licker
Joined: Jul 10, 2009
|
 Posted:
Dec 13, 2011 - 00:37 |
 
|
| Garion wrote: | Is there any need to have this thread?
This is just a two way conversation, I'm sure it could all be worked out using private messaging.
Lock it before it reaches 20 pages please. |
If you don't want to read it don't read it. Why the urge to take away everyone elses entertainment? |
|
|
dode74
Joined: Aug 14, 2009
|
| Posted:
Dec 13, 2011 - 00:55 |
|
| VoodooMike wrote: | | Seems to have plenty of self-conceived validity as far as rejecting the results of tests, though, doesn't it? |
Who is rejecting the results of tests? I am rejecting your null hypothesis as out of context.
| VoodooMike wrote: | | No, what you did was imply that you had actually tested and gotten results with a smaller, less accurate dataset shortly after making a big show of talking about how unusable similarly aggregated data from FUMBBL was. It's duplicitous, and its your usual soft-wording crap. |
What you infer and what I imply are two different things. Don't go attributing your inferences to me.
| VoodooMike wrote: | | You can't test anything with ambiguous success criteria... not with a straight face, anyway. If your hypothesis is that certain inducements cause a reversal of the given trend, then you specify which inducements, and you look at the TV levels they become available. |
Certainly. When I have some data.
| VoodooMike wrote: | | The default position should be that the data we do have remains consistent, until such time as we see evidence to the contrary. |
Only if the environment remains the same, which it does not. It is nonsense to say that the speed of sound is the same at sea level as it is at 60,000 feet because the environment changes.
| VoodooMike wrote: | | The relationship is curvilinear, leveling off around 40%, as far as I can see. |
What makes you think it levels off? Why can the curve not continue curving? The margins of error are such that you simply cannot say. Personally I don't think the general trend reverses, but I'm not going to assume it doesn't.
| VoodooMike wrote: | | Will it remain so? I don't know, but I see no evidence to say it will not, and no logical basis for imagining it will, so my default position will be "no, it won't", and so should everyone elses because that's the only position for which there is any supporting data. There not being any supporting data does not give support to alternate theories, and that seems to be what you argue for, despite claiming you don't. |
Is there or isn't there supporting data? Because if there isn't then the only sensible position is "I don't know". If there is then please allow us to review it.
The data ([B ] data) is not a random sample of all the games of BB ever played (the population). It has a selection bias for low-TV difference games; these are games with minimal amounts of inducements. As such, this compromises the external validity of the data for which we have confidence, meaning your extrapolation is compromised.
I expect you to reply to this, of course, but I see we're going in circles now. You think the extrapolation is valid and I do not know. I suggest we wait for some data. |
|
|
koadah
Joined: Mar 30, 2005
|
| Posted:
Dec 13, 2011 - 01:21 |
|
|
VoodooMike
Joined: Nov 07, 2010
|
| Posted:
Dec 13, 2011 - 10:14 |
|
I could sit there are reply to your waffling (and I did, but realized you're just talking in circles again) but the fact is it can all be broken down into the following:
- Your arguments about my data and methods are insincere, since you used the same data and similar (but less precise) methods in your own stated investigation. Most likely, you're just trying to move the spotlight away from the fact that:
- You don't actually have an opposing hypothesis beyond "nuh uh" at present, based on a vague statement made by Galak that you've interpreted to mean something should be different, maybe.
So what are you arguing for, exactly? Most likely you're stalling for time to try to get something approaching elaboration from Galak so you can pretend you had a solid hypothesis from the start, rather than "something, but I don't know what, will happen, but I don't know when" which is where you stand at the moment.
So here's one last piece of rhetoric for you, pending you having a solid hypothesis (which comes before you even start to worry about data):
"I don't pretend to have all the answers. I don't even pretend to know all the questions. Hey, where am I?" - Dode74/Jack Handy |
|
|
VoodooMike
Joined: Nov 07, 2010
|
| Posted:
Dec 13, 2011 - 10:23 |
|
| koadah wrote: | | So, I did a bit of eye balling. |
Yeah, lets send you to a quick and dirty CI calculator, koadah, and you can throw in the 95% confidence interval for your data:
Use the second box, leave population blank, sample size is your p, percentage is your win percentage
Then put a +/- the number it calculates in the Confidence Interval box and see how you feel about your eyeballing then.
Example:
900 gap- win% is 66.67 +/- 53.34%
400 gap- win% is 32.69 +/- 18.03%
So yeah, we're 95% sure that the average win chance for the underdog at a 900 TV difference lies somewhere between about 13% and 100%.
A solid win for intuition! |
|
|
dode74
Joined: Aug 14, 2009
|
| Posted:
Dec 13, 2011 - 11:18 |
|
| Quote: | | So what are you arguing for, exactly? |
That your extrapolation is invalid. As I said before, the data ([B ] data) is not a random sample of all the games of BB ever played (the population). It has a selection bias for low-TV difference games; these are games with minimal amounts of inducements. As such, this compromises the external validity of the data for which we have confidence, meaning your extrapolation is compromised.
| Quote: | | Your arguments about my data and methods are insincere, since you used the same data and similar (but less precise) methods in your own stated investigation. |
I don't extrapolate beyond the confidence of the data though. It's the extrapolation I take issue with.
| Quote: | | You don't actually have an opposing hypothesis beyond "nuh uh" at present |
Do I need one to be able to say that your choice of null hypothesis is invalid? |
|
|
koadah
Joined: Mar 30, 2005
|
| Posted:
Dec 13, 2011 - 11:20 |
|
|
JimmyFantastic
Joined: Feb 06, 2007
|
| Posted:
Dec 13, 2011 - 11:25 |
|
hahaha dode is never gonna give up, I love it! |
_________________
Pull down the veil - actively bad for the hobby! |
|
VoodooMike
Joined: Nov 07, 2010
|
| Posted:
Dec 13, 2011 - 11:27 |
|
| dode74 wrote: | | That your extrapolation is invalid. As I said before, the data ([B ] data) is not a random sample of all the games of BB ever played (the population). It has a selection bias for low-TV difference games; these are games with minimal amounts of inducements. As such, this compromises the external validity of the data for which we have confidence, meaning your extrapolation is compromised. |
You can (and likely will, if it doesnt bear out for what you're trying to find) use that excuse nomatter what TV ranges appear since the data will always be "not what the game was designed around" owing to the TV matching system. The TV ranges will always cluster around even, resulting in remarkably little data for high ranges.
It's the creationist's argument again - have we looked everywhere to see if God is there? No? Then we can't say he isn't!
The extrapolation is based on the only data that exists. It makes it a more valid extrapolation than any other that might be created based on... what?
| dode74 wrote: | | I don't extrapolate beyond the confidence of the data though. It's the extrapolation I take issue with. |
Every single thing you said you investigated is going being the 95% confidence interval, so yeah, you do. You didn't even know about it until I corrected you, I might add - you're learning this stuff one mistake and wiki lookup at a time, so lets not pretend the topic isn't perpetually at the limit of your understanding of the subject.
| dode74 wrote: | | Do I need one to be able to say that your choice of null hypothesis is invalid? |
Yes, actually, you do - if you lack a hypothesis then you can't say that any null hypothesis is inappropriate.
So tell me.. what precisely are you seeking to investigate, and is it really a statistical question? You've gotten more and more vague on that since you began arguing, and now you're sinking into circular softspeak. |
|
|
dode74
Joined: Aug 14, 2009
|
| Posted:
Dec 13, 2011 - 21:00 |
|
| Quote: | | You can (and likely will, if it doesnt bear out for what you're trying to find) use that excuse nomatter what TV ranges appear since the data will always be "not what the game was designed around" owing to the TV matching system. The TV ranges will always cluster around even, resulting in remarkably little data for high ranges. |
Accusations of intellectual dishonesty aside, so long as there is enough to at least bring down the high margin to below 50% then I can't really argue, can I?
| Quote: | | The extrapolation is based on the only data that exists. It makes it a more valid extrapolation than any other that might be created based on... what? |
All the data R, B and L (and perhaps what we can get from Cyanide). Then we need enough data to get better margins of error. I'm willing to wait for that.
| Quote: | | So tell me.. what precisely are you seeking to investigate |
Whether or not the design objectives have been met. One of those is that some teams do better with more inducements than others at certain TVs and with certain opponents. There are multiple instances of this, so when there is more data I'll take a look at various relationships. That's pretty woolly enough for you, I'm sure. The intent is to look at as many of the relationships as I can.
| Quote: | | Or hey, we can repeat this back and forth another twenty or thirty times, saying exactly the same thing. |
Looks like this was a valid extrapolation. Feel free to have the last word  |
|
|
VoodooMike
Joined: Nov 07, 2010
|
| Posted:
Dec 13, 2011 - 22:54 |
|
| dode74 wrote: | | Accusations of intellectual dishonesty aside, so long as there is enough to at least bring down the high margin to below 50% then I can't really argue, can I? |
Hasn't stopped you thus far, has it? Do keep in mind that you're talking about the data when aggregated to a single TV. You can, in fact, combine multiple TV categories, which will narrow the margins, or not aggregate them, if looking for significance rather than strength in terms people can relate to.
| dode74 wrote: | | All the data R, B and L (and perhaps what we can get from Cyanide). Then we need enough data to get better margins of error. I'm willing to wait for that. |
And once again, my extrapolation is based on all the data that exists. Any contrary position is based on no data we know of (supposed design goals are not data). That doesn't mean my extrapolation is guaranteed correct, but it means it is the only one supported in any way by actual, verifiable data.
To say that a statement is true until proven, beyond all doubt, to be false... is religion.
| dode74 wrote: | | Whether or not the design objectives have been met. One of those is that some teams do better with more inducements than others at certain TVs and with certain opponents. There are multiple instances of this, so when there is more data I'll take a look at various relationships. That's pretty woolly enough for you, I'm sure. The intent is to look at as many of the relationships as I can. |
And again, without specific relationships to investigate, you aren't actually investigating. Some teams - which teams? some inducements - which inducements? certain TVs - which TVs? If there are no solid answers to those questions then you can't test them... all you're doing is sitting there waiting for anything to happen so you can then say "yup, I think that qualifies!" and until that happens saying "We just don't have enough data yet"... it's the gypsy curse.
| dode74 wrote: | | Looks like this was a valid extrapolation. Feel free to have the last word |
I did, when you stormed off back on page 8. |
|
|
Beerox
Joined: Feb 14, 2008
|
| Posted:
Dec 13, 2011 - 23:54 |
|
| JimmyFantastic wrote: | | hahaha dode is never gonna give up, I love it! |
He does this for a living I think |
|
|
Sutherlands
Joined: Aug 01, 2009
|
| Posted:
Dec 14, 2011 - 00:40 |
|
| VoodooMike wrote: |
It's the creationist's argument again - have we looked everywhere to see if God is there? No? Then we can't say he doesn't. |
people make that argument to you? Hm, either you're being disingenuous or you hang out with some really stupid people. (like seeks like?  ) |
|
|
VoodooMike
Joined: Nov 07, 2010
|
| Posted:
Dec 14, 2011 - 00:52 |
|
| Sutherlands wrote: | | people make that argument to you? Hm, either you're being disingenuous or you hang out with some really stupid people. (like seeks like? Wink ) |
Either you are being disingenuous or you're somehow unaware that one need not directly experience something to know about it. Did you ever meet Stalin or Gandhi? Were you aware they existed, and have some idea of what makes them stand out?
Also, careful about jumping on your own grenades. In your "like seeks like" implication that I'm stupid, you missed the fact that you came and posted that to me of your own free will.. |
|
|
VoodooMike
Joined: Nov 07, 2010
|
| Posted:
Dec 14, 2011 - 00:58 |
|
| koadah wrote: | | Cool. So what is it telling us about the difference between getting 150 and 200 in inducements? |
Well, two things. First, we're moving back into the realm of the other thread topic, so we may wish to relocate any real discussion of TD difference/UD victory chance to that. Second, I'm not going to guess based on eyeballing the numbers - there are specific statistical techniques to be used to get an idea of the relationship between different variables, most of which we can't use based solely on your chart.
Trying to talk about relationships and trends based solely on looking at the numbers results in this kind of mistake, so its better to do things right(er - there's no absolute certainties in stats, only probabilities). |
|
|
|
|
| |
Designer's Comm...
Search
