|
United States12235 Posts
On November 20 2012 17:21 skeldark wrote:
@Excalibur_Z & @Mendelfist
Games have a modus tag. So if they do it right they will add an "unranked 1v1" modus.
Than its no problem at all. If they do it the lazy & dirty way, like always, they let them look like ladder games.
In this case they can look like +- 0 games.
The problem is that its hard to tell if their was an mistake in the calculation or if its unranked.
Their is no clear "Ladder" Ram entry. Its hard to find out the gametyp ingame.
I was thinking you could read not from the gametype but from the league icon that's displayed at the end of a match. If there is an icon shown, it's a ranked match. Otherwise, it's an unranked match (or a placement match but those are also unusable). I don't know if the RAM Scanner is capable of delving this deep but this seems like the only way to do it.
On November 21 2012 00:33 BurningRanger wrote:
For the offrace reason I'd expect them to take the ranked MMR for your first few unranked games. Then split off your unranked MMR. I think that's how they want to do it. Even though it just creates a more hidden MMR.
But the most important point, as skeldark pointed out already, is that ranked should never ever play against unranked or it will mess up ranked MMR so much.
- players dropping games to get a specific matchup will push their ranked opponents up the ladder
- players losing from shitty/stupid/funny/i don't care strategies will push their ranked opponents up the ladder
- high MMR players playing unranked but still honest that win will push their ranked opponents down the ladder
- high MMR players dropping unranked MMR, smurfing in low leagues, crushing lowleaguers will push their ranked opponents down the ladder
There are probably more scenarios like this.
Where you end up in the ladder (with your ranked MMR) doesn't depend on skill anymore, but on being lucky or unlucky with your unranked opponents.
Yeah, this is why things are going to get muddy. Alan Dabiri confirmed in Antoine's TL interview that your unranked MMR will bootstrap from your ranked MMR then move independently from there (it actually works in both directions but it's still a true statement). I'm sure it's probably a safe assumption that players in the Unranked pool aren't going to be trying as hard to win generally speaking, otherwise they would play Ranked. You already mentioned the possible and probably likely scenarios that could destabilize the ladder, and I'm sure Blizzard has anticipated them, but I'd be interested to hear the reasoning behind their decision to keep it this way in spite of these potential issues. Maybe they think it won't have a very large effect, or maybe it will be something they'll continue to revise and for now they're just taking a wait-and-see approach.
|
I can imagine that the logic is that it'll all come out in the wash anyways. Each person's unranked MMR is, in effect, one extra person that will perform at most as well as that person in "ranked mode". How they perform in unranked mode, regardless of their motivation, will be where ever their MMR stabilizes which is the intention of the system in the first place.
|
When i first heard that there will be ranked vs "unranked" matched together i was shocked.
That means if i play a ranked game against an opponent who plays unranked it's highly depending on luck if i win. Maybe i play against an opponent who plays serious with an mmr tanked account or the opposite. Maybe he just fools around on an account he usually plays serious on. In both cases will the outcome of the match not be decided by myself.
I also doubt it will be fun for bronze players to play against fun strategys like nuke rushes or only sentries if they want to play "seriously".
I am not sure if it is necessarily a problem for the ranked ladder ranking so, if those inaccuarte results don't happen too often. For example: we have one "bad" game result and then nine accurate ones. At least from my understanding should the ladder ranking evened out after those games.
@Mendelfist
Alan Dabiri said that there won't be seperate MMR's for each race in HOTS.
I am disappointed about that and i think it will screw the unranked mmr up (which will also affect the ranked mmr because they are matched against each other). This will result in a lot of bad matches.
I think the optimal solution for the players would look like this:
2 seperate player pools for ranked and unranked.
Seperate mmr for each player and each race on ranked.
Real unranked play where you can select: your skill, opponent's race, opponent's skill, map, etc.
@skeldark
I am not sure if i understood correctly what you meant. You said it would be problematic for the skill system if there are more than one MMR per player. I don't see how that would influence the match making. The only thing that would change imho is that you had a lot more "accounts". You couldn't say anymore i am at 50% of all players. You would need to say my Zerg mmr is at X% of all accounts. This is imho an unproblematic adjustment because nobody can give accurate data about that anyway with the system blizzard uses.
Btw sad that you won't continue your project with HOTS. I can certainly understand it so. =/
i hope you reconsider when HOTS is out. =)
|
I had 1580 MMR or something, had won 10 games in a row, and won an additional game, then i dropped down to 1480 MMR.
Seems kind of weird.
Any information i can supply you with to find the root of the problem, to improve the plugin further?
|
On November 21 2012 08:11 OrbitalPlane wrote:
When i first heard that there will be ranked vs "unranked" matched together i was shocked.
That means if i play a ranked game against an opponent who plays unranked it's highly depending on luck if i win. Maybe i play against an opponent who plays serious with an mmr tanked account or the opposite. Maybe he just fools around on an account he usually plays serious on. In both cases will the outcome of the match not be decided by myself.
I also doubt it will be fun for bronze players to play against fun strategys like nuke rushes or only sentries if they want to play "seriously".
I am not sure if it is necessarily a problem for the ranked ladder ranking so, if those inaccuarte results don't happen too often. For example: we have one "bad" game result and then nine accurate ones. At least from my understanding should the ladder ranking evened out after those games.
The problem for low leagues is, you're already in a swamp with people playing "not seriously", either because they don't care, don't know better, want to mess you up or want to cheese their way up. It's already a pain, if you want to play "seriously" and get better. Adding unranked vs. ranked will just add more of that crap to lower leagues, imo.
On a sidenote... has anyone ever thought about making an own ranking system (like the one for Broodwar with A to D ranking or so, can't remember the name)? One that isn't so messed up. Blizzards Matchmaking System could still be used to find opponents. Just the outcome could affect such a custom ranking system.
|
On November 21 2012 09:13 BEViking wrote:
I had 1580 MMR or something, had won 10 games in a row, and won an additional game, then i dropped down to 1480 MMR.
Seems kind of weird.
Any information i can supply you with to find the root of the problem, to improve the plugin further?
Most likely you only have 2 good games and one of them is a mistake.
Enable Memory Reader and ladder on. With a 3. good game you see which one is the mistake. The autocorrection function perhaps even find and correct it.
---
Will read rest later. This discussion is fast moving Oo.
|
On November 21 2012 19:56 skeldark wrote:Show nested quote +On November 21 2012 09:13 BEViking wrote:
I had 1580 MMR or something, had won 10 games in a row, and won an additional game, then i dropped down to 1480 MMR.
Seems kind of weird.
Any information i can supply you with to find the root of the problem, to improve the plugin further? Most likely you only have 2 good games and one of them is a mistake. Enable Memory Reader and ladder on. With a 3. good game you see which one is the mistake. The autocorrection function perhaps even find and correct it. --- Will read rest later. This discussion is fast moving Oo. What measures the MMR? Is it how much i win, or how good i play in each game?
It says i'm @ 1410 if i pick the latest 5 games, but i've been playing against top 50 masters and won those games.
Also, i can't seem to find those options that you spoke about.
|
On November 21 2012 22:41 BEViking wrote:What measures the MMR? Is it how much i win, or how good i play in each game?
It says i'm @ 1410 if i pick the latest 5 games, but i've been playing against top 50 masters and won those games.
Also, i can't seem to find those options that you spoke about.
MMR is only calculated from wins and losses, no matter if it was close or 1 rolled the other, no matter if you cheesed or have been cheesed. Just win or loss counts in the end.
The one option you can activate (if not active already) is the "Memory Reader (more data)" in Config -> Settings at the bottom.
Gratz for beating top 50 masters. Go on like this and your MMR will rise quickly and you get promoted to Master. Good luck 
|
On November 21 2012 22:41 BEViking wrote:Show nested quote +On November 21 2012 19:56 skeldark wrote:On November 21 2012 09:13 BEViking wrote:
I had 1580 MMR or something, had won 10 games in a row, and won an additional game, then i dropped down to 1480 MMR.
Seems kind of weird.
Any information i can supply you with to find the root of the problem, to improve the plugin further? Most likely you only have 2 good games and one of them is a mistake. Enable Memory Reader and ladder on. With a 3. good game you see which one is the mistake. The autocorrection function perhaps even find and correct it. --- Will read rest later. This discussion is fast moving Oo. What measures the MMR? Is it how much i win, or how good i play in each game? It says i'm @ 1410 if i pick the latest 5 games, but i've been playing against top 50 masters and won those games. Also, i can't seem to find those options that you spoke about.
MMR is the result of an skill function like the elo number in chess.
It depends on the MMR of your opponent and if you win or lose. E.g. Loosing against someone weaker than you will drop your MMR big because you should be able to beat him. If you Losse against a higher player you dont loose much MMR because the system expect you to lose this game.
When you say TOP 50 master i think you mean his place. This is a fake number of blizzard and says nothing about how good he is in sc2. He can be on of the best players or the worst player of the world. Even leagues like "master" can be total wrong.
His MMR matters, nothing else and you can not see it in battelnet. Thats why i made this tool.
You can enable memory reader in the options. Frist tab down right.
|
On November 21 2012 22:54 skeldark wrote:Show nested quote +On November 21 2012 22:41 BEViking wrote:On November 21 2012 19:56 skeldark wrote:On November 21 2012 09:13 BEViking wrote:
I had 1580 MMR or something, had won 10 games in a row, and won an additional game, then i dropped down to 1480 MMR.
Seems kind of weird.
Any information i can supply you with to find the root of the problem, to improve the plugin further? Most likely you only have 2 good games and one of them is a mistake. Enable Memory Reader and ladder on. With a 3. good game you see which one is the mistake. The autocorrection function perhaps even find and correct it. --- Will read rest later. This discussion is fast moving Oo. What measures the MMR? Is it how much i win, or how good i play in each game? It says i'm @ 1410 if i pick the latest 5 games, but i've been playing against top 50 masters and won those games. Also, i can't seem to find those options that you spoke about. MMR is the result of an skill function like the elo number in chess. It depends on the MMR of your opponent and if you win or lose. E.g. Loosing against someone weaker than you will drop your MMR big because you should be able to beat him. If you Losse against a higher player you dont loose much MMR because the system expect you to lose this game. When you say TOP 50 master i think you mean his place. This is a fake number of blizzard and says nothing about how good he is in sc2. He can be on of the best players or the worst player of the world. Even leagues like "master" can be total wrong. His MMR matters, nothing else and you can not see it in battelnet. Thats why i made this tool. You can enable memory reader in the options. Frist tab down right. Oh, i see. Thank you for the help.
I went in to the settings of Sc2gears, but with your help i understood that i should open the actual plugin.
Thank you!
|
Thanks Skeletor for the tool!
PS. He-Man sucks!
|
On December 04 2012 22:04 Popiel wrote:
Thanks Skeletor for the tool!
PS. He-Man sucks!
Thank you!
PS: Yeah he did.
Did you noticed that you can not find him on TV or in toystores anymore?
Its because i killed him some years ago!
Without this stupid idiot trying to destroy my empire,
i finally could work on important tasks, like back-engineer mmr values.
|
I understand you were thinking about dropping support for the MMR tool, skeldark - is that still the case? If you need someone to take over the project I'd be happy to do it (after some briefing and explanation )
|
On December 07 2012 04:16 Hairy wrote:I understand you were thinking about dropping support for the MMR tool, skeldark - is that still the case? If you need someone to take over the project I'd be happy to do it (after some briefing and explanation )
Have couple of people who want to keep it alive.
I think i will publish all i have an pick a group of people.
You guys can than figure out yourself, who does what.
Should be the best solution because some have knowledge about the mmr calculation some in programming.
I know that most will drop out fast anyway. (always like this in group projects)
The code is written in modules so programming with more than one guy should be no problem.
I will create a list of all people who are interested and set a meeting before hots release.
I have to pull out the webserver too because its not my own and i cant give the accounts away.
EDIT:
Just checked the gamecount after long time.
Looks like i have a buffer overflow in the value Oo
I will count the gamefiles later but should reached 1 Million by now.
|
Well I am a programmer by profession, but only really have C++ and C# experience at the moment - java is basically an easier c#, right? 
|
On December 07 2012 17:08 Hairy wrote:Well I am a programmer by profession, but only really have C++ and C# experience at the moment - java is basically an easier c#, right?
oO
I know a lot of people who would go full rage mode for this sentence.
Gladly im not one of them 
Java is 100% typesecure and also more strict in other parts.
You can say, it force you to work cleaner.
But, its also harder/impossible to interact with some system stuff in java.
So :
- easier functional part, (less possibility - e.g. all old c+ orders that work in c#)
- harder object part if you like it quick and dirty( because you are forced to do it cleaner)
-less power.
- 98% same syntax
The only thing you have to relearn -if you dont want to reinvent the wheel- is like allays: the Bib.
|
Hi, I've just started using this tool - its really great!
I have one question. I managed to get a good game but I'm so far 1/12
I notice the % is based on 4 criteria (its always above 50% but can't figure the logic behind the last 2 stages)
22:45:42 Player: Reading_OK: true Adjusted_OK: true Cap_OK: false Dev._OK: true
22:45:42 Opponent: Reading_OK: true Adjusted_OK: true Cap_OK: false Dev._OK: true
Could you explain what each of the 4 things means (I've looked through the thread but couldn't find the answer) and is there any way other than the ram scanner to improve the %?
Thanks
|
On December 08 2012 08:16 Aimconquest wrote:
Hi, I've just started using this tool - its really great!
I have one question. I managed to get a good game but I'm so far 1/12
I notice the % is based on 4 criteria (its always above 50% but can't figure the logic behind the last 2 stages)
22:45:42 Player: Reading_OK: true Adjusted_OK: true Cap_OK: false Dev._OK: true
22:45:42 Opponent: Reading_OK: true Adjusted_OK: true Cap_OK: false Dev._OK: true
Could you explain what each of the 4 things means (I've looked through the thread but couldn't find the answer) and is there any way other than the ram scanner to improve the %?
Thanks
Thanks.
Reading_OK: Report if the data gathering worked
Adjusted_OK: Calculation process passed without errrors / also test if the gamedata is valid
Dev._OK: Ceck if the errormargin is in an acceptable area
Cap_OK: The one thats the reason for your low goodgame count:
This detects caped games. Caped games are games, where the system give you more points than you deserve.
This is another hiding method. If the system give you fake points, i can not know how much real points you would get.
And this information is critical to calculate MMR.
There is nothing on your side to do against it. It depends on your MMR and Opponent MMR and your and opponent league.
You have many caped games if your MMR is in the Cap area .
But As long as you have 1 out of 10 games "good" the prediction should stay reasonable accurate.
|
Thanks for the answer much appreciated,
So if my understanding is correct, you need an opponent who has a very close MMR to yours in order to get a 100% game as well as ensuring all the other criteria are met. But ofc the matchmaker picks opponents who have similar but not quite similar enough as well hence why non-good games occur
Is that correct or am I missing something?
|
On December 08 2012 08:45 Aimconquest wrote:
Thanks for the answer much appreciated,
So if my understanding is correct, you need an opponent who has a very close MMR to yours in order to get a 100% game as well as ensuring all the other criteria are met. But ofc the matchmaker picks opponents who have similar but not quite similar enough as well hence why non-good games occur
Is that correct or am I missing something?
No. The cap test is a complicated function. The function hide lower point count
Simple:
If your Opponent have less MMR than your league+ around 80 , he will get free points to hide this fact, from him.
Because i need HIS change to calculate YOUR MMR, i calculate a wrong result and have to throw it away.
If you are in this MMR area, the chance that your opponent is in it too, is obvious high, so the chance for a caped game is high.
|
|
|
|
|
|
EPT
