EPT(Korean)August TLPD Win Rates - Page 3
| Forum Index > SC2 General |
|
Grampz
United States2147 Posts
| ||
|
NotSorry
United States6722 Posts
| ||
|
EnderSword
Canada669 Posts
On September 16 2012 02:21 smileface wrote: No, it is not... With a size of 30 you already reached the sweet spot, where the results are fairly representative. Thus 28 is absolutely OK to do a monthly statistic on. And inb4 source/link/you liar/selfmade statistic on the spot bla bla: HERE Edit: I also just checked the data and 28 is just the PvZ losses... add the 42 wins in PvZ and you got a whopping 70 sample size. 30 is a good sample size when you're making a predictive model or a function when dealing with a divisible value. 30 Absolutely sucks if you're recording Yes/No or some binary result. So if you took 30 months of Stock movement, you'll get a pretty good standard deviation and model. But if you go Ask 30 people if they're voting for Romney or Obama, your poll sucks. This is especially true in cases where you're looking for small deviation from a desired rate. So if you're expecting 50/50 then you want 15Wins, 15 loses...but 16 wins 14 loses is 3.33% off, a single record distorts you by a large amount. This is even worse in rare cases, you couldn't use 30 people to determine if Carrots cause Diabetes, because the expected rate is too low. 1 in 30 people won't be expected to have Diabetes at all, let alone caused by Carrots. So in these cases you must raise the Sample size to look for Deviation cases. If you're seeing a 10% variation in Win/Loss, that only takes 3 records to deviate by that amount. The idea is to create a sample size large enough that the deviation pool is 30, not the total sample size. | ||
|
blade55555
United States17423 Posts
On September 16 2012 03:13 NotSorry wrote: People placing what too much importance on such a tiny sample size International may have more sample size but it has a lot more no name joes playing pro's. Hard to take it seriously, also all the games in korea are from top players where as international a lot of them are not from top players  . That's why korean stats are looked at more seriously then international. | ||
|
eleaf
526 Posts
| ||
|
ProxyKnoxy
United Kingdom2576 Posts
| ||
|
TechNoTrance
Canada1007 Posts
On September 16 2012 03:32 ProxyKnoxy wrote: I've stopped paying attention to these winrates... sample sizes are too small to be relevant at all Yup, all these winrates seem to do is give more fodder to balance whiners. | ||
|
NOOBALOPSE
Canada802 Posts
On September 16 2012 00:02 bo1b wrote: I'd say a 60% win rate in pvz is indicative of zerg struggling with protoss... hahaha ya. But imo ZvP is pretty straight forward and ez... for both sides. It should not reallly be that one sided for either race. | ||
|
canikizu
4860 Posts
On September 16 2012 00:02 bo1b wrote: I'd say a 60% win rate in pvz is indicative of zerg struggling with protoss... Haha true. Zerg: always manage to stay underpowered since 2010. | ||
|
spacecoke
Sweden112 Posts
| ||
|
stangstang
Canada281 Posts
DRG is still the best zerg atm but he is no taeja/mvp that's for sure. Protoss have been trading tops frequently. Squirtle, seed, mc, parting. | ||
|
Yonnua
United Kingdom2331 Posts
On September 16 2012 06:17 stangstang wrote: Zerg is severely missing their overlord. DRG is still the best zerg atm but he is no taeja/mvp that's for sure. Protoss have been trading tops frequently. Squirtle, seed, mc, parting. By "Squirtle, Seed, Mc, parting" I assume you mean: "Creator." | ||
|
nkr
Sweden5451 Posts
protoss high winrates -> omg tiny sample size | ||
|
superstartran
United States4013 Posts
On September 16 2012 03:10 Grampz wrote: Stats should be looked at over a 4-6 month period at least, I think. Taking stats from just 1 month is such a low sample size. That's funny because alot of balance changes were done on a complete whim. | ||
|
convention
United States622 Posts
On September 16 2012 07:10 nkr wrote: protoss low winrates -> omg imbalance protoss high winrates -> omg tiny sample size Yep, only protoss players do that, surely not all players of their respective race try to argue identically the same as that? I guess at least this time protoss can finally use the age old argument: Protoss is the highest skill race, look at how foreign protoss are stuggling but korean protoss are doing incredibly well. Guess protoss just benefits the most from good mechanics/micro/macro. | ||
|
Blezza
United Kingdom191 Posts
On September 16 2012 07:10 nkr wrote: (X Race) low winrates -> omg imbalance (X Race) high winrates -> omg tiny sample size Fixed. | ||
|
Lysergic
United States355 Posts
Korea PvT: 43-27 (61.4%) ZvT: 22-25 (46.8%) ZvP: 28-42 (40%) International PvT: 259-301 (46.3%) ZvT: 290-278 (51.1%) ZvP: 364-322 (53.1%) ![[image loading]](http://i.imgur.com/B9Rww.jpg) Note: I calculated the # of games for each matchup using the winrate * total games on the international graph; they might not be 100% accurate. These statistics, especially the korean ones, are pretty meaningless. Even 1 game has a significant impact on the winrates. For example, if Jaedong beat MarineKing 2-0 rather than losing 1-2 in GSL, then ZvT would be 50%. | ||
|
Assirra
Belgium4169 Posts
The reason Korean numbers matter is because the top plays there, and the top will never get a much bigger sample. Also, i don't see how it is misleading when its a fact its 60% | ||
|
DarkPlasmaBall
United States46561 Posts
On September 16 2012 08:37 lysergic wrote: You should include the number of games rather than just the percentage in the OP. It's misleading to just say PvZ 60%. Korea PvT: 43-27 (61.4%) ZvT: 22-25 (46.8%) ZvP: 28-42 (40%) International PvT: 259-301 (46.3%) ZvT: 290-278 (51.1%) ZvP: 364-322 (53.1%) All I needed to see were those sample sizes. Thanks | ||
|
TommyP
United States6231 Posts
On September 16 2012 01:30 LuckyMacro wrote: I know there are many strong EU zergs (can think of 5-6 sick good ones), but do people really believe that EU zergs are better than Korean zergs?...Just wondering. But yeah low number of games...still interesting I guess Nobody believes this except hardcore Stephano fanboys. No foreign zerg could ever win Code S (inlcuding Stephano) while DRG+ NesTea have 4 GSLs together to their name. | ||
| ||
