|
I'm having a problem with the Plugin;
+ Show Spoiler +4:51:8 MMR-Stats 5.6 4:51:8 --------INIT START-------- 4:51:8 OK Encoding: OK 4:51:8 OK Loading DB 4:51:8 UPDATE: New DB avalable! 4:51  Downloading new DB File... 4:51 OK Downloaded new Database! 4:51 Ok Loading DB 4:51 OK Settings loaded! 4:51 ERROR No Data Found! Time to ladder... 4:51 OK Triggers ready! 4:51:10 OK Gameobserver: ready! 4:51:10 OK Timeclient: Systemtime offset: 4 sec 4:51:11 OK: Program is up to date 4:51:11 --------INIT END-------- 5:28:58 -------------------------- 5:28:58 Game Start detected 5:29:47 ERROR Webgrabber: Cant catch Site giveup! 5:29:47 Webgrabber: Player false 5:29:47 No Profile, skip 5:29:47 Webgrabber: Opponent false 5:29:47 PRE-Trigger done 5:54:17 -------------------------- 5:54:17 Game Start detected 5:55:5 ERROR Webgrabber: Cant catch Site giveup! 5:55:5 Webgrabber: Player false 5:55:5 No Profile, skip 5:55:5 Webgrabber: Opponent false 5:55:5 PRE-Trigger done
Other information;
+ Show Spoiler +Windows 7, 64bit
Java 64bit, I belive 32bit was required for the memory scan, but it should work without
sc2gears 12.1
started sc2gears as administrator and without
Diagnostic tool shows everything set to OK, no fails or errors
and ofc, MMRplugin 5.6 version
(Having Ðallan account selected and played on it aswell)
Can my firewall affect this profile look-up? or is it that my profile names has invalid characters perhaps?
Profiles;
+ Show Spoiler +
|
On August 16 2012 10:00 skeldark wrote:
I have no idea what happens if its a tie, never thought about it oO
But the value can only be 1 or 0.
I think a tie can be dropped like it was a custom game, because the MMR shouldn't change anyways.
|
On August 16 2012 11:13 korona wrote:Show nested quote +On August 16 2012 10:00 skeldark wrote:On August 16 2012 09:49 korona wrote:Now as games.dat structure was mentioned couple of posts ago, I would like to know what does the 9th value in the new format (v.5.1->) mean? Also is there a value for tie games (1 = win, 0 = loss)? Here are the formats as I understood: Old format (<- v. 5.0): + Show Spoiler +1) tool version, 2) db version, 3) long time in millis, 4) server, 5) result: win = 1 loss = 0, 6) map, 7) player data: ok = 1 bad = 0, 8) profile URL, 9) profile id, 10) league, 11) division name, 12) wins after, 13) losses after, 14) Points, 15) Unspend Bonuspool, 16) max bonuspool (master), 17) adjusted points, 18) point change without bonus, 19) player data: ok = 1 bad = 0, 20) profile URL, 21) profile id, 22) league, 23) division name, 24) wins after, 25) losses after, 26) Points, 27) Unspend Bonuspool, 28) max bonuspool (master), 29) Adjusted Points, 30) point change without bonus, 31) nothing, 32) user id in tool, 33) season New format (v 5.1->): + Show Spoiler +1) G, 2) tool version, 3) db version, 4) long time in millis, 5) Server, 6) Map, 7) user id in tool, 8) season, 9) ???, 10) result: win = 1 loss = 0, 11) profile URL, 12) player data: ok = 1 bad = 0, 13) profile id, 14) league, 15) division name, 16) main race, 17) race, 18) wins after, 19) losses after, 20) Points, 21) Unspend Bonuspool, 22) max bonuspool (master), 23) Adjusted Points, 24) point change without bonus, 25) profile URL, 26) player data: ok = 1 bad = 0, 27) profile id, 28) league, 29) division name, 30) main race, 31) race, 32) wins after, 33) losses after, 34) Points, 35) Unspend Bonuspool, 36) max bonuspool (master), 37) Adjusted Points, 38) point change without bonus I tinkered with games.dat just couple of days ago. Out of curiosity I manually added some crash games that had not been logged in between as "bad games" (with correct time values, results (win-loss) and player id:s). The additions did nicely fit in the graphs and the analysis based on the originally recorded values did not change (only some finetuning for individual values). The algorithm only care about win or loose on bad games ( some data must be valid but cant tell from memory what exactly there are to many datacheck functions) The game-time is not used at the moment (bonuspool is already calculated) To get everything in the same new format click re-save games in Config-Menu. I have no idea what happens if its a tie, never thought about it oO But the value can only be 1 or 0. What you posted is correct. The "G" is just a marker for the new data format. The 9) ??? is the gametime in seconds. The timestamp is unix time = milisec past since 0:00 1.1.1970 All data is from the posttrigger ( after the game) Notice that the line-breaker is "\r\n". Independent of your system! The file is UTF-8 encoded. The new gameformt: (Sorry im to lazy i just copy paste the output code): + Show Spoiler +text += "G,";
text += gamedata.progversion + "," + gamedata.dbversion + ",";
text += gamedata.time + "," + a.server + "," + a.mapname.toLowerCase() + ",";
text += gamedata.id + "," + gamedata.season + "," + gamedata.gametime + "," + won + ",";
text += a.bnetlink + "," + aok + ",";
text += a.id + "," + a.league.toLowerCase() + "," + a.division_name.toLowerCase() + "," + a.race_main.toLowerCase()+ ","+ a.race.toLowerCase() + ",";
text += a.game_wincount + "," + a.game_loosecount + "," + a.points + "," + a.bonus + "," + a.totalbonuspool + ",";
text += a.adjusted_points + "," + a.change_points + ",";
//
int bok = 0;
if (b.OKmmr) {
bok = 1;
}
text +=b.bnetlink + "," + bok + "," ;
text += b.id + "," + b.league.toLowerCase() + "," + b.division_name.toLowerCase() + "," + b.race_main.toLowerCase()+ ","+ b.race.toLowerCase() + ",";
text += b.game_wincount + "," + b.game_loosecount + "," + b.points + "," + b.bonus + "," + b.totalbonuspool + ",";
text += b.adjusted_points + "," + b.change_points + ",";
//
text += "\r\n";
Thanks! The field I had not understood was indeed: 9) Length of the game in 'Blizzard seconds' (game speed) Last tie game I had was on 2012-06-07 (the only tie game while using the tool. Tool version was 3.6.2 & db 7.2). That match was missing from the games.dat, but I found a its data from the log.txt. The post trigger had read my correct info from the web server (also had change_points: 0). But there had been some problem regarding reading the opponents info. His wins, loses, points, bonus & adjusted points had not been read/calculated (all -9999). His league & division had been read. The post trigger had decided: "ERROR Bad Game!" I can PM the log of that match, if needed.
0 changepoint is not accepted. So the game is bad. ( have to do that many mistakes on bnet server lead to 0 changepoints)
But the analyser thinks : not won = loose
Have to change that.
On August 16 2012 13:07 Malpractice.248 wrote:Show nested quote +On August 16 2012 07:34 skeldark wrote:
BIG, BIG BUG!
This season did not end in time.
Over 720 bonuspool already and my tool thinks that 720 is maximum bonus pool because of an seasonend.
So it goes into "season cap mode".
All latest data is wrong because of this.
I updated the database file ( its automatic updated when you start the program )
Should be ok now. Sorry for the mistake oO
Thanks for pointing the wrong data out guys.
Sorry i dont play much sc2 anymore and did not look in the data for some time,. or i would notice myself. So... i just loaded the mmr tool; and normally theres a "download new version" button, which there isnt atm... So maybe it installed automatically, as you said, however, im assuming its not retroactive? since I still have a drop of 46 pts vs someone who was above me >.> And as i look at it more, (about 2/3 the time) i lose to someone with higher MMR, i lose between 20-50 MMR... Sooooo idk. (However, if i BEAT someone with much higher MMR, I gain 0-5 MMR...) Also, since i hadnt played in like 3 seasons, it started me at 2k MMR exactly (i only did the 1 placement each season) and idk if thats accurate, but who knows. ha ha. Also just exported data, if that gives u my data, im not sure.
Yes. A result of the wrong bonuspool. He calculate changpoints diffrent on win and loose so wrong bonuspool affect win and loose diffrent. Its fine from now on. I updated the database not the client. The maximum bonus pool for each season is stored in the database not directly in the program code.
If you have statistic upload enabled i get your data.
The export function is only if you want to export your mmr value.
On August 16 2012 16:29 BurningRanger wrote:Show nested quote +On August 16 2012 10:00 skeldark wrote:
I have no idea what happens if its a tie, never thought about it oO
But the value can only be 1 or 0. I think a tie can be dropped like it was a custom game, because the MMR shouldn't change anyways.
Custom game drop only happens if the posttrigger dont run at all or the memory id is not equal to settings id ( more than 2 players and your playernumber is bigger than 2 )
The problem is the code will detect no win, and no win is a loose for him.
I will change it
@DLN.sc2
1) pls post logs in spoilers.
2) try to enter your profile this way:
http://eu.battle.net/sc2/en/profile/924220/1/Ðallan/
http://eu.battle.net/sc2/en/profile/2684368/1/DLNǂUndrcovR/
3) set your replay folder in sc2gears. THe posttirgger dont run so the replay store function of sc2gears is not set up correct
4) Firewall is ok i think because the log say you downloaded the new database file without problems.
|
AFAIK ties are ignored by the ladder system.
On August 16 2012 05:32 Cambam wrote:
Ok, so I looked at the "games" excel file and looked only at masters players who had spent all their bonus pool (720 in the file I'm looking at). I figured that once you've spent all your BP, your points should basically just be your MMR minus a constant. However, upon looking for this constant, I found it varied a lot more than I expected. It's approximately 1217 (therefore, MMR = points + 1217), but it varied all the way from 1116 to 1300. This leads to weird situations like: one player has 1498 points but has a higher MMR than another player with 1542 points, though both have spent all their bonus pool. Another example is a player with 764 points with a lower MMR than a player with 711 points. Lastly, there is a slight positive correlation between a player's points and the value of his constant, meaning that players with higher points tend to have higher constants.
What are your thoughts on this? Shouldn't points match up with MMR pretty nicely after BP has been spent? What's with the correlation between points and the constant?
Adjusted points have a force that acts upon them pushing them towards your performance MMR (i.e. if you beat 1500 MMR opponents 50% of the time, your points get pushed towards 1500). The bigger the difference between your adj points and performance MMR, the stronger your adj points get pushed. This is due to how the ladder system awards points.
On the other side MMR and adj points constantly get desynced with one another. MMR changes on average by 16 points whereas adj points closer to ~12. In general adj points tend to lag behind MMR changes.
Look at this image:
![[image loading]](http://s14.postimage.org/bdk5bte9d/astraflame.png)
This is a player who's going rapid MMR increase in some ~140 games. Blue line being adj points Red line being MMR approximation. Adj points fall behind until the MMR is sufficiently above them for the force that pushes them towards MMR is strong enough for them to stop falling behind. That happens when you start getting ~16 adj points per game, or looking at F function, when your adj points are roughly 105 to 136 points away points away from your MMR (plus league offset, naturally).
To this you need to add the fact that everyone's adj points start desynced with their MMR at the beginning of each new season, when adj points are reset to 0. You used players who finished their bonus pool as an indicator of when their adj points 'caught up', but using F function, you can find a better indicator. Let's look at an extreme case of a Master player who has 2920 MMR at the beginning of a new season. This player adj points 'want' to be at 1000, but are 0 (new season). Assuming the player receives 16 MMR for a win and losses 16 MMR for a loss, we can calculate how many games it will take for the adj points to catch up. For each win the adj points increase by 24 and catch up with the MMR by 8, and each loss decrease by 0, catching up with MMR by 16. This will go on until the adj points are sufficiently close to the MMR when the player will start receiving 23 points per win and losing 1 per loss, which happens at 668 points difference.
This is an interesting thought experiment, but you will always have cases of players who are like the one in the picture (and the opposite case). Also just because that is an extreme case does not mean similar things do not happen for more 'normal' players. And do not forget about promotions / demotions as well, which reset adj points to 73.
BTW throw away the 1217 number. The number you're looking for is 1920 which is our estimate of Master league offset. The 1217 number is actually 1920 - 720. Never use actual ladder points in ANYTHING that you do, they will just add confusion. Always use adjusted points instead.
|
@DLN.sc2
1) pls post logs in spoilers.
2) try to enter your profile this way:
http://eu.battle.net/sc2/en/profile/924220/1/Ðallan/
http://eu.battle.net/sc2/en/profile/2684368/1/DLNǂUndrcovR/
3) set your replay folder in sc2gears. THe posttirgger dont run so the replay store function of sc2gears is not set up correct
4) Firewall is ok i think because the log say you downloaded the new database file without problems.
[/QUOTE]
I actually happen to have 4 different sc2 clients for 4 different regions, NA, EU, KR, Taiwan, that was probably the reason why it didn't recognize the folders, I try to fix it now
Thanks
|
|
|
I have a question
the last 7 games, my opponent always was 10 MMR above me, is it because the plugin is having trouble finding their MMR
Here is the log
+ Show Spoiler +18:11:48 Game Start detected 18:11:48 OK Found Memory-Profile in Active-Profiles! 18:12:31 Webgrabber: Player true 18:12:33 Webgrabber: Opponent true 18:12:33 PRE-Trigger done 18:30:32 Webgrabber: Player true 18:30:35 Webgrabber: Opponent true 18:30:35 Adjused Calculate Player:true 18:30:35 Adjused Calculate Opponent: true 18:30:35 Gamecheck: true 18:30:35 POST-Trigger done 18:30:35 Player OK?: MMR: true Cap: true Dev.: true 18:30:35 Opponent OK?: MMR: true Cap: false Dev.: true 18:30:35 Player-data-Qualitiy: 100% 18:30:35 Opponent-data-Qualitiy: 75% 18:30:37 OK Uploader: game uploaded 18:30:37 Ladder Game Done!
|
so your update isnt retroactive? so my mmr is higher than shown atm?
|
On August 17 2012 07:35 Maxamix wrote:I have a question the last 7 games, my opponent always was 10 MMR above me, is it because the plugin is having trouble finding their MMR Here is the log + Show Spoiler +18:11:48 Game Start detected 18:11:48 OK Found Memory-Profile in Active-Profiles! 18:12:31 Webgrabber: Player true 18:12:33 Webgrabber: Opponent true 18:12:33 PRE-Trigger done 18:30:32 Webgrabber: Player true 18:30:35 Webgrabber: Opponent true 18:30:35 Adjused Calculate Player:true 18:30:35 Adjused Calculate Opponent: true 18:30:35 Gamecheck: true 18:30:35 POST-Trigger done 18:30:35 Player OK?: MMR: true Cap: true Dev.: true 18:30:35 Opponent OK?: MMR: true Cap: false Dev.: true 18:30:35 Player-data-Qualitiy: 100% 18:30:35 Opponent-data-Qualitiy: 75% 18:30:37 OK Uploader: game uploaded 18:30:37 Ladder Game Done!
Thats only filler points. 10 points above = i have no idea what his mmr is.
Its 10 above because you can not see it if its your mmr.
If your MMR rise you climb the "hierarchy". If you get a nice icon from blizzard for that or not.
But you get your dia icon if you get stable above the promotion line. For promotion you have to be stable.
So if you keep rising you dont get promoted. That was blizzards idea not my!
Just dont give a XXXX about the icon and you are fine 
BUT read why i answered Malpractice
On August 17 2012 07:59 Malpractice.248 wrote:
so your update isnt retroactive? so my mmr is higher than shown atm?
Yes. No.
I will correct old games with an update soon ( i will just mark them bad)
But here comes the bad news. I calculated to less bonuspool. So you have to high adjusted points.
So the real mmr is lower than shown atm....
|
 I dont see how, when im losing 30-50 mmr per loss vs higher MMR, but only gaining 1-5 per win >.> ha ha. Oh well.
|
Also, how come some dots are filled in, while others are not?
|
On August 17 2012 12:05 Malpractice.248 wrote: I dont see how, when im losing 30-50 mmr per loss vs higher MMR, but only gaining 1-5 per win >.> ha ha. Oh well.
Mhh good question. Its very complicated. The bonuspool error set the adjusted points to high but on the other hand he screw with the changepoints.
i download the new dataset and look what happens to your graph with my fix.
On August 17 2012 12:07 Malpractice.248 wrote:
Also, how come some dots are filled in, while others are not?
filled = won, not filled = loose
EDIT: in the end result not much change. The graph looks obvious way better but in average no big change.
But with last 20 games bad games your graph will be very unstable. You did not play a game with the new database file?
|
Okay, so heres a link of what im talking about
http://www.freeimagehosting.net/xjqe9
I did purple circles on major spots that are huge jumps (30+ points) (usually down, naturally, lol)
The red circles, is where i dont understand the "Filled = win, Empty = loss"
As you can see in the middle, i lost 1 (dropped) Won 3 (...Went up, then down.. then back up..) Then lost one, and went back down..
Its showing wins, yet losing MMR...
Im just confused lol
But i love u for this program :p
|
On August 17 2012 12:27 Malpractice.248 wrote:Okay, so heres a link of what im talking about http://www.freeimagehosting.net/xjqe9I did purple circles on major spots that are huge jumps (30+ points) (usually down, naturally, lol) The red circles, is where i dont understand the "Filled = win, Empty = loss" As you can see in the middle, i lost 1 (dropped) Won 3 (...Went up, then down.. then back up..) Then lost one, and went back down.. Its showing wins, yet losing MMR... Im just confused lol But i love u for this program :p
Like i said all this data is wrong because of the wrong bonus pool calculation
Thats your graph with my fix: gray games are bad games = assumptions
+ Show Spoiler +
You did not play a game after i fixed the database file?
|
No i havent played yet :p been busy, about ot in about 5 min tho :p
|
Also, does it just "give" the MMR of 2k when i first started, since i hadnt played in a few seasons? I mean, i was high masters few seasons ago :p
I dont really mind, though. I just use the tool to see progress Figured id lose a bunch coming back
|
5.7 out.
Markes games that are wrong because of the database-bug as bad games!
On August 17 2012 12:39 Malpractice.248 wrote:Also, does it just "give" the MMR of 2k when i first started, since i hadnt played in a few seasons? I mean, i was high masters few seasons ago :p I dont really mind, though. I just use the tool to see progress Figured id lose a bunch coming back
No if you played your placement matches he remember your mmr.
He dont reset it to a special amount as far as we know.
|
Ill start playing I just gained 10 mmr :p ha ahaa
But only 1 is grey :p allsssssss good!
|
On August 17 2012 12:43 Malpractice.248 wrote:Ill start playing I just gained 10 mmr :p ha ahaa But only 1 is grey :p allsssssss good!
restart sc2gears and update to 5.7
To see assumptions you have to enable this options in the setup menu.
If not enabled you only see the last game as assumption. This one is allways a "fake" game because i calculate the MMR of the LAST game. So i show one "fake" game at the end with the assumption of game you just played.
|
I didnt get a setup menu after i restarted :p Its alllll good
|
|
|
|
|
|
EPT![[image loading]](http://i.imgur.com/krKOA.png)
