EPT
Blizzard Blog: Balance Snapshot - Page 25
| Forum Index > SC2 General |
|
KristianJS
2107 Posts
| ||
|
Slunk
Germany768 Posts
On September 23 2011 18:29 sleepingdog wrote: if P was immensely buffed, some better Ts would fall out of GM and some worse Ps would get in. Then - since those Ps would still suck compared to gosu-Ts - would probably end up with a win/loss-ration below 50%. Balanced? I agree that Blizz "knows" it, but they use the data to "convince" the community, everything is right - when it's not. No offense, but this is the srupidest attempt at logical thinking I've ever read in my life. | ||
Jibba
United States22883 Posts
On September 23 2011 19:23 mTw|NarutO wrote: It actually doesn't matter how strong one race is in some stages of the game. If Zerg can't reach lategame because of imbalance or disadvantages there's no point in not adjusting the game. Ofcourse you have to take into account the Korean players. Ofcourse NesTea with his individual skill makes for good games, but most Zergs for example still suffer and die easily. You have to take a look into the proscene, but do not take a few examples out of there that are probably outstanding. If both races have even chances of reaching all stages of the game you can begin adjusting the strength of different units and aspects which will lead to hopefully - just the lategame. As in nerf broodlords in damage output etc. I understand what you mean I think, but I still don't think I agree with what you are saying. And I do also hate TvZ Lategame. :o! I can imagine on absolute highest level TvZ 25+ will lead in insanely close games, as Zerg is sick powerful, but Terran with good control and upgrades as well as bases can make good use of the ghost. That's part of the analysis though. If Zergs win 60% lategame, but only 20% of Zergs get past the 15th minute, then that's something they should look into. I'd really like to know just how deep Blizzard's numbers go. If it's just win %s, then that's really bad. | ||
|
RvB
Netherlands6261 Posts
On September 23 2011 21:08 Jibba wrote: That's part of the analysis though. If Zergs win 60% lategame, but only 20% of Zergs get past the 15th minute, then that's something they should look into. I'd really like to know just how deep Blizzard's numbers go. If it's just win %s, then that's really bad. I doubt they only have win %.. Blizzard can't be that stupid, besides they made this huge ass formula to calculate skill in it etc. If you do that much effort into getting win rates I doubt that's all the statistics they got. | ||
|
Holykitty
Netherlands246 Posts
seems that if everyone improves a bit more zvt atleast would be a fine match up. | ||
|
sleepingdog
Austria6145 Posts
On September 23 2011 21:05 Slunk wrote: No offense, but this is the srupidest attempt at logical thinking I've ever read in my life. You've convinced me, I take back what I said. Clearly we need to make terran stronger, since PvT has a win-rate above 50% over all leagues in NA/EU. | ||
|
Big J
Austria16289 Posts
There are approximatly 100times more games being played in master's than in GM, so you only have a everyone and a corrupted Masters table... Really doesn't say anything about balance among good players... PLEASE JUST GIVE US THE GRANDMASTER STATS and PLEASE TELL US HOW MANY GAMES IN WHICH TIME PERIOD AND ON WHICH PATCHES HAVE BEEN PLAYED! | ||
|
HaXXspetten
Sweden15718 Posts
PvT still not perfect though | ||
|
paralleluniverse
4065 Posts
On September 23 2011 01:52 Drunken.Jedi wrote: Well, this doesnt really say much. Blizzard bascially gave us a bunch of numbers, but didn't explain what those numbers actually mean, except in very vague terms. I would be very interested in how Blizzard gets those "adjusted" win percentages. As far as I can see, the only way to determine player skill accurately and objectively is to look at match results, but match results are already influenced by balance and if those numbers are based on that, they aren't useful for assessing balance. If it says in the table that PvT is 53%, then that means in a match between a Protoss and Zerg player of exactly equal skill, the probability of the Protoss player winning is 0.53. As for how it's calculated that's already been explained in a few threads, for example this one: http://www.teamliquid.net/forum/viewmessage.php?topic_id=250789 | ||
|
FeyFey
Germany10114 Posts
EU = Macro KR = Early aggression with no afterwards plan (sounds nicer then pure allin) NA = kinda everything played randomly. If you want high level take tournament winrates after around 500 played games taking out the imbalanced maps :p. But since i know zergs love to get to much drones in us/eu. And that terrans in KR only know allins. Those winrates in master/gm are not really an suprise. Its quiet interesting though that PvT is close to 50% in NA/EU. thought terrans would get poked to death there thanks to chargelots vs marauder. Anyway they can't buff zergs early game anymore, just like they can't nerf terrans early game to strong, or toss early game. | ||
|
MilesTeg
France1271 Posts
On September 23 2011 21:17 sleepingdog wrote: You've convinced me, I take back what I said. Clearly we need to make terran stronger, since PvT has a win-rate above 50% over all leagues in NA/EU. Well... I know you're not a bad poster but honestly your post made no sense at all. On September 23 2011 21:38 paralleluniverse wrote: If it says in the table that PvT is 53%, then that means in a match between a Protoss and Zerg player of exactly equal skill, the probability of the Protoss player winning is 0.53. As for how it's calculated that's already been explained in a few threads, for example this one: http://www.teamliquid.net/forum/viewmessage.php?topic_id=250789 It hasn't really been explained actually. No one knows what they are using to measure skill, if they take another parameter... I don't see how they can calculate "clean" win percentages that get rid of the MMR effect with the data they have. | ||
|
paralleluniverse
4065 Posts
On September 23 2011 21:50 MilesTeg wrote: It hasn't really been explained actually. No one knows what they are using to measure skill, if they take another parameter... I don't see how they can calculate "clean" win percentages that get rid of the MMR effect with the data they have. In very simple terms: If you recall conditional probability from school, they calculate the probability of winning conditional on MMR. | ||
|
paralleluniverse
4065 Posts
On September 23 2011 21:50 MilesTeg wrote: It hasn't really been explained actually. No one knows what they are using to measure skill, if they take another parameter... I don't see how they can calculate "clean" win percentages that get rid of the MMR effect with the data they have. I would also add that what they use is from the theory of Bayesian Inference, which is usually taught as a honors year math/stats course at most universities, so you won't understand the fine details of this method unless you have a university-level background in statistics. However, it is still very possible to understand the overall "big picture" view of what they're doing. Suppose there are 5 numbered balls in a bag. 1 - red 2 - red 3 - blue 4 - blue 5 - blue (A) What is the probability of drawing a ball numbered 4? Clearly it's 1/5. (B) For the balls that are blue balls, what is the probability of drawing a ball numbered 4? It's the number of balls that are both blue and numbered 4, out of the number of balls that are blue, that is 1/3. Now replace "less than 4" with "winning" and "blue ball" with "MMR equal to, say, 3000", then (A) corresponds to simply looking at unadjusted win ratios and (B) is in principle the method that Blizzard uses. The idea is to get the probability of drawing a ball numbered 4 when fixing the color of the ball, or in Blizzard terms, getting the probability of winning when fixing MMR, and that's done by taking the probability of both having a certain MMR and winning, divided by the probability of having that certain MMR, i.e. what we did in (B). Please tell me if my explanation isn't clear, I do actually want people to understand this idea. | ||
|
Steel
Japan2283 Posts
On September 23 2011 01:38 chadissilent wrote: Nerf Fungal Growth. Haha nice post especially when the matchups seem balanced... The only problem seems to be PvT but that's mainly because of 1/1/1 being strong even at the tip top level. If this goes on for longer there might be a balance issue (with T IMO) but maybe toss haven't figured out the optimal response yet... Or maybe it's a map issue. I mean, on GSL maps they get SO much more time to prepare whereas on Xel'naga Caverns, the lava map...most maps!! The Terran has to screw up hard to lose. Blizz should make the ladder maps more GSL-like and eliminate spawn position in my opinion, I think it would make 1/1/1 much weaker..and balance discussions could go from there. | ||
|
ElusoryX
Singapore2047 Posts
On September 23 2011 01:52 Andreas wrote: Interesting how NA and EU PvZ and TvZ winrates are so much higher than in Korea. Are Korean Zergs really that good? yes they are. they are highly aggressive and do pretty much everything lol | ||
|
flodeskum
Iceland1267 Posts
On September 23 2011 21:27 HaXXspetten wrote: Nice finally seeing Blizz release something like this, and also nice to see that it does reflect everything that we were expecting PvT still not perfect though Even if the win percentage was 50/50 I feel like PvT needs a bit of work. The games just look bad and are uncomfortable to play. Both races are allining each other and if the games go into the lategame they are often decided by a single one sided fight on either side. It's just bad. | ||
|
polysciguy
United States488 Posts
On September 23 2011 19:41 Paladia wrote: In all studies I've seen, there is a very strong correlation between APM and skill. It isn't perfect but in many ways it is a better indicator than win ratio in a match made-scenario. Using Dreamhack as an example. The average finishing position of the people with the top 10 highest APM was 16.5. The average position of the bottom 10 APM was 28. As such, there is a very strong correlation between APM and how far they got in the tournament. So, to prove your point, please come up with a real life statistical analysis that shows that higher APM on average does not indicate a generally higher skill. A thought experiment you can do on yourself is. If you got to choose who to represent you in a tournament out of two groups, would you rather randomly get one out of a group with very low APM (50) or out of a group with high APM (300)? The answer should be quite obvious, and it is obvious due to APM on average correlates to higher skill. correlation=/= causation. there is a strong correlation between ice cream sales and crime rates, does that meant that ice cream is the cause....no, in fact there is a 3rd factor that controls both, being weather. correlation doesn't mean anything more than that there is a relationship between the two. | ||
|
Ninja [X]
Korea (South)40 Posts
| ||
|
sluggaslamoo
Australia4494 Posts
On September 23 2011 21:08 Jibba wrote: That's part of the analysis though. If Zergs win 60% lategame, but only 20% of Zergs get past the 15th minute, then that's something they should look into. I'd really like to know just how deep Blizzard's numbers go. If it's just win %s, then that's really bad. Pretty much exactly what I was thinking, these stats are irrelevant to me. | ||
|
Tommie
China658 Posts
On September 23 2011 21:05 Slunk wrote: No offense, but this is the srupidest attempt at logical thinking I've ever read in my life. No it is very true. If one of the races is imbalanced it will show up in the number of players in the top leagues. We are seeing this both in code S and in the grandmaster leagues. | ||
| ||
