looking at master-GM and master-diamond correct? What if they think these boundaries are ok, much of the community would probably agree here. So we still need to wait to see if they changed only the lower leagues. My personal view is that its gold down where there are the serious problems. My further personal view is that this isn't a good way to deal with those problems.
So we just need to wait for the mmr stats to roll in I assume. From reading between the lines on bnet it seems they've shifted boundaries but conservatively, and they'll possibly be shifting more. Everything would be a lot easier if they just told us exactly what they've been doing.
|
United States12224 Posts
Looks like people are being shuffled around a little more aggressively now. Time to check those offsets again.
|
On January 25 2014 07:49 Excalibur_Z wrote:
Looks like people are being shuffled around a little more aggressively now. Time to check those offsets again.
If and when things start happening I hope I get good timestamps for each server as cannot do the archiving in the future based on the version number or maintenance break time anymore (as initial changes seemed non-existent). Let's see if there has been movement during the past hour like some people in some forums claim (still the increase of numbers in higher leagues in last day based on nios.kr are about 'normal').
Edit: some user graphs have become erratic and there are promotions in unexpected places. This has potentially started on US maybe around 20 pm CET, which was 5 hours ago (now ~1 am CET). --> offsets/thresholds have been likely changed (or e.g changes directly to MMR based on historical data). Changes to EU happened likely later.
|
Just got promoted into master YAAAY! Anyway just wanted to share since I got promoted when my mmr is actually at a low point so maybe they did additional changes to the boundaries?
http://postimg.org/image/wnvqwzidb/
The last game isn't good but the two proceeding it is while also being losses so it's hard seeing I won more mmr in one game than i lost in two and I didn't get promoted on any of the highpoints previous but after my first win today.
Also I really wonder about this mmr decay. I had opponents last game who was top 250 master on the region (in points) with an mmr in the mid/lowish range.
|
On January 25 2014 14:00 Incand wrote:Just got promoted into master YAAAY! Anyway just wanted to share since I got promoted when my mmr is actually at a low point so maybe they did additional changes to the boundaries? http://postimg.org/image/wnvqwzidb/The last game isn't good but the two proceeding it is while also being losses so it's hard seeing I won more mmr in one game than i lost in two and I didn't get promoted on any of the highpoints previous but after my first win today. Also I really wonder about this mmr decay. I had opponents last game who was top 250 master on the region (in points) with an mmr in the mid/lowish range.
Congrats! Calculations regarding matches played during last ~10 hours may be incorrect (also there is an error margin regarding results from 'good games') as Blizzard finally did some changes to the ladder (these changes were expected already for patch 2.1, but there was no offset changes until now). If the changes were offset / threshold changes, then it requires that new offsets are calculated & published. After that the calculations are automatically fixed. To gather enough data it may take 2 to 3 weeks (for diam-master offset couple of days, but will not publish them until there are values for all offsets. But in couple of days it is known if value for that offset differs from the old value). But at this point it is unknown if the changes were offsets changes or something else (e.g. direct changes to player MMR e.g. based on historical data).
On January 25 2014 07:07 Spirit09 wrote:Show nested quote +On January 25 2014 06:06 Mendelfist wrote:
That being said: That korona hasn't seen any changes is really strange. I wonder about this. korona's data so far is only looking at master-GM and master-diamond correct? What if they think these boundaries are ok, much of the community would probably agree here. So we still need to wait to see if they changed only the lower leagues. My personal view is that its gold down where there are the serious problems. My further personal view is that this isn't a good way to deal with those problems. So we just need to wait for the mmr stats to roll in I assume. From reading between the lines on bnet it seems they've shifted boundaries but conservatively, and they'll possibly be shifting more. Everything would be a lot easier if they just told us exactly what they've been doing.
There was only enough data to be able to calculate diamond-master & master-gm offsets. But based on the user graphs the other offsets were the same or very close. Considerable offset changes are almost immediately reflected on many user graphs (the graphs become erratic).
But now ~10 hours ago we finally got the changes we were waiting for (were immediately reflected on the user graphs).
|
2014-01-25: Blizzard made ladder changes --> MMR calculations values and league borders may be incorrect until new league offsets are calculated
When v. 2.1 was published, it was expected that Blizzard would make league offset / threshold changes, but it seems that there was none. But about 10 hours ago Blizzard made ladder changes to all servers (likely league offset/threshold changes). As a result of these changes many players are expected to be promoted to higher league after they win one match.
As the new offsets & thresholds are unknown, the MMR calculations & league borders will likely be incorrect for matches played after the changes. Lots of data needs to be collected before it is possible to calculate the new offsets. This may take 2 to 3 weeks. When the new offsets are calculated & published, the MMR calculations will be automatically fixed.
Please use the MMR tool even if the calculations are likely to be incorrect for couple of weeks. Lots of match data is needed to be able to calculate the new offsets.
--
Also note, that after SC2 v. 2.1 the MMR tool requires the latest version of Sc2gears. If you did not update Sc2gears already couple of days ago, please do it now (to update Sc2gears go to its 'Help' menu and choose 'Check Update').
|
![[image loading]](http://i.imgur.com/LiYcbbs.png)
Felt crazy.
Of my seven past losses (two to the same guy), two of them got promoted after my match to Platinum League. I'm gold past two seasons.
My first win after those 13 losses (last 13 real dots there) in a row promoted me
|
On January 25 2014 19:46 Danglars wrote:+ Show Spoiler +Felt crazy. Of my seven past losses (two to the same guy), two of them got promoted after my match to Platinum League. I'm gold past two seasons. My first win after those 13 losses (last 13 real dots there) in a row promoted me
Yes. You can be promoted only if you win a match (placement matches are exception to this rule).
Btw. I suggest you to turn off the moving average as it usually makes analyzing the graph quite useless + also makes it harder to spot possible incorrect values.
|
Ran some offset calculations for games played after Blizzard made league offset changes some 40 hours ago (little less than 4000 matches by 574 accounts). New diamond-master offset seems to be from 400 to 500 (I have more precise idea than that. The old value was 250). New master-grandmaster offset is likely 73 or something quite close (73 is Blizzard's 'magic number'. Old value was ~ from 400 to 420).
For couple of hours I tried out finding the other offsets based on user graphs, but did not end up with values that I would be satisfied with. Let's say that the current guesses are bronze-silver: close to old value, silver-gold: much smaller than old value, gold-plat: close to old value, plat-diam: smaller or much smaller than old value. I guess it is better to just wait for more data, so there would be more calculation results for other offsets (potentially will try again in next weekend if I have time).
|
On January 26 2014 22:43 korona wrote:
Ran some offset calculations for games played after Blizzard made league offset changes some 40 hours ago (little less than 4000 matches by 574 accounts). New diamond-master offset seems to be from 400 to 500 (I have more precise idea than that. The old value was 250). New master-grandmaster offset is likely 73 or something quite close (73 is Blizzard's 'magic number'. Old value was ~ from 400 to 420).
For couple of hours I tried out finding the other offsets based on user graphs, but did not end up with values that I would be satisfied with. Let's say that the current guesses are bronze-silver: close to old value, silver-gold: much smaller than old value, gold-plat: close to old value, plat-diam: smaller or much smaller than old value. I guess it is better to just wait for more data, so there would be more calculation results for other offsets (potentially will try again in next weekend if I have time).
![[image loading]](http://i.imgur.com/xuqXP61.jpg)
It looks like the re-enabled showing points lost. IIRC you said it made things more difficult when they were not shown. Does this help people get "GOOD" game data post 2.1?
http://us.battle.net/sc2/en/profile/362351/1/JustRageQuit/matches#filter=solo
It shows it on the website tooo!
|
|
|
|
|
|
EPT![[image loading]](http://postimg.org/image/wnvqwzidb/)
