|
On October 07 2011 23:23 Logros wrote:Show nested quote +On October 07 2011 23:19 Chewbacca. wrote: It's pretty crazy how much a difference there is in the win % of the pro races vs the win % of master/GMaster players on ladder.
On NA/EU Masster/GMaster level P beats Z 57% of the time while here Z beats P about 57% of the time, a complete reversal. Well I'm a Masters player as well and I can tell you most Masters players really suck and are nowhere near the level of real pro's. So those Blizzard statistics grouping Masters and GM together don't really say anything.[/QUOTE They are still important, because many people take the gsl code s balance problems as the reason why they cant win ZvT, PvZ or PvT in their diamond/master division.
Hell, i know silver league players who think they belong into diamond and that it is only the op'ness of terran what hinders them getting there.
|
On October 07 2011 23:28 decaf wrote: People have to take into consideration that those winrates ignore the fact that maps are imbalanced and that some races might have better players than others.
the map argument is reasonable and could be taken into account. the 'players of one race are better' argument is non-sense: you have to assume that every race has players of the same skill. otherwise balancing between different races becomes meaningless unless the mechanics of the races are virtually equivalent.
|
On October 07 2011 23:43 dooge wrote:Show nested quote +On October 07 2011 23:28 decaf wrote: People have to take into consideration that those winrates ignore the fact that maps are imbalanced and that some races might have better players than others. the map argument is reasonable and could be taken into account. the 'players of one race are better' argument is non-sense: you have to assume that every race has players of the same skill. otherwise balancing between different races becomes meaningless unless the mechanics of the races are virtually equivalent. Get real, there is no Mvp for protoss.
|
On October 07 2011 23:43 decaf wrote:Show nested quote +On October 07 2011 23:43 dooge wrote:On October 07 2011 23:28 decaf wrote: People have to take into consideration that those winrates ignore the fact that maps are imbalanced and that some races might have better players than others. the map argument is reasonable and could be taken into account. the 'players of one race are better' argument is non-sense: you have to assume that every race has players of the same skill. otherwise balancing between different races becomes meaningless unless the mechanics of the races are virtually equivalent. Get real, there is no Mvp for protoss.
can you proof that? ...no?! ok
|
The color blind version is a fantastic addition thanks for that.
|
Where is the sad zergling and the happy marine? :D jk
Thank you guys for this work, it's pretty intersting to see how the game goes month after month.
|
What? Terran is still crushing everything still? I never would have expected that....
/sarcasm.
|
On October 07 2011 23:45 dooge wrote:Show nested quote +On October 07 2011 23:43 decaf wrote:On October 07 2011 23:43 dooge wrote:On October 07 2011 23:28 decaf wrote: People have to take into consideration that those winrates ignore the fact that maps are imbalanced and that some races might have better players than others. the map argument is reasonable and could be taken into account. the 'players of one race are better' argument is non-sense: you have to assume that every race has players of the same skill. otherwise balancing between different races becomes meaningless unless the mechanics of the races are virtually equivalent. Get real, there is no Mvp for protoss. can you proof that? ...no?! ok
Quite easy to prove, just take a look at the gsl, tons of rising protoss stars and mc who is inconsitent as fuck and code b, no MV.
|
On October 07 2011 22:46 humbre wrote: no offense but tlpd is missing like half of the games
Doesn't matter, if you randomly remove half the games it should still give an accurate result, give or take a couple of percentage points. With a 50% sample that means you're getting pretty damn close to precision (as long as you don't pick and choose which games to remove).
|
On October 07 2011 23:58 kankerganker wrote:Show nested quote +On October 07 2011 23:45 dooge wrote:On October 07 2011 23:43 decaf wrote:On October 07 2011 23:43 dooge wrote:On October 07 2011 23:28 decaf wrote: People have to take into consideration that those winrates ignore the fact that maps are imbalanced and that some races might have better players than others. the map argument is reasonable and could be taken into account. the 'players of one race are better' argument is non-sense: you have to assume that every race has players of the same skill. otherwise balancing between different races becomes meaningless unless the mechanics of the races are virtually equivalent. Get real, there is no Mvp for protoss. can you proof that? ...no?! ok Quite easy to prove, just take a look at the gsl, tons of rising protoss stars and mc who is inconsitent as fuck and code b, no MV.
It's impossible to prove, because those Protoss players might be consistent and in Code S if they're Terran. The best you can do is eyeball it and give your opinion.
|
On October 07 2011 23:58 kankerganker wrote:Show nested quote +On October 07 2011 23:45 dooge wrote:On October 07 2011 23:43 decaf wrote:On October 07 2011 23:43 dooge wrote:On October 07 2011 23:28 decaf wrote: People have to take into consideration that those winrates ignore the fact that maps are imbalanced and that some races might have better players than others. the map argument is reasonable and could be taken into account. the 'players of one race are better' argument is non-sense: you have to assume that every race has players of the same skill. otherwise balancing between different races becomes meaningless unless the mechanics of the races are virtually equivalent. Get real, there is no Mvp for protoss. can you proof that? ...no?! ok Quite easy to prove, just take a look at the gsl, tons of rising protoss stars and mc who is inconsitent as fuck and code b, no MV.
His point is that if MVP played protoss he might not be as good, or that if a Protoss player (eg. MC) played Terran he could have been even better than MVP.
No way to prove it, but you do have to go with the expectation that players of all race, ON AVERAGE, are equal skill. While I doubt that you'll have protoss players who could've been on Nestea or MVP's level, on average you can't say that Korean Terrans are better than Protosses... because there's a good chance that the reason Korean Terrans are better on average (excluding cases like MVP) is that the race is stronger.
Coming from a Zerg player. I think the game is pretty balanced at our level (I'm at Top 300 masters) but at the top of the top it feels like there's a skew towards Terrans (and Zerg is about average, and Protoss feels a bit weak atm even with the buffs).
|
sick of the race whine, it only hurts the game we all care about...
the brood war equivalencies are great to look at because we're making a big fuss over nothing. BW was in a similar situation at a similar time in its life time most likely (or possibly worse)
|
keep crying idra + fans, you're doing well!
|
On October 07 2011 23:21 Ctuchik wrote:Show nested quote +On October 07 2011 22:56 Rannasha wrote: I don't like the polynomial that was fitted to the data. It's stated in the picture that it's a 3rd order polynomial (so: a x^3 + b x^2 + c x + d), which is a fairly arbitrary choice. I don't think you can just fit a polynomial function to data like this, since the data do not just come from continuous natural evolution, but also discrete jumps (changes in map-pool and mostly balance-patches can create bumps in the graphs that can't easily be approximated by a function fit).
Just provide the raw data without any attempts at fitting it next time. Well, it's a trend line, nothing more. I guess it could be de-emphasized a bit. I'm actually fairly happy with how it trends the data points.
No, it's not a trend line. A 3-month moving average like someone else mentioned earlier would be a trend line. This is fitting a very specific type of function to the data. You only make a fit if you have reasons that the data behaves in a certain way (with some stochastic noise). The fact that a 3rd degree polynomial was used to fit implies that the person doing the fit either believes that winrates follow a 3rd degree polynomial over time or that he doesn't know what a fit is for
|
On October 08 2011 00:06 ilikeLIONZ wrote: keep crying idra + fans, you're doing well!
Idra knows he's doing well, he just won IEM!
|
Looking good IMO, the game would be nothing but coinflipping if it was just 50%
|
On October 08 2011 00:02 FairForever wrote:Show nested quote +On October 07 2011 22:46 humbre wrote: no offense but tlpd is missing like half of the games Doesn't matter, if you randomly remove half the games it should still give an accurate result, give or take a couple of percentage points. With a 50% sample that means you're getting pretty damn close to precision (as long as you don't pick and choose which games to remove).
Depends on the sample size. 50% of a sample size of 4 would mean mean pretty much absolutely nothing, albeit mostly because the entire sample size doesn't mean anything either. This is a sample size that is at least large enough that it should be reasonably precise, but a larger sample size is always better.
Also based on the number of games recorded in August, this is probably closer to 1/3rd of the total games that took place in September that should be included, barring a big drop in September tournaments that I'm not aware of.
|
On October 08 2011 00:09 MrSalamandra wrote:Show nested quote +On October 08 2011 00:02 FairForever wrote:On October 07 2011 22:46 humbre wrote: no offense but tlpd is missing like half of the games Doesn't matter, if you randomly remove half the games it should still give an accurate result, give or take a couple of percentage points. With a 50% sample that means you're getting pretty damn close to precision (as long as you don't pick and choose which games to remove). Depends on the sample size. 50% of a sample size of 4 would mean mean pretty much absolutely nothing, albeit mostly because the entire sample size doesn't mean anything either. This is a sample size that is at least large enough that it should be reasonably precise, but a larger sample size is always better. Also based on the number of games recorded in August, this is probably closer to 1/3rd of the total games that took place in September that should be included, barring a big drop in September tournaments that I'm not aware of.
You're right.
Sorry, I meant assuming a reasonable sample size. In this case, 900 samples out of 1800 (or even 2700) is more than sufficient.
In your case the sample of 2 of 4 would mean nothing, I wouldn't refer to 4 as a sample size though because theoretically that would be the population, no?
A larger sample size would be better but could be cost-efficient and would have extremely marginal effects on the win %.
So it's still not a huge deal. But yes, I mean based on the numbers we have here I think it's pretty clear to say that, statistically, the ZvP and TvZ winrates clearly reject the hypothesis that the win % in these matchups are 50%. I haven't calculated for TvP but I imagine it would come ot the same conclusion.
EDIT: I haven't calculated for any of them, but from inspection it's clear of the first two statistically. I think it's pretty clear from the TvP matchup but unsure.
|
On October 08 2011 00:08 HwangjaeTerran wrote: Looking good IMO, the game would be nothing but coinflipping if it was just 50%
Dumbest comment i've read on TL EVER.
|
On October 08 2011 00:14 affliction wrote:Show nested quote +On October 08 2011 00:08 HwangjaeTerran wrote: Looking good IMO, the game would be nothing but coinflipping if it was just 50% Dumbest comment i've read on TL EVER.
Its true.
|
|
|
|