|
On April 09 2013 01:27 BronzeKnee wrote:Show nested quote +On April 09 2013 01:19 tenklavir wrote:
It is not that we have a problem with data...we have a problem with garbage data and people that use it to attempt to justify their balance whining. Data is defined as "values of qualitative or quantitative variables, belonging to a set of items." In this instance what the OP showed fits the definition of data. And then there are things which aren't data which people sometimes call data. I assume this is what you mean regarding garbage data. This is akin to the fact that people have been calling themselves Christians for thousands of years and using that to justify terrible deeds which are decidedly against the Christian faith. People have and will try and abuse something of authority in order to convince others of something, science and statistics are not immune to this. But just because someone calls it Christianity, statistics, science or whatever, doesn't mean it actually is. The solution to this problem is not the that Christianity, statistics or science are bad and shouldn't be used because people have abused them, the problem lies in the people that abuse them. And we need to call them out and correct them, which is what I've been trying to do the whole time.
It's garbage because of sample size, noise, attempts to correlate random variables on arbitrarily chosen parameters...shall I continue?
Odd that you want to call people out when you conveniently ignored my questions, so I'll try again:
For instance you asked the OP to add the Widow Mine % of Active Army @ 15:00 data to the OP, describing it as "revealing". What does it reveal, exactly? Especially when one page before that I show you that the Rsq of unweighted correlation between WM% and Win rate is 0.097. From this you can effectively conclude that WM% has no effect on Win rate. What else could you possibly want to discuss from that?
|
On April 09 2013 01:32 tenklavir wrote:Show nested quote +On April 09 2013 01:27 BronzeKnee wrote:On April 09 2013 01:19 tenklavir wrote:
It is not that we have a problem with data...we have a problem with garbage data and people that use it to attempt to justify their balance whining. Data is defined as "values of qualitative or quantitative variables, belonging to a set of items." In this instance what the OP showed fits the definition of data. And then there are things which aren't data which people sometimes call data. I assume this is what you mean regarding garbage data. This is akin to the fact that people have been calling themselves Christians for thousands of years and using that to justify terrible deeds which are decidedly against the Christian faith. People have and will try and abuse something of authority in order to convince others of something, science and statistics are not immune to this. But just because someone calls it Christianity, statistics, science or whatever, doesn't mean it actually is. The solution to this problem is not the that Christianity, statistics or science are bad and shouldn't be used because people have abused them, the problem lies in the people that abuse them. And we need to call them out and correct them, which is what I've been trying to do the whole time. It's garbage because of sample size, noise, attempts to correlate random variables on arbitrarily chosen parameters...shall I continue? Odd that you want to call people out when you conveniently ignored my questions, so I'll try again: For instance you asked the OP to add the Widow Mine % of Active Army @ 15:00 data to the OP, describing it as "revealing". What does it reveal, exactly? Especially when one page before that I show you that the Rsq of unweighted correlation between WM% and Win rate is 0.097. From this you can effectively conclude that WM% has no effect on Win rate. What else could you possibly want to discuss from that?
Adding the chart to the OP made what you just described above clear to everyone. It also showed a different problem, Terran is winning a lot than Zerg in TvZ (which your data did not show), though we don't know if it is statistically significant yet! That is why it should be added to the OP.
Now what I find odd is that you conveniently ignored my entire post.
|
I mean, heck, the OP doesn't really even qualify its proposed findings (other than saying it doesn't prove anything, which is true of everything that has ever happened in reality). It'd be a really good move to put in a little blurb stating the potential pitfalls of this sort of analysis IMO.
|
On April 09 2013 01:35 BronzeKnee wrote:Show nested quote +On April 09 2013 01:32 tenklavir wrote:On April 09 2013 01:27 BronzeKnee wrote:On April 09 2013 01:19 tenklavir wrote:
It is not that we have a problem with data...we have a problem with garbage data and people that use it to attempt to justify their balance whining. Data is defined as "values of qualitative or quantitative variables, belonging to a set of items." In this instance what the OP showed fits the definition of data. And then there are things which aren't data which people sometimes call data. I assume this is what you mean regarding garbage data. This is akin to the fact that people have been calling themselves Christians for thousands of years and using that to justify terrible deeds which are decidedly against the Christian faith. People have and will try and abuse something of authority in order to convince others of something, science and statistics are not immune to this. But just because someone calls it Christianity, statistics, science or whatever, doesn't mean it actually is. The solution to this problem is not the that Christianity, statistics or science are bad and shouldn't be used because people have abused them, the problem lies in the people that abuse them. And we need to call them out and correct them, which is what I've been trying to do the whole time. It's garbage because of sample size, noise, attempts to correlate random variables on arbitrarily chosen parameters...shall I continue? Odd that you want to call people out when you conveniently ignored my questions, so I'll try again: For instance you asked the OP to add the Widow Mine % of Active Army @ 15:00 data to the OP, describing it as "revealing". What does it reveal, exactly? Especially when one page before that I show you that the Rsq of unweighted correlation between WM% and Win rate is 0.097. From this you can effectively conclude that WM% has no effect on Win rate. What else could you possibly want to discuss from that? Adding the chart to the OP made what you just described above clear to everyone. It also showed a different problem, Terran is winning a lot than Zerg in TvZ. That is why it should be added to the OP.
No it doesn't. None of the very-basic correlation analysis I posted was added to the OP which would immediately tell anyone looking at it why it's all bunk.
I can't tell if you're intentionally being dense or you truly lack the mathematical background to understand what is being discussed here.
It also showed a different problem, Terran is winning a lot than Zerg in TvZ. That is why it should be added to the OP
On what basis are you concluding this? Please point to the exact metric and parameters you are using to make this claim.
|
On April 09 2013 01:39 tenklavir wrote:Show nested quote +On April 09 2013 01:35 BronzeKnee wrote:On April 09 2013 01:32 tenklavir wrote:On April 09 2013 01:27 BronzeKnee wrote:On April 09 2013 01:19 tenklavir wrote:
It is not that we have a problem with data...we have a problem with garbage data and people that use it to attempt to justify their balance whining. Data is defined as "values of qualitative or quantitative variables, belonging to a set of items." In this instance what the OP showed fits the definition of data. And then there are things which aren't data which people sometimes call data. I assume this is what you mean regarding garbage data. This is akin to the fact that people have been calling themselves Christians for thousands of years and using that to justify terrible deeds which are decidedly against the Christian faith. People have and will try and abuse something of authority in order to convince others of something, science and statistics are not immune to this. But just because someone calls it Christianity, statistics, science or whatever, doesn't mean it actually is. The solution to this problem is not the that Christianity, statistics or science are bad and shouldn't be used because people have abused them, the problem lies in the people that abuse them. And we need to call them out and correct them, which is what I've been trying to do the whole time. It's garbage because of sample size, noise, attempts to correlate random variables on arbitrarily chosen parameters...shall I continue? Odd that you want to call people out when you conveniently ignored my questions, so I'll try again: For instance you asked the OP to add the Widow Mine % of Active Army @ 15:00 data to the OP, describing it as "revealing". What does it reveal, exactly? Especially when one page before that I show you that the Rsq of unweighted correlation between WM% and Win rate is 0.097. From this you can effectively conclude that WM% has no effect on Win rate. What else could you possibly want to discuss from that? Adding the chart to the OP made what you just described above clear to everyone. It also showed a different problem, Terran is winning a lot than Zerg in TvZ. That is why it should be added to the OP. No it doesn't. None of the very-basic correlation analysis I posted was added to the OP which would immediately tell anyone looking at it why it's all bunk. I can't tell if you're intentionally being dense or you truly lack the mathematical background to understand what is being discussed here. Show nested quote +It also showed a different problem, Terran is winning a lot than Zerg in TvZ. That is why it should be added to the OP On what basis are you concluding this? Please point to the exact metric and parameters you are using to make this claim.
So you said: "For instance you asked the OP to add the Widow Mine % of Active Army @ 15:00 data to the OP, describing it as "revealing". What does it reveal, exactly? Especially when one page before that I show you that the Rsq of unweighted correlation between WM% and Win rate is 0.097."
What your saying (correct me if I am wrong) is you found that that "WM% has no effect on Win rate" (in your own words). And the chart basically shows that (though there is some deviation, but I haven't calculated whether or not it is a significant difference).
So you're asking me what the chart reveals that your calculation didn't. And I answered that (chart says T wins more than Z).
Then you just ask on what basis am I concluding that T wins more than Z. That is based on the chart which again, says T wins more than Z.
Is the chart wrong? No, based on his statistics that he collected that is what the chart says. But let us discuss it and why his data might be skewed if it is or how we can explain said winrate and show that it isn't an issue. This is a discussion forum after all! Do you have anything useful to add?
I'd like to see data based on Medivac usage, especially those who use afterburners (though I'm not sure he can get data based on afterburner usage).
|
|
|
Oh mines are op. Who would've thought lol.
|
On April 09 2013 01:44 BronzeKnee wrote:Show nested quote +On April 09 2013 01:39 tenklavir wrote:On April 09 2013 01:35 BronzeKnee wrote:On April 09 2013 01:32 tenklavir wrote:On April 09 2013 01:27 BronzeKnee wrote:On April 09 2013 01:19 tenklavir wrote:
It is not that we have a problem with data...we have a problem with garbage data and people that use it to attempt to justify their balance whining. Data is defined as "values of qualitative or quantitative variables, belonging to a set of items." In this instance what the OP showed fits the definition of data. And then there are things which aren't data which people sometimes call data. I assume this is what you mean regarding garbage data. This is akin to the fact that people have been calling themselves Christians for thousands of years and using that to justify terrible deeds which are decidedly against the Christian faith. People have and will try and abuse something of authority in order to convince others of something, science and statistics are not immune to this. But just because someone calls it Christianity, statistics, science or whatever, doesn't mean it actually is. The solution to this problem is not the that Christianity, statistics or science are bad and shouldn't be used because people have abused them, the problem lies in the people that abuse them. And we need to call them out and correct them, which is what I've been trying to do the whole time. It's garbage because of sample size, noise, attempts to correlate random variables on arbitrarily chosen parameters...shall I continue? Odd that you want to call people out when you conveniently ignored my questions, so I'll try again: For instance you asked the OP to add the Widow Mine % of Active Army @ 15:00 data to the OP, describing it as "revealing". What does it reveal, exactly? Especially when one page before that I show you that the Rsq of unweighted correlation between WM% and Win rate is 0.097. From this you can effectively conclude that WM% has no effect on Win rate. What else could you possibly want to discuss from that? Adding the chart to the OP made what you just described above clear to everyone. It also showed a different problem, Terran is winning a lot than Zerg in TvZ. That is why it should be added to the OP. No it doesn't. None of the very-basic correlation analysis I posted was added to the OP which would immediately tell anyone looking at it why it's all bunk. I can't tell if you're intentionally being dense or you truly lack the mathematical background to understand what is being discussed here. It also showed a different problem, Terran is winning a lot than Zerg in TvZ. That is why it should be added to the OP On what basis are you concluding this? Please point to the exact metric and parameters you are using to make this claim. So you said: "For instance you asked the OP to add the Widow Mine % of Active Army @ 15:00 data to the OP, describing it as "revealing". What does it reveal, exactly? Especially when one page before that I show you that the Rsq of unweighted correlation between WM% and Win rate is 0.097." What your saying (correct me if I am wrong) is you found that that WM% has no effect on Win rate. And the chart basically shows that (though there is some deviation, but I haven't calculated whether or not it is a significant difference). So you're asking me what the chart reveals that your calculation didn't. And I answered that (T wins more than Z). Then you just ask on what basis am I concluding that T wins more than Z. That is based on the chart.
Well, let's be more specific. When you say "T wins more than Z", what do you really mean? Are you saying that:
- Terran players, when playing against Zerg players, win more than 50% of the time across the entire population of SC2 players?
- The above statement, for only the top N% of players with regards to skill*?
- A Terran player, when playing Zerg player of equal skill*, will win more than 50% of the time?
* note that skill in SC2 is still very much ill-defined, so it's going to be very tough for you to claim that this sort of conclusion is supported by any extant evidence
|
On April 09 2013 01:49 shaldengeki wrote:Show nested quote +On April 09 2013 01:44 BronzeKnee wrote:On April 09 2013 01:39 tenklavir wrote:On April 09 2013 01:35 BronzeKnee wrote:On April 09 2013 01:32 tenklavir wrote:On April 09 2013 01:27 BronzeKnee wrote:On April 09 2013 01:19 tenklavir wrote:
It is not that we have a problem with data...we have a problem with garbage data and people that use it to attempt to justify their balance whining. Data is defined as "values of qualitative or quantitative variables, belonging to a set of items." In this instance what the OP showed fits the definition of data. And then there are things which aren't data which people sometimes call data. I assume this is what you mean regarding garbage data. This is akin to the fact that people have been calling themselves Christians for thousands of years and using that to justify terrible deeds which are decidedly against the Christian faith. People have and will try and abuse something of authority in order to convince others of something, science and statistics are not immune to this. But just because someone calls it Christianity, statistics, science or whatever, doesn't mean it actually is. The solution to this problem is not the that Christianity, statistics or science are bad and shouldn't be used because people have abused them, the problem lies in the people that abuse them. And we need to call them out and correct them, which is what I've been trying to do the whole time. It's garbage because of sample size, noise, attempts to correlate random variables on arbitrarily chosen parameters...shall I continue? Odd that you want to call people out when you conveniently ignored my questions, so I'll try again: For instance you asked the OP to add the Widow Mine % of Active Army @ 15:00 data to the OP, describing it as "revealing". What does it reveal, exactly? Especially when one page before that I show you that the Rsq of unweighted correlation between WM% and Win rate is 0.097. From this you can effectively conclude that WM% has no effect on Win rate. What else could you possibly want to discuss from that? Adding the chart to the OP made what you just described above clear to everyone. It also showed a different problem, Terran is winning a lot than Zerg in TvZ. That is why it should be added to the OP. No it doesn't. None of the very-basic correlation analysis I posted was added to the OP which would immediately tell anyone looking at it why it's all bunk. I can't tell if you're intentionally being dense or you truly lack the mathematical background to understand what is being discussed here. It also showed a different problem, Terran is winning a lot than Zerg in TvZ. That is why it should be added to the OP On what basis are you concluding this? Please point to the exact metric and parameters you are using to make this claim. So you said: "For instance you asked the OP to add the Widow Mine % of Active Army @ 15:00 data to the OP, describing it as "revealing". What does it reveal, exactly? Especially when one page before that I show you that the Rsq of unweighted correlation between WM% and Win rate is 0.097." What your saying (correct me if I am wrong) is you found that that WM% has no effect on Win rate. And the chart basically shows that (though there is some deviation, but I haven't calculated whether or not it is a significant difference). So you're asking me what the chart reveals that your calculation didn't. And I answered that (T wins more than Z). Then you just ask on what basis am I concluding that T wins more than Z. That is based on the chart. Well, let's be more specific. When you say "T wins more than Z", what do you really mean? Are you saying that: - Terran players, when playing against Zerg players, win more than 50% of the time across the entire population of SC2 players?
- The above statement, for only the top N% of players with regards to skill*?
- A Terran player, when playing Zerg player of equal skill*, will win more than 50% of the time?
* note that skill in SC2 is still very much ill-defined, so it's going to be very tough for you to claim that this sort of conclusion is supported by any extant evidence
Let's us be really clear. I didn't say anything, the chart said it. I am just repeating it. If I wasn't on this forum right now, it would still say exactly the same thing. It is independent of me.
![[image loading]](http://i.imgur.com/8Y75XGl.png)
This isn't you versus me or him or whatever, this is us trying to understand the data and the game.
Now the chart didn't answer any of the questions you asked, but maybe we can get some answers, because it would sure be useful. Just make sure there isn't any bias here in the questions you are asking. We aren't trying to show everything is or isn't balanced, we are trying to find out if things are balanced or not.
|
Now compare this with a different analysis of, say, 15, 20 and 25+ widow mines to see if the winrate continues up the trend. By your assertion, the win rate should be higher as more widow mines are built.
I'd be interested to see this data actually.
|
![[image loading]](http://i.imgur.com/JIgXeQb.png)
I'm kind of confused. The first one shows 961 TvZ games in master's league with ( 616/961 x 0.49 + 345/961 x 0.59 ) = 52.6% winrate, while the second one shows 736 games and 425 wins = 57.7% winrate. Are these two different sets of data, or were some games excluded from the second sample?
|
This may be a bit OT but I would be very interested on the PvZ winrate for protoss presented in terms of # of void rays (or even air units in general) that are made. Could show a trend that Zerg really lacks anti-air if the winrate is high enough.
|
On April 09 2013 01:50 BronzeKnee wrote:Show nested quote +On April 09 2013 01:49 shaldengeki wrote:On April 09 2013 01:44 BronzeKnee wrote:On April 09 2013 01:39 tenklavir wrote:On April 09 2013 01:35 BronzeKnee wrote:On April 09 2013 01:32 tenklavir wrote:On April 09 2013 01:27 BronzeKnee wrote:On April 09 2013 01:19 tenklavir wrote:
It is not that we have a problem with data...we have a problem with garbage data and people that use it to attempt to justify their balance whining. Data is defined as "values of qualitative or quantitative variables, belonging to a set of items." In this instance what the OP showed fits the definition of data. And then there are things which aren't data which people sometimes call data. I assume this is what you mean regarding garbage data. This is akin to the fact that people have been calling themselves Christians for thousands of years and using that to justify terrible deeds which are decidedly against the Christian faith. People have and will try and abuse something of authority in order to convince others of something, science and statistics are not immune to this. But just because someone calls it Christianity, statistics, science or whatever, doesn't mean it actually is. The solution to this problem is not the that Christianity, statistics or science are bad and shouldn't be used because people have abused them, the problem lies in the people that abuse them. And we need to call them out and correct them, which is what I've been trying to do the whole time. It's garbage because of sample size, noise, attempts to correlate random variables on arbitrarily chosen parameters...shall I continue? Odd that you want to call people out when you conveniently ignored my questions, so I'll try again: For instance you asked the OP to add the Widow Mine % of Active Army @ 15:00 data to the OP, describing it as "revealing". What does it reveal, exactly? Especially when one page before that I show you that the Rsq of unweighted correlation between WM% and Win rate is 0.097. From this you can effectively conclude that WM% has no effect on Win rate. What else could you possibly want to discuss from that? Adding the chart to the OP made what you just described above clear to everyone. It also showed a different problem, Terran is winning a lot than Zerg in TvZ. That is why it should be added to the OP. No it doesn't. None of the very-basic correlation analysis I posted was added to the OP which would immediately tell anyone looking at it why it's all bunk. I can't tell if you're intentionally being dense or you truly lack the mathematical background to understand what is being discussed here. It also showed a different problem, Terran is winning a lot than Zerg in TvZ. That is why it should be added to the OP On what basis are you concluding this? Please point to the exact metric and parameters you are using to make this claim. So you said: "For instance you asked the OP to add the Widow Mine % of Active Army @ 15:00 data to the OP, describing it as "revealing". What does it reveal, exactly? Especially when one page before that I show you that the Rsq of unweighted correlation between WM% and Win rate is 0.097." What your saying (correct me if I am wrong) is you found that that WM% has no effect on Win rate. And the chart basically shows that (though there is some deviation, but I haven't calculated whether or not it is a significant difference). So you're asking me what the chart reveals that your calculation didn't. And I answered that (T wins more than Z). Then you just ask on what basis am I concluding that T wins more than Z. That is based on the chart. Well, let's be more specific. When you say "T wins more than Z", what do you really mean? Are you saying that: - Terran players, when playing against Zerg players, win more than 50% of the time across the entire population of SC2 players?
- The above statement, for only the top N% of players with regards to skill*?
- A Terran player, when playing Zerg player of equal skill*, will win more than 50% of the time?
* note that skill in SC2 is still very much ill-defined, so it's going to be very tough for you to claim that this sort of conclusion is supported by any extant evidence Let's us be really clear. I didn't say anything, the chart said it. I am just repeating it. If I wasn't on this forum right now, it would still say exactly the same thing. It is independent of me. This isn't you versus me or him or whatever, this is us trying to understand the data and the game. Now the chart didn't answer any of the questions you asked, but maybe we can get some answers, because it would sure be useful. Just make sure there isn't any bias here in the questions you are asking. We aren't trying to show everything is or isn't balanced, we are trying to find out if things are balanced or not.
The problem with this is that the chart doesn't say that T wins more against Z. It says that these games happened, not that they're representative of (or generalize-able to) any larger superset of games between players. So you see, what you're specifically claiming when you say "T wins more than Z" is really important, because it's clearly a more general statement than what the chart indicates!
|
On April 09 2013 01:50 BronzeKnee wrote:Show nested quote +On April 09 2013 01:49 shaldengeki wrote:On April 09 2013 01:44 BronzeKnee wrote:On April 09 2013 01:39 tenklavir wrote:On April 09 2013 01:35 BronzeKnee wrote:On April 09 2013 01:32 tenklavir wrote:On April 09 2013 01:27 BronzeKnee wrote:On April 09 2013 01:19 tenklavir wrote:
It is not that we have a problem with data...we have a problem with garbage data and people that use it to attempt to justify their balance whining. Data is defined as "values of qualitative or quantitative variables, belonging to a set of items." In this instance what the OP showed fits the definition of data. And then there are things which aren't data which people sometimes call data. I assume this is what you mean regarding garbage data. This is akin to the fact that people have been calling themselves Christians for thousands of years and using that to justify terrible deeds which are decidedly against the Christian faith. People have and will try and abuse something of authority in order to convince others of something, science and statistics are not immune to this. But just because someone calls it Christianity, statistics, science or whatever, doesn't mean it actually is. The solution to this problem is not the that Christianity, statistics or science are bad and shouldn't be used because people have abused them, the problem lies in the people that abuse them. And we need to call them out and correct them, which is what I've been trying to do the whole time. It's garbage because of sample size, noise, attempts to correlate random variables on arbitrarily chosen parameters...shall I continue? Odd that you want to call people out when you conveniently ignored my questions, so I'll try again: For instance you asked the OP to add the Widow Mine % of Active Army @ 15:00 data to the OP, describing it as "revealing". What does it reveal, exactly? Especially when one page before that I show you that the Rsq of unweighted correlation between WM% and Win rate is 0.097. From this you can effectively conclude that WM% has no effect on Win rate. What else could you possibly want to discuss from that? Adding the chart to the OP made what you just described above clear to everyone. It also showed a different problem, Terran is winning a lot than Zerg in TvZ. That is why it should be added to the OP. No it doesn't. None of the very-basic correlation analysis I posted was added to the OP which would immediately tell anyone looking at it why it's all bunk. I can't tell if you're intentionally being dense or you truly lack the mathematical background to understand what is being discussed here. It also showed a different problem, Terran is winning a lot than Zerg in TvZ. That is why it should be added to the OP On what basis are you concluding this? Please point to the exact metric and parameters you are using to make this claim. So you said: "For instance you asked the OP to add the Widow Mine % of Active Army @ 15:00 data to the OP, describing it as "revealing". What does it reveal, exactly? Especially when one page before that I show you that the Rsq of unweighted correlation between WM% and Win rate is 0.097." What your saying (correct me if I am wrong) is you found that that WM% has no effect on Win rate. And the chart basically shows that (though there is some deviation, but I haven't calculated whether or not it is a significant difference). So you're asking me what the chart reveals that your calculation didn't. And I answered that (T wins more than Z). Then you just ask on what basis am I concluding that T wins more than Z. That is based on the chart. Well, let's be more specific. When you say "T wins more than Z", what do you really mean? Are you saying that: - Terran players, when playing against Zerg players, win more than 50% of the time across the entire population of SC2 players?
- The above statement, for only the top N% of players with regards to skill*?
- A Terran player, when playing Zerg player of equal skill*, will win more than 50% of the time?
* note that skill in SC2 is still very much ill-defined, so it's going to be very tough for you to claim that this sort of conclusion is supported by any extant evidence Let's us be really clear. I didn't say anything, the chart said it. I am just repeating it. If I wasn't on this forum right now, it would still say exactly the same thing. It is independent of me. This isn't you versus me or him or whatever, this is us trying to understand the data and the game. Now the chart didn't answer any of the questions you asked, but maybe we can get some answers, because it would sure be useful. Just make sure there isn't any bias here in the questions you are asking. We aren't trying to show everything is or isn't balanced, we are trying to find out if things are balanced or not. i quoted this post because it contains the graph, i'm not saying anything in relation to quoted post.
When looking at the graph, there are a number of things that cross my mind.
1. sample sizes go down fast with rising widow mine usage. The effects of this have to be taken into account (larger uncertainty of the data point.)
2. How is the widow mine percentage defined? Is it via overall recources? Or supply? When the rest of the army is only marines, this doesn't have an effect, but when medivacs and a somewhat larger amount of marauders come into play, it might be worth it having a clear view on this.
3. Hots is about a month old. I think it isn't a strange thought that many experiments took place right after launch, more so than now. If the date is accumulated over the course of hots from release onward, i would be interested to see the change in composition over the course of the month. If for example (might not be real, but w/e) mass widow mines would be shut down hard by a trick, developed only a week ago, i feel this will taint the percentages in the graph.
Maybe these have been addressed already.
When i look at the graph itself (only percentages), i'd say the conclusion that more widow mines lead to a higher win rate is quite weak, seeing the lowest 3 rates are at in the 15%-35% range. I haven't calculated anything, but i think this graph alone won't be enough to prove such a statement.
|
On April 09 2013 01:50 BronzeKnee wrote:Show nested quote +On April 09 2013 01:49 shaldengeki wrote:On April 09 2013 01:44 BronzeKnee wrote:On April 09 2013 01:39 tenklavir wrote:On April 09 2013 01:35 BronzeKnee wrote:On April 09 2013 01:32 tenklavir wrote:On April 09 2013 01:27 BronzeKnee wrote:On April 09 2013 01:19 tenklavir wrote:
It is not that we have a problem with data...we have a problem with garbage data and people that use it to attempt to justify their balance whining. Data is defined as "values of qualitative or quantitative variables, belonging to a set of items." In this instance what the OP showed fits the definition of data. And then there are things which aren't data which people sometimes call data. I assume this is what you mean regarding garbage data. This is akin to the fact that people have been calling themselves Christians for thousands of years and using that to justify terrible deeds which are decidedly against the Christian faith. People have and will try and abuse something of authority in order to convince others of something, science and statistics are not immune to this. But just because someone calls it Christianity, statistics, science or whatever, doesn't mean it actually is. The solution to this problem is not the that Christianity, statistics or science are bad and shouldn't be used because people have abused them, the problem lies in the people that abuse them. And we need to call them out and correct them, which is what I've been trying to do the whole time. It's garbage because of sample size, noise, attempts to correlate random variables on arbitrarily chosen parameters...shall I continue? Odd that you want to call people out when you conveniently ignored my questions, so I'll try again: For instance you asked the OP to add the Widow Mine % of Active Army @ 15:00 data to the OP, describing it as "revealing". What does it reveal, exactly? Especially when one page before that I show you that the Rsq of unweighted correlation between WM% and Win rate is 0.097. From this you can effectively conclude that WM% has no effect on Win rate. What else could you possibly want to discuss from that? Adding the chart to the OP made what you just described above clear to everyone. It also showed a different problem, Terran is winning a lot than Zerg in TvZ. That is why it should be added to the OP. No it doesn't. None of the very-basic correlation analysis I posted was added to the OP which would immediately tell anyone looking at it why it's all bunk. I can't tell if you're intentionally being dense or you truly lack the mathematical background to understand what is being discussed here. It also showed a different problem, Terran is winning a lot than Zerg in TvZ. That is why it should be added to the OP On what basis are you concluding this? Please point to the exact metric and parameters you are using to make this claim. So you said: "For instance you asked the OP to add the Widow Mine % of Active Army @ 15:00 data to the OP, describing it as "revealing". What does it reveal, exactly? Especially when one page before that I show you that the Rsq of unweighted correlation between WM% and Win rate is 0.097." What your saying (correct me if I am wrong) is you found that that WM% has no effect on Win rate. And the chart basically shows that (though there is some deviation, but I haven't calculated whether or not it is a significant difference). So you're asking me what the chart reveals that your calculation didn't. And I answered that (T wins more than Z). Then you just ask on what basis am I concluding that T wins more than Z. That is based on the chart. Well, let's be more specific. When you say "T wins more than Z", what do you really mean? Are you saying that: - Terran players, when playing against Zerg players, win more than 50% of the time across the entire population of SC2 players?
- The above statement, for only the top N% of players with regards to skill*?
- A Terran player, when playing Zerg player of equal skill*, will win more than 50% of the time?
* note that skill in SC2 is still very much ill-defined, so it's going to be very tough for you to claim that this sort of conclusion is supported by any extant evidence Let's us be really clear. I didn't say anything, the chart said it. I am just repeating it. If I wasn't on this forum right now, it would still say exactly the same thing. It is independent of me. This isn't you versus me or him or whatever, this is us trying to understand the data and the game. Now the chart didn't answer any of the questions you asked, but maybe we can get some answers, because it would sure be useful. Just make sure there isn't any bias here in the questions you are asking. We aren't trying to show everything is or isn't balanced, we are trying to find out if things are balanced or not.
Looking at that chart, I see absolutely no correlation that would suggest that Widow Mine is imbalanced. Your data is all over the place and generally hovers in the mid-50 % winrate, which is perfectly fine.
|
On April 09 2013 02:18 Wen_Jie wrote:
Are these two different sets of data, or were some games excluded from the second sample?
The second set have only the games that lasted to 15:00.
|
Im pretty sure if you ran the stats for any many other units, you would see a similar %
for example, try games where T makes 10+ medivacs vs T makes less than 10 medivacs. I bet it shows 10+ medivacs will have a higher win %.
or 10+ tanks, or 10+ ravens, or 10+ BCs
|
On April 08 2013 23:25 dsjoerg wrote:This has been lots of fun and I will be doing more stats posts at some point. To make the next discussion more productive and a little less flaming: - Would you be willing to review a stats post in advance? There were several insightful / valid / intelligent comments. "Peer review" will help ensure that the post is better thought out before wide circulation. If so, please PM me.
- Are there other stats you'd like to see? I can run custom queries for you and you can write your own post based on them. There's a lot of info in GGTracker (my site). PM me...
Could you break the data down a little further based on what strategies zerg have employed? For example, could you determine the win rates, in games where a T builds 10 or more mines, and the zerg player builds, says, 6 to 10 swarm hosts? 10 - 20 swarm hosts? I'm curious to know whether swarm hosts are an answer to mine usage. I saw a GSTL game where a small number of SHs were used to trigger mines, etc. The zerg player lost but it seemed like it might be an efficient way to deal with the mines. Also I'd be curious to see how effective mines are v. roach hydra viper compositions.
|
bleh these stats posts are riddled with incompetent statistics people.
Drawing much conclusions from these data is horrible though, such correlations have very little value for determining problems. Winrates itself can be really useful, almost anything else should be avoided by non-experts really, you can't really say anything useful about causation with these sort of data gathering. Not even mentioning all the forms of bias which can't be excluded with this sort of data gathering.
|
Please do not set your mind one way or the other based on this "data".
|
|
|
|
|
|
EPT
