EPTMatch Making Rating Tool - Page 47
| Forum Index > SC2 General |
|
Gyaru
Sweden16 Posts
| ||
|
meijin
United States141 Posts
| ||
|
skeldark
Germany2223 Posts
On June 18 2012 05:45 Excalibur_Z wrote: Yeah and I just wanted the tiers in for analytical purposes so we could compare the offsets and see a player get promoted by stabilizing within a tier. If that's not going to be accurate and there are better ways to error-check the offsets then that's fine, the tiers don't have to be shown. I do think it would make explaining the existence of tiers easier though (since you would actually be able to see them on a chart) but it's no problem. I dont think you have to. I mean if someone wants to know what it is yes, but in the end the people want to see their skill. And tiers dont help here. So i think this is more something for us than for the normal player. There is no need that every player understand the league system. they want to know their skill they look at my graph. If they are interested in math or statistic they can read all our articles and look at our data. On June 18 2012 05:45 Cheshyr wrote: Ahh, I misunderstood. I think we're running into a difference in needs between people using as a gaming tool and people using it as an analytical tool. I can definitely see why Excalibur would want tier info, even if that info is novelty for me. yes and i dont think this program is a good analytical tool. I put in the results of our analytic into it. So if you analyse with this you run into danger to proof possible mistakes, with the mistake itself, true. I published the playerdata in the season 7 dataset that includes all the analysed mmr. To find out something about promotions this is the better way. On June 18 2012 06:37 Gyaru wrote: Hey, i was wondering about the E+/F analyzer. According to E+ analyzer my mmr is 1825, and according to the F analyzer 1987. Which one is more accurate? :p It feels weird with the F one, as according to it i am only 6 mmr points from master league, however i almost only face diamond opponents. That being said, my understanding of the league system isnt the best, other than mmr decides which league you are in etc. Sounds intresting. In generell f is better but can be that in your case f makes a mistake and e+ not. Can you give me your profile link i try to check it. On June 18 2012 11:04 meijin wrote: just had something slightly odd. i've been tracking my last 30 games and everything seemed to work perfectly but when i lost game #26, it showed my mmr going down, then when i won game 27, the graph changed and stayed level. after i won game #28, my mmr went down. Nice! I know that sounds bad for you. Sounds wonder-full for me as long as the mmrloose on the win is not more than 20-30 points It sounds good because this is your only complain and you dont have 100 point jumps or? Also the dots are the result before the game! thats why i add a fake point in the end! Mmr always increase if you win. However there is a deviation. So lets say: game 1 L 1000 MMR +- 24 game 2 W 980 MMR +-24 game 3 W 970 MMR +-40 so in real your mmr is = 1 L 1012 2 W 996 3 W 1010 Notice that in longer rows this mistake will equal itself out. It can make single dots wrong but will not affect the hole graph. Its an local mistakes. global mistakes exist too but you notice this one when he is unable to correct them. Lets say i show you to low for 6 games in a row. Sooner or later you play a game with a high minimum value. The analyser notice its impossible that you are as low as he thought in this game. So he will re-judge all games with this new information. . If there is still a very big jump, this means it is impossible to correct the mistake. He will complain in the log that he found a mistake he can not correct. This can mean that a) the offsets are total wrong /they are not b) one of the games is wrong and show wrong values that confuse the analyser c) you played not recorded games between this games | ||
|
vaderseven
United States2556 Posts
Thank you again. | ||
|
skeldark
Germany2223 Posts
I notice something in user skill that is well known. This information is the key why Blizzard invented the hole fake points system. No one change! You are at a level of mmr and most likely stay there forever... This is how skill works in everything. You learned what you can and stay at a level and increase your skill slowly but not more than the other guys on the server. Remember = stay at same mmr means you improve ! To raise you have to improve more then the average guy at your level! So this sounds bad. But its not. Lets say you loose 20 games in a row. You think you suck and it will be hard to comeback. But in real you just had a bad day and you will easy climb back to your real level. Knowing that, make ladder more relaxing and any kind of improving over longer time a bigger success. | ||
|
skeldark
Germany2223 Posts
On June 18 2012 06:37 Gyaru wrote: Hey, i was wondering about the E+/F analyzer. According to E+ analyzer my mmr is 1825, and according to the F analyzer 1987. Which one is more accurate? :p It feels weird with the F one, as according to it i am only 6 mmr points from master league, however i almost only face diamond opponents. That being said, my understanding of the league system isnt the best, other than mmr decides which league you are in etc. omg. Your graph is absolute worst case! E+ is right and F wrong. I searched for the mistake but than i found out he dont make one. You dont have a single 100% game against a master or platinum. With only the information you have the graph can be in 3 areas. Its the lowest one of the 3 i know this, because you dont play enough masters. BUT from the data it accidental fits best to the highest one. So he dont make any mistake. Its just unlucky that the errormargin fits so perfect in the wrong one. i think about an solution. He notice that all 3 areas fits good its just that the highest fits a little little bit better. On the good site this means your graph is unstable. He could easyly drop the hole graph if you have one good game against a master. EDIT: After watching all the data i think F is right... This one is strange really strange. if i hardcode some offsets one think never change. In the end he want you to be near master... EDIT2: 2 of your opponents are in the same division! thats good to see how good the analyser is. Edit2: i think you start lower than f but in the end f is right. please think about enable the memoryreader. I will watch your graph in future closely. ------------------------------------------------------------------------ Some questions and ideas that come to my mind while watching mmr data for 2 hours: 1) Who is autorage, the guy that mass ladder allways update the version but never posted here? 2) why noone commented on my easteregg yet? 3) can you all please switch to P, train 4 gate 100 times and than fast ladder up to master so i have better graphs? | ||
|
AmericanUmlaut
Germany2598 Posts
On June 18 2012 18:11 skeldark wrote: 3) can you all please switch to P, train 4 gate 100 times and than fast ladder up to master so i have better graphs? Hey, no balance whining in here  . | ||
|
Neelia
Germany599 Posts
If I want to update the plugin and click the update button in the main window I have to confirm to start the download and confirm to start the installation later, the last 2 steps seem a bit superfluous? Thank you soo much for your hard work on the plugin  | ||
|
Gyaru
Sweden16 Posts
On June 18 2012 18:11 skeldark wrote: omg. Your graph is absolute worst case! E+ is right and F wrong. I searched for the mistake but than i found out he dont make one. You dont have a single 100% game against a master or platinum. With only the information you have the graph can be in 3 areas. Its the lowest one of the 3 i know this, because you dont play enough masters. BUT from the data it accidental fits best to the highest one. So he dont make any mistake. Its just unlucky that the errormargin fits so perfect in the wrong one. i think about an solution. He notice that all 3 areas fits good its just that the highest fits a little little bit better. On the good site this means your graph is unstable. He could easyly drop the hole graph if you have one good game against a master. EDIT: After watching all the data i think F is right... This one is strange really strange. if i hardcode some offsets one think never change. In the end he want you to be near master... EDIT2: 2 of your opponents are in the same division! thats good to see how good the analyser is. Edit2: i think you start lower than f but in the end f is right. please think about enable the memoryreader. I will watch your graph in future closely. Its weird, today when i looked at the graph it had totally changed, not surpassing 1980 mmr once, and in general lower :p guess it somehow corrected itself...? Thanks for the awesome replies tho, cool to see that you reply and look at everyone individually :D | ||
|
Garf
Belgium8 Posts
Edit: or am I running into some LibreOffice XLS limitation... | ||
|
dareKEU
Spain16 Posts
On June 18 2012 18:11 skeldark wrote: 2) why noone commented on my easteregg yet? What easteregg? I cannot find it :-( | ||
|
StreetWise
United States594 Posts
| ||
|
meijin
United States141 Posts
On June 18 2012 17:22 skeldark wrote: Nice! I know that sounds bad for you. Sounds wonder-full for me as long as the mmrloose on the win is not more than 20-30 points It sounds good because this is your only complain and you dont have 100 point jumps or? Also the dots are the result before the game! thats why i add a fake point in the end! Mmr always increase if you win. However there is a deviation. So lets say: game 1 L 1000 MMR +- 24 game 2 W 980 MMR +-24 game 3 W 970 MMR +-40 so in real your mmr is = 1 L 1012 2 W 996 3 W 1010 Notice that in longer rows this mistake will equal itself out. It can make single dots wrong but will not affect the hole graph. Its an local mistakes. global mistakes exist too but you notice this one when he is unable to correct them. Lets say i show you to low for 6 games in a row. Sooner or later you play a game with a high minimum value. The analyser notice its impossible that you are as low as he thought in this game. So he will re-judge all games with this new information. . If there is still a very big jump, this means it is impossible to correct the mistake. He will complain in the log that he found a mistake he can not correct. This can mean that a) the offsets are total wrong /they are not b) one of the games is wrong and show wrong values that confuse the analyser c) you played not recorded games between this games ok i thought it was something like that. figured i'd ask to see in case it wasn't supposed to happen. | ||
|
hpty603
United States262 Posts
That just keeps blinking. I've tried it with two accounts and my replays auto save. I've made sure the links to my bnet accounts are correct | ||
|
StreetWise
United States594 Posts
| ||
|
ErrantMind
United States63 Posts
| ||
|
Rube_Juice
Canada348 Posts
| ||
|
skeldark
Germany2223 Posts
just think 4 gate is more powerfull than 3 rax or bannling bust ^^ its more the strategie than the race On June 18 2012 19:51 Neelia wrote: A small suggestion: If I want to update the plugin and click the update button in the main window I have to confirm to start the download and confirm to start the installation later, the last 2 steps seem a bit superfluous? Thank you soo much for your hard work on the plugin last version should have only step 1 and 2 i think On June 18 2012 20:19 Gyaru wrote: Its weird, today when i looked at the graph it had totally changed, not surpassing 1980 mmr once, and in general lower :p guess it somehow corrected itself...? Thanks for the awesome replies tho, cool to see that you reply and look at everyone individually :D it corrected itself with each new data. Basic the algorithm claculate what he thinks the data should look like and than try to be as close to it as possible. When its impossible to get close he make a new assumption. In short he trys alot of combination and pick the best one. On low data the wrong one can be better than the right one because the errormargins fit in the wrong one better. On June 18 2012 22:11 Garf wrote: Is it a coincidence that your Season 7 data has exactly 16384 games in it? Or did you run into the limit of some tool? Edit: or am I running into some LibreOffice XLS limitation... Its near 20k i think its your program. + Show Spoiler + Even if you leave the best behind you,it does not mean you are good. On June 19 2012 02:38 StreetWise wrote: I am in masters league with an MMR of 0? Not sure what this means? you suck! ^^ nah just need a good game to judge. before that its on 0. On June 19 2012 03:02 meijin wrote: ok i thought it was something like that. figured i'd ask to see in case it wasn't supposed to happen. allways welcome On June 19 2012 07:52 hpty603 wrote: I'm trying to use this but it gets stuck at "Post Trigger Running" That just keeps blinking. I've tried it with two accounts and my replays auto save. I've made sure the links to my bnet accounts are correct argh not again. whats the last message he output in logs before he stuck? On June 19 2012 11:25 StreetWise wrote: So did anyone else go on a massive tilt after installing this program? I think having is show 0 MMR got to me. perhaps i should show 9001 as first value? On June 19 2012 11:27 ErrantMind wrote: found the easter egg... its under 9000. ^^ On June 19 2012 12:14 Rube_Juice wrote: Is the GM promotion line semi-accurate? It didn't move from season to season change, but getting picked at GM time in a new season is a lot different than being promoted midseason is it not? no. you are right. Its the midseason one. The startseassion one is total gamble. i think 800 points above the master line is a chance. gm is a little bit fucked up. You need luck to make it in it at the start if you are not top 20 on the server | ||
|
hpty603
United States262 Posts
3:34:14 --------INIT START-------- | ||
|
skeldark
Germany2223 Posts
On June 19 2012 16:36 hpty603 wrote: There are like 4 fatal errors :o 3:34:14 --------INIT START-------- use java 32 bit and allow the this program in your firewall its blocked | ||
| ||
