EPT![[image loading]](http://img.photobucket.com/albums/v321/RainMan5419/MMRspike_zpsaada7d50.png)
Any idea what would cause such a huge spike?
Match Making Rating Tool - Page 90
| Forum Index > SC2 General |
|
Rainman5419
United States92 Posts
Any idea what would cause such a huge spike? | ||
|
xaoteca
Czech Republic34 Posts
On September 21 2012 14:56 Rainman5419 wrote: I'm getting a small number of "good" games like the others, but after ~50 games played I was surprised to see this. + Show Spoiler + Any idea what would cause such a huge spike? I would assume its the program predicting something and then correcting itself because of that one "good" game. | ||
|
graNite
Germany4434 Posts
Even if it would be only for all MMR Stats Users, It would be cool to see how good I am compared to others. Another cool feature would be to be able to load another Users ranking. I would ike to see my friends' ranking in my graph as well so we can compete  | ||
|
skeldark
Germany2223 Posts
-You can not compare MMR-points form last season to MMR-points from this season. Offsets changed = MMR values change -I will check whats wrong with the export function -I dont care for position in divisions. It say NOTHING about your skill. -Spikes are often mistake in the datacollection. The analyser will tell you if he thinks that the game is wrong. You can mark games that you think are wrong in the dataeditor. NEVER do it with the last good game. You dont know yet if the old value or the new is wrong. Wait unti you know for sure which one is wrong. -To see a different user you need his game_new.dat and add his account in your settings. When i publish userdata its never include the account names. -In the past promotion line where 73 mmr above the offset 0 line. It can be that this is different now but im not sure yet. Offsets looks fine, Promotion line does not. So have to investigate this. | ||
|
InfCereal
Canada1759 Posts
I'd say the master promotion line is a bit too high. | ||
|
skeldark
Germany2223 Posts
On September 24 2012 01:34 InfCereal wrote: Just got promoted. This is the graph directly after: http://i.imgur.com/2f6fB.jpg I'd say the master promotion line is a bit too high. Yes i think i will lower them all to the 0 line. I have a lot to do at the moment,so next update will take some time. | ||
|
InfCereal
Canada1759 Posts
On September 24 2012 04:46 skeldark wrote: Yes i think i will lower them all to the 0 line. I have a lot to do at the moment,so next update will take some time. If you have a chance, can you update the source code, too? | ||
|
EnderSword
Canada669 Posts
On September 24 2012 01:34 InfCereal wrote: Just got promoted. This is the graph directly after: http://i.imgur.com/2f6fB.jpg I'd say the master promotion line is a bit too high. That looks like its almost exactly accurate, doesn't it? | ||
|
barwick11
44 Posts
Do you open up the database for searching by anyone? I'm curious to see the win rates by ranges of MMR. And ideally, I'm curious to see the MMR increase of the average Zerg player since around late spring, early summer, as opposed to the MMR increase of the average Terran player in the same timeframe. Not sure how that would be searchable, possibly picking 500 random players from each race between 1000 and 1400 MMR in May, and compare that to their MMR today. Actually I'm curious on a lot more than that, but that's top of things I'm thinking of. My contention is, Zerg win rates may be "balanced" now, but I think it's because of a metagame shift that allowed both good and not-so-good Zerg players to go way higher in MMR than they otherwise would have. It almost seems like there's a crapton of Protoss left in plat through mid diamond, the same # of Terran, and a few formerly gold Zergs keep sprouting up once they learn the metagame and how to macro. | ||
|
Gemini_19
United States1238 Posts
On September 24 2012 05:02 EnderSword wrote: That looks like its almost exactly accurate, doesn't it? Yeah I'm confused as to what's wrong here. | ||
|
skeldark
Germany2223 Posts
On September 24 2012 05:09 barwick11 wrote: So, question... Do you open up the database for searching by anyone? I'm curious to see the win rates by ranges of MMR. And ideally, I'm curious to see the MMR increase of the average Zerg player since around late spring, early summer, as opposed to the MMR increase of the average Terran player in the same timeframe. Not sure how that would be searchable, possibly picking 500 random players from each race between 1000 and 1400 MMR in May, and compare that to their MMR today. Actually I'm curious on a lot more than that, but that's top of things I'm thinking of. My contention is, Zerg win rates may be "balanced" now, but I think it's because of a metagame shift that allowed both good and not-so-good Zerg players to go way higher in MMR than they otherwise would have. It almost seems like there's a crapton of Protoss left in plat through mid diamond, the same # of Terran, and a few formerly gold Zergs keep sprouting up once they learn the metagame and how to macro. check out: http://www.teamliquid.net/forum/viewmessage.php?topic_id=351786 I dont calculate this data anymore. Any kind of balance talk on this website is to ... lets call it emotional. On September 24 2012 05:32 Gemini_19 wrote: Yeah I'm confused as to what's wrong here. He moves into master and get promoted. But you have to be stable in master to get promoted. The promotion line must be lower so that his last 3-8 games are in master already before he got promoted. On September 24 2012 04:51 InfCereal wrote: If you have a chance, can you update the source code, too? Sure. I will make a update in 1-2 weeks with full collected data and clean source code of latest version. | ||
|
barwick11
44 Posts
On September 24 2012 05:51 skeldark wrote: check out: http://www.teamliquid.net/forum/viewmessage.php?topic_id=351786 I dont calculate this data anymore. Any kind of balance talk on this website is to ... lets call it emotional. Yeah I saw that thread, didn't know if you'd done it since.  So no way for a fellow computer scientist to dig through that data huh? Oh well I still think my theory may have weight... | ||
|
Lysenko
Iceland2128 Posts
On September 21 2012 17:01 skeldark wrote: -I dont care for position in divisions. It say NOTHING about your skill. You should know better than to say this. Among players who play enough to use their bonus as they earn it, division ranking is a close approximation to ranking by MMR. Those players may only be the top ten or so in many divisions, so it's a little challenging to interpret, but it's wrong to say it says "NOTHING." | ||
|
skeldark
Germany2223 Posts
On September 24 2012 14:59 Lysenko wrote: You should know better than to say this. Among players who play enough to use their bonus as they earn it, division ranking is a close approximation to ranking by MMR. Those players may only be the top ten or so in many divisions, so it's a little challenging to interpret, but it's wrong to say it says "NOTHING." So if they used up their bonus pool and you can see their adjusted points, how exactly does this help you? Adjusted point != skill. You come one step in the right direction but their are many to go. You assume heigher adjusted = heigher mmr. That is not the case. You asume: "division ranking is a close approximation to ranking by MMR". Thats not always the case! That are lies from blizzard to let you think points are more meaningful than they are. It can be an indicator but its way more complicated than just bonus-pool. Bonuspoint is only one of many hiding functions in the battlenet. Calculating it out is the frist step on finding mmr. But their are still many left. I would be surprised if the top ten in a division are the top 10 in mmr in this order. Its possible but unlikely. It say SOMETHING. It gives data to calculate mmr. But what i wanted to say is, that on watching the numbers or the position in your league without calculation, you can not say ANYTHING ACCURATE about skill. I said it once in this thread: You try to count the clouds, you drive by with your car to measure your speed. This give you "some" kind of indication about speed. But as someone who build speedometers, im not really interested in this kind of information. People ask me all the time: "Why do you say i drive 100km/h when i counted 5 clouds today? " Thats why i said so dramatic: "Clouds you drive by, say NOTHING about your speed." Its right on the one site because you have to find out how fast the clouds are and how big they are first and its wrong on the other site because when you collect this information you can really calculate your speed. The whole system is designed to hide your skill and show you a number that let everyone think he is better than he is. And whatever i think about this task, i must admit, they do this part good. On September 24 2012 07:47 barwick11 wrote: Yeah I saw that thread, didn't know if you'd done it since. So no way for a fellow computer scientist to dig through that data huh? Oh well I still think my theory may have weight...I can give you the anomynised gamedata i have. Much to do at moment, perhaps in 1 week. | ||
|
sour_eraser
Canada932 Posts
Ya. Divison rank doesnt mean much. Im 2nd place with 300 points with 30 wins in my Diamond league while 1st place is a guy with 50 wins but with 310 points. Now, do you honestly think rank 1 player is still better player than me? Points are mainly there to make you feel good about yourself. | ||
|
Lysenko
Iceland2128 Posts
On September 24 2012 15:07 skeldark wrote: You assume heigh adjusted = heigh mmr. That is not the case. You think points follow mmr. Thats not allways the case! The feedback between MMR and adjusted point score is quite strong. Adjusted points converge on MMR if MMR is stable. Yes, the two can be out of whack if MMR is moving rapidly, but most players have pretty stable MMRs because active players usually improve pretty slowly. I get the feeling that you look at the differences between adjusted points and MMR and see them as more meaningful than they are. MMR is a pretty rough way to guess the outcome of a game to begin with -- a wide range of players can take a game off someone at a given MMR. And yeah, the point system is designed to give people a feeling of progressing by playing more even if their MMR remains stable. You call this deception, I call it a reward system. | ||
|
Lysenko
Iceland2128 Posts
On September 24 2012 15:18 jidolboy wrote: ^ Ya. Divison rank doesnt mean much. Im 2nd place with 300 points with 30 wins in my Diamond league while 1st place is a guy with 50 wins but with 310 points. Now, do you honestly think rank 1 player is still better player than me? Points are mainly there to make you feel good about yourself. Both of you have used up your bonus pool, so playing those extra 20 games don't necessarily earn him extra points. If the system finds him good matches and his MMR is staying put, he'll earn as many points as he loses playing those extra 20 games. A 10-point difference is in the noise, so chances are if the two of you played it would be pretty even. Also, if he has a relatively short crazy winning streak and his points increase more than his MMR, he'll get fewer and fewer points per win and lose more for losses until his points wind up in equilibrium with his MMR again. A lot of people seem to have this idea that playing more games earns you more points ad infinitum. That's not true. Playing more games only guarantees you additional points to the extent that you've accrued bonus pool, and once that's run out, playing more will not change your point score unless your MMR is going up. | ||
|
skeldark
Germany2223 Posts
On September 24 2012 15:22 Lysenko wrote: The feedback between MMR and adjusted point score is quite strong. Adjusted points converge on MMR if MMR is stable. Yes, the two can be out of whack if MMR is moving rapidly, but most players have pretty stable MMRs because active players usually improve pretty slowly. I get the feeling that you look at the differences between adjusted points and MMR and see them as more meaningful than they are. MMR is a pretty rough way to guess the outcome of a game to begin with -- a wide range of players can take a game off someone at a given MMR. And yeah, the point system is designed to give people a feeling of progressing by playing more even if their MMR remains stable. You call this deception, I call it a reward system. I edited my post. Read it again. Bonuspool is not the ONLY hiding function! Its just the one that most people know about. "The feedback between MMR and adjusted point score is quite strong" Wrong! The feedback between MMR and adjusted point can be not existed if some other rules trigger! Stability + Bonuspool is not all either! In fact blizzards stability function is so plain simple it does not care after 30 games. | ||
|
Lysenko
Iceland2128 Posts
On September 24 2012 15:26 skeldark wrote: I edited my post. Read it again. I say it again: Bonuspool is not the ONLY hiding function! As you know, I find your posts interesting, but the degree to which you assume that Blizzard's making their choices explicitly to make your life more difficult is very tiresome. Wrong! The feedback between MMR and adjusted point can be not existed if some other rules trigger! Describe these rules to us in detail, let's get Excalibur_Z to discuss them with the developers, and get his post updated, if they're a real thing. That is, unless you're talking about weird edge cases at the extreme high end or low end of the ladder, or how things like promotion or demotion events are handled, in which case it wouldn't be an interesting discussion in any case, because of course those kinds of things have to be handled specially. | ||
|
skeldark
Germany2223 Posts
On September 24 2012 15:29 Lysenko wrote: As you know, I find your posts interesting, but the degree to which you assume that Blizzard's making their choices explicitly to make your life more difficult is very tiresome. Yes as i know. And we know the system by now. I know this could sound arrogant but i know it well better than you. (if you are not the one blizzard employ who inverted it) I know their are more hiding functions and i can prove that. Blizzard dont try to make my life hard. They try to hide some facts from every user. They dont do it, so i can not calculate it. They do it so you dont see your skill in your points. That the whole task! On September 24 2012 15:29 Lysenko wrote: Describe these rules to us in detail, let's get Excalibur_Z to discuss them with the developers, and get his post updated, if they're a real thing. That is, unless you're talking about weird edge cases at the extreme high end or low end of the ladder, or how things like promotion or demotion events are handled, in which case it wouldn't be an interesting discussion in any case, because of course those kinds of things have to be handled specially. We describe them in detail! check out op and the links in it. I talk with Excalibur about this and other rules a lot. He updated his post. He ask blizzard about it. You have to read between some lines in his post to understand it because his post is focus on explaining the basics to the average tl user. We calculated and analysed the ladder system for month. You come here, read Excalibur guid that is written for people who dont know anything about the ladder and only explain the basics and than you try to explain me how it works. Thats very arrogant on your part. | ||
| ||
