| Forum Index > SC2 General |
|
kafkaesque
Germany2006 Posts
On May 02 2013 03:25 Algar wrote: "I have never taken a university level statistics course." Didn't your university level statistics course teach you how rare people who take university level statistics courses are? | ||
|
achan1058
1091 Posts
On May 02 2013 05:28 kafkaesque wrote: Didn't your university level statistics course teach you how rare people who take university level statistics courses are? No, since all engineers/CS students in my school are required to take the entry level course. I don't know about his school though. No, there's nothing wrong with the sample size. On the other hand, this is in no way indicative of balance neither. It merely states that "among these tournaments, X race has Y win percentage against Z", nothing more, nothing less. | ||
|
MrMatt
Canada225 Posts
On May 02 2013 05:28 -VapidSlug- wrote: Probably because: "all pro data" = too small sample size "all ladder data" = way too noob "all tourny data" = korea vs foreign skew "all top ladder" = too noob/cross region issues "all top korea" = terran-culture/too small sample/(variable complaint here) In short, I don't think the SC2 community will ever agree on any statistical analysis. This is how I feel. Or like now. people are like well zerg just won the last tournament, while others argue that they were the only zerg and much better than their opponents. Same with when taeja was winning and some said terran is fine, terran just won. then people argued that one terran won and was better than their opponents. Peeple will use or ignore statistics based on their belief. | ||
|
MarlieChurphy
United States2063 Posts
Widow mine range needs decrease. Void ray prismatic alignment needs it's buff/debuff timers drastically modified. (possiblly 4-5 seconds buff, 10-30 seconds recharge) Hellbats could use a nerf or two. HP, Armor, but more importantly the way the damage and unit type modifier is layed out. (they need to do less to all and big boost to small/bio) Ravens - not sure on these yet, but I rarely lose a game where I get a nice ball of them added up. Especially in TvT Swarm Hosts with proper dichotomy of AA and creep seem really hard for protoss to deal with. (possible nerf of locust range, damage, or hp with increase in spawn time.) | ||
|
Sissors
1395 Posts
| ||
|
Prplppleatr
United States1518 Posts
Something is off here | ||
|
BronzeKnee
United States5212 Posts
On May 02 2013 03:27 Integra wrote: So by other words, it could mean ANYTHING, and thus is just useless statitsics with no real intent or clear focus or aim. @BronzeKnee; thank you for proving my point regarding speculation and making their own shit up regarding what it it could mean. Yes by systemically disagreeing with you and showing why you were wrong I proved my point... Every statistical discussion on this website turns out like this. People rant and rave and don't think anyone can be objective, that everyone has some hidden agenda... Lovebuzz has it right: On May 02 2013 03:44 LoveBuzz wrote: Actually, what they literally mean is simple: zergs are the least winning race right now. Period. Do we know why? No, but that doesn't matter right now. The important thing is to watch and see if this trend continues or changes. If win rates are different next month or 6 months from now, then there is no cause for alarm. However, if this trend keeps up, THEN we need to look for explanations. Though I think waiting 6 months is way too long. If feasible we should find out now why the winning percentages look the way they do and work to correct them if indeed there is a significant difference based on the sample size. | ||
|
_SpiRaL_
Afghanistan1636 Posts
| ||
|
Luepert
United States1933 Posts
On May 02 2013 02:56 Blezza wrote: Well the EU quals are in there becuase they actaully happened? That's like saying Hitler was a nice guy if it wasn't for the holocaust. EU qualifiers aren't the highest level of play. I'm really only interested in Korean winrates tbh. | ||
|
Sissors
1395 Posts
Actually if this is based on the same spreadsheet that also circulated in other topics that is the case. First of all because we aren't talking here about thousands of games. Second if a single tournament has a very 'unbalanced' win rate that can have a significant influence. Case in point WCS europe qualifiers. Removing that from the list didn't make everything magically at 50%, but it did have a significant influence on the stats. Since there more zergs were invited than terran for example, you can also debate if it should be included. @Bronze, Lovebuzz has it almost correct: Actually, what they literally mean is simple: zergs are the least winning race right now in the games included in these statistics. Period @Luepert, since I am disagreeing with everyone I also disagree with you, for consistency. Two reasons. First of all it is nice if it is balanced for the top 10 worldwide, but also the millions of others like it balanced at their level. And of course at lower levels you can accept larger balance differences, but that doesn't mean it isn't relevant at all. Probably more import for you is reason 2: If you only look at code S the sample size is simply too small to have any relevant data. Then if Life decides to play LoL tomorrow we should boost zerg? And if he then returns 2 months later nerf zerg again? | ||
|
Eventine
United States307 Posts
On May 02 2013 06:04 Luepert wrote: EU qualifiers aren't the highest level of play. I'm really only interested in Korean winrates tbh. Then you should probably construct that data set | ||
|
Eventine
United States307 Posts
On May 02 2013 06:07 Sissors wrote: @Spiral Actually if this is based on the same spreadsheet that also circulated in other topics that is the case. First of all because we aren't talking here about thousands of games. Second if a single tournament has a very 'unbalanced' win rate that can have a significant influence. Case in point WCS europe qualifiers. Removing that from the list didn't make everything magically at 50%, but it did have a significant influence on the stats. Since there more zergs were invited than terran for example, you can also debate if it should be included. @Bronze, Lovebuzz has it almost correct: Can you point me to the spreadsheet? Or better yet, is there spreadsheets for each months results by tournament? Thanks. | ||
|
BronzeKnee
United States5212 Posts
On May 02 2013 06:02 _SpiRaL_ wrote: Why are people pointing to samples of a few dozen results as if that is skewing the results of thousands of games particularly strongly in one way or another? It isn't. ~126,000,000 people voted in the 2012 election. Reuters polls polled an average of less than 5000 voters, and were within 2% of the final results. Just thought you should know that. | ||
|
Sandermatt
Switzerland1365 Posts
| ||
|
BronzeKnee
United States5212 Posts
On May 02 2013 06:07 Sissors wrote: @Spiral Actually if this is based on the same spreadsheet that also circulated in other topics that is the case. First of all because we aren't talking here about thousands of games. Second if a single tournament has a very 'unbalanced' win rate that can have a significant influence. Case in point WCS europe qualifiers. Removing that from the list didn't make everything magically at 50%, but it did have a significant influence on the stats. Since there more zergs were invited than terran for example, you can also debate if it should be included. @Bronze, Lovebuzz has it almost correct: @Luepert, since I am disagreeing with everyone I also disagree with you, for consistency. Two reasons. First of all it is nice if it is balanced for the top 10 worldwide, but also the millions of others like it balanced at their level. And of course at lower levels you can accept larger balance differences, but that doesn't mean it isn't relevant at all. Probably more import for you is reason 2: If you only look at code S the sample size is simply too small to have any relevant data. Then if Life decides to play LoL tomorrow we should boost zerg? And if he then returns 2 months later nerf zerg again? You're right. However, adding such a qualification might be weaker than you think, assuming those tournaments represent nearly all high level play as is stated. We need a lot more information regarding the sample size and tournaments that were used, and how far back this goes... ChaosTerran didn't seem to release much in the Reddit thread. On May 02 2013 06:10 Sandermatt wrote: With this sample size the standard deviation is about 2% (If I calculated properly). So only the TvZ matchup is statistically significant. How did you find out the sample size? | ||
|
JustPassingBy
10776 Posts
edit: Anyways, nothing that was not expected. | ||
|
Adreme
United States5574 Posts
On May 02 2013 06:11 BronzeKnee wrote: You're right. However, adding such a qualification might be weaker than you think, assuming those tournaments represent nearly all high level play as is stated. We need a lot more information regarding the sample size and tournaments that were used, and how far back this goes... ChaosTerran didn't seem to release much in the Reddit thread. How did you find out the sample size? The 6th chart says the sample size. I have to say though I have not seen many matchups go a way different than expected or at least different enough than expected to be considered an upset and that's usually first light. | ||
|
Targe
United Kingdom14103 Posts
| ||
|
BronzeKnee
United States5212 Posts
On May 02 2013 06:16 Adreme wrote: The 6th chart says the sample size. I have to say though I have not seen many matchups go a way different than expected or at least different enough than expected to be considered an upset and that's usually first light. Yeah it is really hard to get it exact though. I found it in the Reddit thread: 415 TvZ games 483 TvP games 555 ZvP games I also found this: Chaos Terran wrote: I had the same problem when working on these statistics. I realized that some tournaments had alot of amateur players in their brackets (Dreamhack, WCS NA Qualifiers, etc..) so I decided to set the parameters for any given tournament in a way that would allow me to sidestep this problem. WCS EU: Ro32+ WCS NA: Round 5+ Dreamhack: Groupstage 2+ All other tournaments either had no amateur players participating or very, very few, so WCS EU, NA and Dreamhack were the only tournaments in which some rounds or group stages did not influence these statistics. edit: Of course, there will always be a skill gap that I cannot account for, but I did my best to sidestep the problem. That is some pretty good work Certainly the data isn't perfect, but it is revealing and the sample size is large enough to warrant more investigation. | ||
|
Sissors
1395 Posts
On May 02 2013 06:09 Eventine wrote: Can you point me to the spreadsheet? Or better yet, is there spreadsheets for each months results by tournament? Thanks. Took some searching (yeah all for you  ), but found it again: https://docs.google.com/spreadsheet/ccc?key=0At0PE4rdhsI9dDE0cEprWkwwMGxQdTczTTlLNW1qX1E#gid=0Main thing to notice imo: If you delete the WCS qualifiers (so EU + NA), then suddenly you get (overall): 49.7% win rate for terran, 48.8% for zerg and 51.5% for toss. So those qualifiers have a huge impact. Not strange since both have TvZ rates at roughly 70%! So yeah I stick to my conclusions that I wouldnt conclude too much from these stats. Edit: It indeeds seems that those stats are from that spreadsheet. So would be nice if the link to the sheet was included in the OP, and in general it is alot more informative than only the pictures | ||
| ||
|
|
SOOP
SKillous vs Spirit
Tenacious Turtle Tussle
PiG Sty Festival
Serral vs TriGGeR
Cure vs SHIN
The PondCast
Replay Cast
PiG Sty Festival
Clem vs Bunny
Solar vs Zoun
Replay Cast
Korean StarCraft League
PiG Sty Festival
herO vs Rogue
ByuN vs SKillous
SC Evo Complete
[ Show More ] |
|
|
EPT
