proving that why "race win rate" is meaningless

I am a chinese student so forget about my poor english,just play attention to meaning
----------------------------------------------------------------------------
this is what a talking about ,"race win rate",blizzard & other org may analyze mass data in ladder or pro game to get that.
we got the result the win rate alway near 50% from mass data.
so we thinked that the 3 races were nearly balance!!

but the true is,even the "same lever zvt win rate" is 70%,the result you got is always 50%,
in ladder,the MMR balance that.
the MMR is a "real point" for every player, and MMR dont care your race.
so,you will always play with a same lever player,if the zvt win rate 70%,the z player will
play with a powerful t player,of course,the win rate is 50%.

I knew this not only in math,also in my experience.
I play terren well so my tvt win rate near 80% in diamond,but my tvz is near 40%,tvp 50%.
From that I keep the balance in my group.to analyze the rep i realized the z player not strong than me(from apm,and micro).
so is the MMR hide the imba data and keep the win rate 50%.

of course,blizzard may have smarter way to get the inside data,not only MMR.
but the other org,the cant get into the deep of database,only they can do is face the massive win/lose data,
and the data is decided by MMR
-------------------------------------------------------------------------------------
that's why dont believe any "race win rate" any more!
if you find something wrong,post it,both my theory and english

Last edit: 2012-07-10 19:11:23

Salivanth Australia. July 10 2012 18:50. Posts 815

Profile Blog #

Yeah, which is why race win rate is only good for looking at short-term changes. If a patch came out that screwed up, say, PvZ, and made it really Protoss favoured, we'd know, because Protoss would start murdering Zergs. If PvZ has been broken for Protoss for 8 months, the ladder win rate won't work.

Of course, the best balance data comes from high-level tournaments where this isn't a concern anyway.

<@Wikt> so you are one of those nega-fans <@Wikt> that hates the company that makes a game and everything they stand for <@Wikt> but still plays the game <@Wikt> (like roughly 30% of blizzard's player base, maybe much more...)

AmericanUmlaut Germany. July 10 2012 18:52. Posts 2251

Profile #

This doesn't need its own thread, it's been discussed many times before. No one takes ladder winrates at face value, especially below high master/GM. Even Blizzard has said that they run their ladder data through some kind of data massaging algorithm to try and determine the "skill-adjusted" winrates at different skill levels.

The frumious Bandersnatch

Bobbias Canada. July 10 2012 18:53. Posts 1006

Profile Blog #

On July 10 2012 18:48 whipjeans wrote:
I am a chinese student so forget about my poor english,just play attention to meaning
----------------------------------------------------------------------------
this is what a talking about ,"race win rate",blizzard & other org may analyze mass data in ladder or pro game to get that.
we got the result the win rate alway near 50% from mass data.
so we thinked that the 3 races were nearly balance!!

but the true is,even the "same lever zvt win rate" is 70%,the result you got is always 50%,
in ladder,the MMR balance that.
the MMR is a "real point" for every player, and MMR dont care your race.
so,you will always play with a same lever player,if the zvt win rate 70%,the z player will
play with a powerful t player,of course,the win rate is 50%.

I knew this not only in math,also in my experience.
I play terren well so my tvt win rate near 80% in diamond,but my tvz is near 40%,tvp 50%.
From that I keep the balance in my group.to analyze the rep i realized the z player not strong than me(from apm,and micro).
so is the MMR hide the imba data and keep the win rate 50%.
-------------------------------------------------------------------------------------
that's why dont believe any "race win rate" any more!

Blizzard adjusts their win rates. They know MMR and can use that to adjust for skill, so when they do their analysis, they can see player skills and adjust the win rates accordingly. Your logic only works for unadjusted winrates.

Aterons_toss Romania. July 10 2012 18:55. Posts 1227

Profile Blog #

Blizzard win rate were always meaningless, only the TLPD one can be considered revelant

A good strategy means leaving your opponent room to make mistakes

whipjeans China. July 10 2012 18:56. Posts 14

Profile #

the tvz is so hard now,so i give up play terren to play z,i post this because blizzard also tell us that
the win rate is 50%

CrtBalorda Slovenia. July 10 2012 18:57. Posts 535

Profile #

Its logical...

4th August 2012...Never forget.....

Hardigan Switzerland. July 10 2012 19:00. Posts 1297

Profile #

we don't know what exactly Blizz is using as their proof. Maybe they just always say it's 50% to prevent massive balance whinde which would hence be justified.

and i like your saying at the beginning "play attention to meaning". It reminds me of the show "Spartacus"

superLanboy Australia. July 10 2012 19:02. Posts 535

Profile #

You don't understand how blizzard calculates win rates. They are much smarter than you give them credit for . Win rates are calculated after skill/rank difference is taken into account. How you could think they could overlook such a thing astounds me.

IdrA quotes video http://www.youtube.com/watch?v=QUGNVpzB2t0

whipjeans China. July 10 2012 19:03. Posts 14

Profile #

On Bobbias Canada. July 10 2012 18:53. Posts 722:

Blizzard adjusts their win rates. They know MMR and can use that to adjust for skill, so when they do their analysis, they can see player skills and adjust the win rates accordingly. Your logic only works for unadjusted winrates.

blizzard cant know the player skills from data,the MMR dont care the race,they have no idea to analyze each race win rate
form one player,or may be they dont admit that

Last edit: 2012-07-10 19:04:06

Ragnarork France. July 10 2012 19:06. Posts 1829

Profile #

The problem is, until we can quantify skill, we can't use the winrate system to see if a match up is balanced or not..

It's true that it can be seen when a player just overwhelm its opponent, and that skill play a role and the little imbalances don't.

But still, around small skill ranges, you can't see that...

I agree though with the face that a huge difference right after a patch may be a sign of imbalance, but it could also mean that players have not adapted themselves to the new game configuration.

ST_Bomber | EG.ThorZaIN | CMStorm_Polt | AcerMMA | Ax.Ryung

whipjeans China. July 10 2012 19:10. Posts 14

Profile #

On July 10 2012 19:02 superLanboy wrote:
You don't understand how blizzard calculates win rates. They are much smarter than you give them credit for . Win rates are calculated after skill/rank difference is taken into account. How you could think they could overlook such a thing astounds me.

blizzard may have smarter way to get the inside data,not only MMR,i will correct that
but the other org,the cant get into the deep of database,only they can do is face the massive win/lose data,
and the data is decided by MMR

iAmiAnC United Kingdom. July 10 2012 19:10. Posts 316

Profile Blog #

The TLPD win rate statistics are coming from tournaments contested by pro players, which MMR has nothing to do with (so you can still read something into these I guess). I understand what you're saying about laddering and how a player's overall win rate is pushed towards 50% (all 3 matchups combined). I certainly get the impression that my 50% win rate is comprised of some very wonky statistics. Something like 70% TvT, 50% TvP and 30% TvZ. I belive that Blizzard work with the win rates in specific matchups (i.e. Terran vs Zerg) rather than the overall value.

It would be nice if a race's overall 50% win rate was comprised of 3 50% rates (e.g. TvT, TvZ, TvP being 50% each) rather than some imbalanced matchups combining to create a "balanced" overall win rate. It would also be nice if a matchup was balanced throughout the entirety of a game. In saying that I mean it would be nice if Terran and Protoss each have an equal chance of winning the matchup in the early, mid and late game. What I think we have at the moment is the illusion of overall balance (through the combination of some skewed matchups to reach a ~50% overall rate) and the illusion of specific matchups being balanced/competitive when in reality the MU favors one race 70%, followed by the other race 70% - based on time elapsed.

I'm sure Blizzard aren't over the moon with X>Y>Z>X balance or with matchups that feel terribly skewed in one race's favor at a certain time period of the match. I'm hoping HoTS will improve the competitiveness of specific matchups over the course of a game.

http://www.twitch.tv/iamianc <- High master EU terran stream /w commentary!

Scrubwave Poland. July 10 2012 19:11. Posts 1315

Profile #

On July 10 2012 19:00 Hardigan wrote:
we don't know what exactly Blizz is using as their proof.

3D-Swifty England. July 10 2012 19:21. Posts 65

Profile #

Its still quite accurate because your MMR is joint between all 3 races.
your other two "stronger" match ups will make your MMR higher making your most difficult race to play against, more difficult.
If other people are having the same difficulty in the MU as you, this could reveal an imbalance within the matchup, within the race stats.

One of the most inaccurate parts this type of data is that, it doesn't take into consideration game length.
The other weakness of this type of test is the levels of players playing. For example a bronze terrans win rate vs zerg is 90% then plat vs plat could be 40%. This could be because of the mechanics/builds for each race are easier at different levels of the ladder.

bhfberserk Canada. July 10 2012 19:37. Posts 229

Profile #

I saw it in an interview that Blizz takes away the MMR 50% winrate to do the balance calculation. I cannot remember which interview. It would be pretty dumb to make a winrate calculation in the ladder that forces 50% winrates.

come on have a bit more faith in Blizzard lol.
I will try to find that interview.

I am a Terran player, and I actually do find the game quite balance. TvZ got really bad lately, but you just need to learn the new metagame, after you caught on the meta, it feels pretty decent again.

Azzur Australia. July 10 2012 19:59. Posts 5839

Profile Blog #

Blizzard does an adjusted win rate, where the win ratios are based on racial balance - the OP's argument is very old and has already been addressed many times.

iky43210 United States. July 10 2012 20:03. Posts 1951

Profile Blog #

winrates will always be flawed because races have different skillfloor and ceiling. There is a huge difference between races at the highest level and lowest level, and even such differences can be spotted within races.

Example: Code S vs MLG open qualifier. And for extreme unit comparison banelings are overpowered against marines in low level and not so much at higher.

Question then becomes where are you going to make that cutoff line for "balance". gold league? NA pros? European pros? MLG qualifiers? Code A? Code S? GSL semi-finalists?

wish I have better time to illustrate my points better, but hopefully its coherent enough

Last edit: 2012-07-10 20:04:51

Morfildur Germany. July 10 2012 20:25. Posts 3486

Profile Blog #

OP, you really think blizzard hasn't thought about that?

There are lots of options to get "skill adjusted winrates", it's mostly a matter of statistics.

As an example, you could calculate an MMR based on the mirror matchups as a baseline for each player and then from there compare the winrates of players of equal "mirror match MMR" (MMMMR). Now you say "but some players are better in mirror matches than others!" and this is where statistics come into play. In the big picture it doesn't matter, there is so much data that the deviation from that is negligible, 50% are better, 50% are worse, so in the end you get very precise MMMMR-adjusted winrates.

Blizzard has some very, very intelligent employees that spend 40 hours a week on doing this stuff and you spend 10 minutes thinking about it and say that they are wrong?

"Remember kids, the 3 most important things for becoming a good player: Micro, Macro and always take your Dailies!" - Rastaban

Nimix France. July 10 2012 20:29. Posts 372

Profile #

What bothers me is more how pro games look than the actual win rates. Zerg destroying terrans in not even close games, with the feeling that terran can never have enough units to actually do something, bothers me. If zerg had 70% win rate in gstl TVZ playing better and/or in close games it wouldn't bother me.

1 2 Next All

Please log in or register to reply.