|
Russian Federation66 Posts
Being inspired of yaoyuan survey results, I decided to count real ratio of each race at NeoGame-i server. Latest 200000 matches were analyzed and here is the real picture:
Overall statistics of choosing particular race in ladder matches:
Terrans: 44% Zergs: 29% Protoss: 27%
For 1100+++ matches:
Terrans: 42% Zergs: 30% Protoss: 28%
And now the most interesting statistics. The percent of win in every matchup:
T>Z 50.2% (i.e. Terrans took a victory over Zergs in 50.2% of all TvZ matches) P>T 51.6% Z>P 54.1%
The portion of LT-matches is 89.5%
|
wow i would think itd be 95% LT
|
haha the T>Z>P>T formula stands true
|
Braavos36362 Posts
50.2 percent is hardly ">" worthy...
|
Russian Federation66 Posts
|
wow p>t i must suck hard :D i was >z
|
MURICA15980 Posts
But it still is mathmatically.
|
Zerg needs to be weakened vs P. 54.1% is ALOT when 200000 are taken into account.
54.1% would be a great % if it was spread over 500 or 1000 games. But 54.1% spread over 200000 games is very significant. Zerg winning 54.1% of the time vs toss is a considerable difference.
Also my neogamei statistics (estimated):
PvT: 75%
PvZ: 33%
PvP: 50%
I'm 1151 high =/
Yeah pvz is a bitch. One small mistake in pvz can many times cost you the game (which alot of times involves the lurker). P has to play almost perfect to win. That's why in the professional level p = z, because pro protosses can play at almost-perfect games but in the semi-pro or amateur level, Z >>> P
|
When its all factored out its about 300 more games where Z>P than P>Z its really not that big of a difference, dont ask me about my math, your all bright young children and can figure it out.
|
On July 17 2004 23:44 tfeign wrote: Yeah pvz is a bitch. One small mistake in pvz can many times cost you the game (which alot of times involves the lurker). P has to play almost perfect to win. That's why in the professional level p = z, because pro protosses can play at almost-perfect games but in the semi-pro or amateur level, Z >>> P
So true. And there are only 2 this good protoss Nal_ra,Reach.
|
Examples:
P successfully corsair-rushed Z: p kills off a few ovies, get to watch Z's tech. Game is still up for grabs.
Z successfully muta-rushed P: Game over. A successful muta rush would be gg 90% of the time.
P successfully high templar drop Z's mineral line and killed off a bunch of drones. Z is slowed down momentarily while he makes another bunch of drones from his hatches. Game still up for grabs.
Z successfully lurker dropped P's mineral line and killed off a bunch of probes (note that since toss don't have many expansions, MANY more probes are gathered in the mineral lines). 20+ probes dies. No way off toss coming back into the game now.
P successfully killed off one of Z's expasion. Z is slowed down a bit.
Z successfully killed off one of P's expansion. Game over. Toss cannot afford to lose any exp vs zerg unless he is way ahead in the game.
P catches Z offguard with fast DT. Zerg just has to pull back and wait until his ovies come along with his army and attack again.
Z catches P offguard with fast lurkers. Now protoss will be contained in his base while Z expands and MASS. In pvz, a contained toss will lose most of the time.
|
I always thought 54% spread over 500 or 10 000 000 000 games has the same significance, it's a good thing we have you here tfeign
|
Yeah that's what I did too, until I took statistic classes in college. So many ways in which statistics aren't what they seem to be
Like, I am very capable of getting 100% winning % over 10 games on ng-i, but there is no way that I can get 100% winning % over 100 games on ng-i. Not even the best koreans can do that.
Basically, the higher the number of experiments (games) are conducted, the closer and closer the % should merge to 50%.
Try flipping a coin and record the number of heads & tails in 100 throws. The head-tail percentage probably won't be 50% -- May vary 55%-45%.
Flipping the coin 1000 times may still vary, but probably no worse than 51%-49%.
And if you try flipping a coin 200000 times and record heads-tail ratio. I can guarantee to you the percentage will be 50%-50% or something extremely close to it (like 50.01%-50.99%)
If you flip a coin 200000 times and 54.1% heads came up, then the coin or experiment must have been rigged.
|
LAW OF LARGE NUMBERS MR. STATISTICS GJ.
I got a 5 on my AP stats exam.
|
the point is 54% is not that much considering all the whining we studied statistics at the academy too, don't worry
|
On July 18 2004 00:48 Breeze wrote: the point is 54% is not that much considering all the whining
Especially since this is 90% on temple, which people better than me seem to think is a Zerg map. Most imbalance problems could probably be solved by switching from temple to a carefully selected group of maps.
|
In a game you cannot have complete perfect balance. I don't expect a perfect 50%. That's why TvZ 50.2% is good.
If it was the coin flipping and 50.2% heads came up out of 200000 throws, that would be way to much.
However in a game like Starcraft, 50.2% out of 200000 is good. But 54.1% crosses the line.
Especially since this is 90% on temple, which people better than me seem to think is a Zerg map. Most imbalance problems could probably be solved by switching from temple to a carefully selected group of maps.
It's true that LT isn't the most balanced maps out there. And although it looks good on paper, it will never happen. 90% of games will always be played on lt...it's like a tradition that won't change
|
|
Is there information about the number of disconnects and draws per race and matchup?
|
Personally, I've made peace with pvz, I know it's arguably imbalanced, and I know they can't change it too drastically in fear of pvt and tvz becoming imbalanced, nor would they, blizz doesn't do anymore balance patches to this gmae. That's why I go gats on 9/10 in most of my games, and later do some crazy kind of cannon contain just to piss them off.
|
|
|
|