EPTMatch Making Rating Tool - Page 164
| Forum Index > SC2 General |
|
korona
1098 Posts
| ||
|
Strahl
United Kingdom6 Posts
Should the line not follow the opponents' dots more or less? This would make more sense to me as I am rank 3 plat currently, beating diamonds sometimes, so my line should be near the top of plat not hanging around at the bottom? | ||
|
korona
1098 Posts
On August 15 2013 09:32 Strahl wrote: Thanks for the quick reply. I've read about the capped opponents, however if you look at my graph, none of my points on the graph are near my opponents points like they used to be before I deleted everything and re-started. To re-itterate, my line is no where near my "good game" opponents. Should the line not follow the opponents' dots more or less? This would make more sense to me as I am rank 3 plat currently, beating diamonds sometimes, so my line should be near the top of plat not hanging around at the bottom? No. Even if opponent's usually have close MMR, their MMR can differ a lot from yours (there can be rare data errors also regarding opponents' MMR). Note that if opponent's MMR cannot be calculated, his MMR is shown 10 over yours in the graph. Also division ranks mean nothing. In a hypothetical division you can be rank 1 and have lower league MMR. And those higher league opponents can have low MMR too as demotions do not happen during the season (only in placement match). Your adjusted points from your last match also suggest that you are likely slightly higher than the platinum line (just like the last estimate MMR value suggests). Of course this is a generalization based on adjusted points (MMR can differ considerably from what adjusted points suggest). Edit: It is also recommended to turn the memory reader on (you likely have it off as opponent's change points have not been calculated for lots of matches). | ||
|
Strahl
United Kingdom6 Posts
| ||
|
korona
1098 Posts
On August 15 2013 10:04 Strahl wrote: Okay, thanks again for the response. My memory reader is on, has been the whole time :/ | ||
|
Val_
Ukraine156 Posts
 ![[image loading]](http://i.imgur.com/CnbQ26S.png) gj authors nice app | ||
|
mythandier
United States828 Posts
Or do I have to wait for auto-uploaded new games to populate the data. | ||
|
korona
1098 Posts
On August 17 2013 07:47 mythandier wrote: Just installed...can I not incorporate all the data from games prior to installation? Even if they are from the current season? Or do I have to wait for auto-uploaded new games to populate the data. You cannot as we don't have enough data for those games. Yes. You need to keep MMR tool on while playing to record new data (auto-upload is not required, but by default on. Uploaded info is kept private and used to e.g. calculate better league offsets). (Technically it is possible manually add 'estimate games', but too complex to explain how. And they would just be estimates based on values from games recorded after starting to use MMR tool). | ||
|
mythandier
United States828 Posts
On August 17 2013 08:04 korona wrote: You cannot as we don't have enough data for those games. Yes. You need to keep MMR tool on while playing to record new data (auto-upload is not required, but by default on. Uploaded info is kept private and used to e.g. calculate better league offsets). (Technically it is possible manually add 'estimate games', but too complex to explain how. And they would just be estimates based on values from games recorded after starting to use MMR tool). Thanks for the great explanation. I used to play a ton but have had to limit my play time lately so I was just trying to be lazy. I'll keep in on as I play from here on out. | ||
|
vaderseven
United States2556 Posts
Specifically, I was recently re-promoted to masters and as such have a very high bonus pool. I went through a reformat during that same time frame. As such, every game in my mmr tracker atm is post promotion. I have played maybe 18 games and still have well over 550 bonus pool since the promotion. Does any of that information skew the MMR reading in one direction or another? I know that combination of factors (especially the low number of games played in the tracker) will create an inaccurate reading for at least a few more games but I wanted to know if the inaccuracy should (i.e. doesn't have to but from a statistics viewpoint should) tend in one direction or the other. Of note is that I do have the memory reader on and all but two of my games have been vs other masters players. **INSERT VERY HIGH PRAISE FOR THIS PLUGIN HERE** | ||
|
korona
1098 Posts
On August 15 2013 09:26 korona wrote: Season lock is supposed to happen next Monday for all servers (also for China). I likely lock the tool DB during Sunday & re-enable it on Monday when final max bonus pools are known (& it is certain that the servers are locked & there are no display errors). When the tool DB is locked match data for new matches is automatically set 'bad'. Those matches are used as estimate matches. At this moment ladders may have already been locked for Korea, China and SEA. Ladder lock for EU is supposed to happen in 5 hours and in US in some ~13 hours (did not check timezones so this may be off hour a two). Thus I just published a new Bnet DB file that locks match data for all servers. Please restart sc2gears to load it. After this match data for all future matches is set 'bad' automatically (these will be used as estimate matches). When ladder lock is in effect for all servers some time tomorrow Monday, I will find out the final bonus pools & check if there are display errors and after that will publish new DB file that unlocks the data. | ||
|
korona
1098 Posts
On August 18 2013 13:30 vaderseven wrote: Does having a high or low bonus pool effect the accuracy of this plugin? Specifically, I was recently re-promoted to masters and as such have a very high bonus pool. I went through a reformat during that same time frame. As such, every game in my mmr tracker atm is post promotion. I have played maybe 18 games and still have well over 550 bonus pool since the promotion. Does any of that information skew the MMR reading in one direction or another? I know that combination of factors (especially the low number of games played in the tracker) will create an inaccurate reading for at least a few more games but I wanted to know if the inaccuracy should (i.e. doesn't have to but from a statistics viewpoint should) tend in one direction or the other. Of note is that I do have the memory reader on and all but two of my games have been vs other masters players. **INSERT VERY HIGH PRAISE FOR THIS PLUGIN HERE** Bonus pool has no effects regarding accuracy. The tool always removes bonus point portion from the match change points. Blizzard shows combined points in match history. In many cases pre-match data is needed to solve _real_ change points. This is also the reason why change points can be solved only in limited situations if memory reader is off. Low number of games played in tracker is neither a reason for possible inaccuracies. If you get a good reading & you and your opponent have played enough games after blank MMR (MMR changes become 'normal after some 25 to 40 matches after starting from blank MMR) & there are no data errors, then the first good result can be accurate. But of course when you have more good readings you can be more sure that there were no problems regarding the first reading. (As a side note: Like you saw in my last post I just locked the DB for all servers. This is needed to make sure that the bonus points are correctly deducted when change points are calculated. We will know the final max bonus pools only after ladders have been locked for each server. And when the max pools are known the DB lock can be released) | ||
|
Aurik
Germany12 Posts
do u have an Idea, why not a single game of 180+ games is set as a good game for my opponents? For me there are about 15/180 games "good games", but as i said, not a single one for my opponents. In addition, there is a very large spike in my graph: ![[image loading]](http://i.imgur.com/pPeDQcR.jpg) Is this large spike correct? If so: Why was the data so wrong bevor this one game? Ty for answer and ur work here, its a nice and helpfull tool! Aurik ps.: There are "good games" for my opponents, but this was bevor the new database at start of S4 was released | ||
|
korona
1098 Posts
On August 19 2013 22:20 Aurik wrote: Hi, do u have an Idea, why not a single game of 180+ games is set as a good game for my opponents? For me there are about 15/180 games "good games", but as i said, not a single one for my opponents. This is easy to answer. Based on the graph you are in platinum, but your MMR has been dropped into gold range. You are 'capped' (Blizzard artificially boosts points gained and lessens points lost). Opponent's MMR is calculated from your data and it cannot be done reliably. Notice that in the early games you were still in platinum range & opponent MMR could often be calculated. Read more e.g. from http://www.teamliquid.net/forum/viewmessage.php?topic_id=334561¤tpage=162#3226 On August 19 2013 22:20 Aurik wrote: In addition, there is a very large spike in my graph: + Show Spoiler + Is this large spike correct? If so: Why was the data so wrong bevor this one game? Ty for answer and ur work here, its a nice and helpfull tool! Aurik ps.: There are "good games" for my opponents, but this was bevor the new database at start of S4 was released Incorrect spikes sometimes happen. It may be a slight problem in cap detection or e.g. due data error (there may be other reasons too). Even 1 point error in change points can shift the result considerable. And if the match ended near bonus point tick time, such calculation errors sometimes (quite rare) happen even if there are safeguards. One possibility also is that you had not downloaded the latest bnet DB file that locks data for all servers (sets match data automatically 'bad'). The ladders were supposed to be locked during last night, which means that bonus pool stopped growing. The tool thinks the bonus pool is still growing & calculates the change points incorrectly. This is why all data is now set automatically 'bad' (can be used as estimate games). Later this evening I will find out the final max bonus pools for all servers & check if there are display bugs (this was very common during WoL that they showed incorrect bonus pool either on web profile or in-game profile during the ladder lock periods). When I know the final bonus pools I will publish new DB file & unlock the DB. Note that these 'spikes' are easy to spot & most are upwards. When you spot such match you can set it bad. | ||
|
korona
1098 Posts
The final bonus pools for S14 were: For EU, US, KR and SEA: master+ 1825, diamond- 1058 For China: master+ 874, diamond- 507 Please note that the max Chinese master pool could be incorrect. It is either 874 or 875, but I cannot check this for sure as I have not found any Chinese WoL master profiles (from Chinese HotS profiles you cannot do this yet for this season, next season you can). But as diamond (or lower) pool is known (507) the master pool can be calculated. And as last season max pools for other servers were master+ 874 and diamond- 507, it is safe to presume master pool is 874 (and not 875). This possible 1 point error only affects if either player depletes his pool during a match. Note this only affects Chinese server. | ||
|
iLevitate
United States225 Posts
| ||
|
KillerofGods
United States34 Posts
On August 23 2013 11:26 iLevitate wrote: Can someone kindly show me how to set this up? I'm not sure what to do.. I tried to run "ram.exe" but it crash every time.. This is a plugin for sc2gears so you have to download that first. | ||
|
korona
1098 Posts
Hopefully on Monday every server has switched to season 15 and Blizzard updates the web profiles. When all web profiles are updated, I will calculate season start times & release new DB file that unlocks the data. Please note that sometimes it takes long time for Blizzard to update the web profiles, so it may take longer. EDIT: Now at Monday evening US and EU web profiles are still in old season. But Korean & SEA profiles have been 'half' switched to new season. For example season number is missing from the ladder profile pages. This may require new version of the MMR tool, if this change is applied to all servers. But will have to wait and see what happens regarding EU & NA profiles. At the moment SEA profile api returns sometimes S14 data and sometimes S15 data, so it also might just be a temporary profile site error too. Edit 2: Now at midnight EU web profiles have been updated & miss the season number too. EU api returns sometimes S14 and sometimes S15 data. US profiles are still in S14 but api returns S15 data. --> Will need new version of MMR tool. Already did most of changes & now resolve the season number via api. But it will be released at earliest late Tuesday evening EU time when I have time to test it in practice. Hopefully US profiles have been fixed by then too... | ||
|
Fennixx85
2 Posts
| ||
|
Wilko
Germany470 Posts
Really appreciate your work  | ||
| ||
