EPTedit: is it because their bonuspoint is "full" and they then get a set amount of bonus points?
Match Making Rating Tool - Page 4
| Forum Index > SC2 General |
|
Greenei
Germany1754 Posts
edit: is it because their bonuspoint is "full" and they then get a set amount of bonus points? | ||
|
Doodsmack
United States7224 Posts
| ||
|
Peleus
Australia420 Posts
On May 04 2012 09:56 skeldark wrote: Not your fault, nothing to fix. The game just dont have enough data to calculate the mmr of your opponent but enough to calculate yours. Thats because you won and have no bonuspool. So i can not tell how much of your +22 was bonuspool and how much was gamewin. Because your enemy is gm your mmr is not really -60 Thats a very interesting game you have there, we need such games to calculate the sea gm offset. By all means I hope data that you gather can support this further, but I can tell you from manual tracking of my MMR over the last few days a value of (very) close to 430 would make sense for the offset. I'd say it's within 2-3 points of this. | ||
|
skeldark
Germany2223 Posts
On May 04 2012 10:07 Greenei wrote: how is it even possible to calculate my mmr even when i lose? i mean how does the programm know the bonuspool of my opponent. only sometimes there is a gap in the plot. edit: is it because their bonuspoint is "full" and they then get a set amount of bonus points? if they have bonus pool i know that he give them 1 bonus for 1 real. so all i have to do it change_points /2 On May 04 2012 10:12 Doodsmack wrote: What's the point of knowing your MMR? Is the idea that it gives you a better approximation of your skill than league and point total? I would say it's still a bad way to measure skill...skill is much more nuanced than ladder results. So that still leaves me wondering what the benefit of knowing my MMR is. perhaps skill is not ladder results. But the mmr is your ladderskill! League and point total is a function of mmr + adding skill indipendent stuffl. League and points = mmr +- random factor. And we delete the random factor that blizzard put in to hide the skill. They want you to show a a number that mostly increase so you don't notice if you are bad. Because they think we are little kids that cant handle the truth. On May 04 2012 10:37 Peleus wrote: By all means I hope data that you gather can support this further, but I can tell you from manual tracking of my MMR over the last few days a value of (very) close to 430 would make sense for the offset. I'd say it's within 2-3 points of this. Interesting. NotHat think sea offset is smaller than us and eu. The problem is you are in gm. So we only have data you to master. A master guy who play alot of other master (so we know his stable mmr) who than play against you would be perfect. But this is only one way to do it. i have the idea of a min max function, that calculate the offset range for every game and we than just take the average off it. Some data of us indicates that Excalibur dia tiers changed with the sessions. we will see when we run data analysers. | ||
|
Alejandrisha
United States6565 Posts
On May 04 2012 10:12 Doodsmack wrote: What's the point of knowing your MMR? Is the idea that it gives you a better approximation of your skill than league and point total? I would say it's still a bad way to measure skill...skill is much more nuanced than ladder results. So that still leaves me wondering what the benefit of knowing my MMR is. through enough data collection i think we could figure out the mmr cutoff of gm which would be valuable for those aspiring to make it in. seems a reasonable enough point | ||
|
Peleus
Australia420 Posts
[/QUOTE] By all means I hope data that you gather can support this further, but I can tell you from manual tracking of my MMR over the last few days a value of (very) close to 430 would make sense for the offset. I'd say it's within 2-3 points of this. [/QUOTE] Interesting. NotHat think sea offset is smaller than us and eu. The problem is you are in gm. So we only have data you to master. A master guy who play alot of other master (so we know his stable mmr) who than play against you would be perfect. But this is only one way to do it. i have the idea of a min max function, that calculate the offset range for every game and we than just take the average off it. Some data of us indicates that Excalibur dia tiers changed with the sessions. we will see when we run data analysers. [/QUOTE] Not sure if I'm interpreting correctly, so sorry! But I'm in Masters (to my annoyance  ). I've also played a lot of games ~420 this season, so I know my MMR is pretty stable. Let me know if I can help. Cheers. | ||
|
NHL Fever
Canada104 Posts
| ||
Excalibur_Z
United States12240 Posts
On May 04 2012 07:28 skeldark wrote: It does. Thats the reason for all the "Bad data not your fault". There are ways to find out what the +x is. For 50% i will code it into next version. For the rest 50% i would be forced to hack the sc2 ram. Dont know if i want to do this. It also recalculate "under master bonus" because its wrong on the bnet page. You don't need to "hack the sc2 ram" at all. You just sync the bonus pool with the Battle.net web server. That way, you already know exactly at what time it will tick over, and can take appropriate measures. Also, the division tiers never changed from season to season. | ||
|
skeldark
Germany2223 Posts
On May 04 2012 12:43 Excalibur_Z wrote: You don't need to "hack the sc2 ram" at all. You just sync the bonus pool with the Battle.net web server. That way, you already know exactly at what time it will tick over, and can take appropriate measures. Also, the division tiers never changed from season to season. i need minimum his points before the game. To get data before the game i have to check the webpage. To check the webpage i have to have his id before the replay is saved. For this i have to read the ram Because i know the id for one of the 2 players ( the guy who enter it in my settings ) i can kill at least 50% error with precheck without hack the ram. -- we just had 1 data that say the top dia tier offset is min 3 higher than the highest in your list. It can be that the algorithm is wrong. We already know he does not work always. It was only 3 points tho. If we run min max analyse on diamonds we will see it. In the end i think we can live with 10 points accuracy because we would work only on linear bases here. | ||
|
skeldark
Germany2223 Posts
On May 04 2012 11:37 Peleus wrote: Not sure if I'm interpreting correctly, so sorry! But I'm in Masters (to my annoyance ). I've also played a lot of games ~420 this season, so I know my MMR is pretty stable. Let me know if I can help. Cheers. Ok than it should be easy to get sea gm. i need a 2 games where mmr is calculated 1 against master and 1 against gm directly after each other. if we have 10 of this sets. we should be able to set the gm offset +-5 points you should give us good data to test it before we run the same system on everything. However , notthats algo is a theorie. First of all we have to improve and check it to make sure its accurate or to filter out where its not. | ||
|
Excalibur_Z
United States12240 Posts
On May 04 2012 12:57 skeldark wrote: i need minimum his points before the game. To get data before the game i have to check the webpage. To check the webpage i have to have his id before the replay is saved. For this i have to read the ram Because i know the id for one of the 2 players ( the guy who enter it in my settings ) i can kill at least 50% error with precheck without hack the ram. -- we just had 1 data that say the top dia tier offset is min 3 higher than the highest in your list. It can be that the algorithm is wrong. We already know he does not work always. It was only 3 points tho. If we run min max analyse on diamonds we will see it. In the end i think we can live with 10 points accuracy because we would work only on linear bases here. You don't need his points before the game, just before the end of the current game. There are already tools for that, such as this: http://www.teamliquid.net/forum/viewmessage.php?topic_id=328318 I'm not sure how much ram hacking is involved in tools like that or StarInfo, but they exist for your benefit and you can probably work with the creators or perform minor adjustments for them to suit your needs. | ||
|
skeldark
Germany2223 Posts
On May 04 2012 14:05 Excalibur_Z wrote: You don't need his points before the game, just before the end of the current game. There are already tools for that, such as this: http://www.teamliquid.net/forum/viewmessage.php?topic_id=328318 I'm not sure how much ram hacking is involved in tools like that or StarInfo, but they exist for your benefit and you can probably work with the creators or perform minor adjustments for them to suit your needs. This tool reads value from the ram.  I am a step ahead. Messaged the maker of the tool 5 hour ago and he gave me the offsets of the data in the ram that i need.  | ||
|
korona
1098 Posts
On May 04 2012 14:05 Excalibur_Z wrote: You don't need his points before the game, just before the end of the current game. There are already tools for that, such as this: http://www.teamliquid.net/forum/viewmessage.php?topic_id=328318 I'm not sure how much ram hacking is involved in tools like that or StarInfo, but they exist for your benefit and you can probably work with the creators or perform minor adjustments for them to suit your needs. Yes. All of them are hacks according to Blizzard's TOS as there are no other way to automatically gain opponent's name & bnet id during a match than reading them from RAM. Shame that many of these programs don't even warn their users about this, which leads that many are now 'hacking' without knowing it. After the match opponent info is available from files (replay file & s2gs file). sc2gears is clean as it only uses replays (and likely in the future s2gs files too). On May 04 2012 14:22 skeldark wrote: I am a step ahead. Messaged the maker of the tool 5 hour ago and he gave me the offsets of the data in the ram that i need. If you implement features that require memory reading please state it clearly and warn the users. The users should be able to make the decision if they willing to violate the TOS. Also a setting where the user could choose if the feature is on or off would be beneficial. | ||
|
skeldark
Germany2223 Posts
On May 04 2012 19:35 korona wrote: If you implement features that require memory reading please state it clearly and warn the users. The users should be able to make the decision if they willing to violate the TOS. Also a setting where the user could choose if the feature is on or off would be beneficial. Th plugin works anyways this option only make all data valid. So you can toggle it on/off in the settings and get a warning if you turn it on. | ||
|
mnck
Denmark1518 Posts
| ||
|
korona
1098 Posts
On May 04 2012 21:32 skeldark wrote: Th plugin works anyways this option only make all data valid. So you can toggle it on/off in the settings and get a warning if you turn it on. Yes. Meant if you add validation feature that uses memory reads there would be a toggle setting to turn it off (default off). The tool is beneficial to users even if the validation feature that requires memory reading is off. | ||
|
skeldark
Germany2223 Posts
On May 04 2012 22:11 mnck wrote: This is cool! Sc2gears just keep getting improved, and these contributions are amazing ^_^ wrong time i read this. sc2gear dont want to call my pregame trigger. Argh... Looks like the user have to enable apm alert. I dont want this so i will write my own game detector. | ||
|
Geos13
437 Posts
| ||
|
AcesAnoka
Belgium262 Posts
This is truly amazing | ||
|
Archontas
United States319 Posts
| ||
| ||
