NTBB: Stats

Got some ideas for rules? Maybe a skill change or something completely different!!! Tell us here.

Moderator: TFF Mods

Post Reply
User avatar
VoodooMike
Emerging Star
Emerging Star
Posts: 434
Joined: Thu Oct 07, 2010 8:03 am

Re: NTBB: Stats

Post by VoodooMike »

koadah wrote:So Dode, where are all the statistics and analysis that the BBRC used to decide that their job was done?
A question I've asked many times, though I'm not sure why you're asking Dode - unless he was a secret member of the BBRC? Tell me, Koadah... where is Jimmy Hoffa's body buried?
koadah wrote:Did they satisfy VMs rigorous standards?
My only "rigorous standard" is that there be a legitimate statistical basis for changes if you're claiming that there is one. You're not too stupid to have read that in one of the 10 posts I've pointed out that I don't oppose houserules, just things that claim to be stats based when they're not... you're just, as usual, being deliberately obtuse.
koadah wrote:Unless you are going to prove that they are already within Plasmoid's tiers then yes, I'm for him to call it narrow tiers.
He doesn't have to prove they are - plasmoid would have to prove that they are NOT. The base assumption shouldn't be that things are broken, but that things aren't broken, and the analysis is what you use to show that things are likely not as they should be. Luckily, a hundred years of brilliant mathematicians have spent their lives developing methods to do it if you can be bothered to learn how - if you can't, then don't be surprised when the legitimacy of the stuff you DO bother to do, is questioned.
koadah wrote:So far your stats don't seem to be any more real than our witchcraft.
"Seem" meaning... you don't understand them, so they must not be real. Your logic is, as always, impeccable.
Darkson wrote:You want to break the World down into statistics and data, go ahead. You already killed one discussion with your demands that people quantify their feelings with cold hard facts.
Yeah, man, why are you letting the truth get in the way of a good story? Darkson has it right - if he doesn't understand it, and doesn't like it, then it has no place in his world and just serves to muddy the waters. Can those thick books stop a punch, college boy???
Darkson wrote:I'll be sitting in the corner playing a game with people who "feel" something., and don't need reams of data to suggest changes that work for them.
I bet neither you nor those people would be in a big hurry to use a medical system built on "feel", or drive cars designed by "feel", or live in houses that were built on "feeling-based" principles, etc. Your entire life is data-driven, and just because statistics and data analysis confuse you does not mean that they're some hard and confusing thing that should be reserved for what you perceive to be "critical" portions of life.
dode74 wrote:Either do the maths or remove the implication. I believe plasmoid is doing the maths.
I don't. He stated, outright, that he couldn't be bothered to even look up how to calculate CIs, much less learn statistical analysis of any relevant sort. That doesn't suggest he's going to suddenly decide to learn and then actually do the work.
koadah wrote:It might imply that to a handful of stats nuts but I suspect most of the average punters think he'll just eyeball some results data pretty much as we would.
Most people will just eyeball numbers and draw conclusions from them... because they don't know how to do otherwise. Those same people will, however, tend to side with statistical analysis if they're offered the results of that, and just someone saying "well, I think..." when they don't already have a well-established opinion themselves... and as well they should. The dangerous thing is when people misrepresent their simple opinions as statistics.. or just throw out meaningless numbers to convince undecided people that those mean something they don't, hence the adage "there are lies, damned lies, and statistics". Numbers don't lie, but people can use numbers to tell lies, as is the case with plasmoid's unqualified means on the first page. I don't think he's deliberately lying with them (though some of his statements made in this and the other thread do suggest he'd be willing to misrepresent data if it would support his ideas) but the numbers are being used to misrepresent the truth regardless.
koadah wrote:Gut feel, witchcraft & hand waving FTW
So long as the W stands for "worthless" or "wrong", sure.
Pluisje wrote:Ok, I have some numbers for Amazons only. I took the 250.000+ games in the file kindly supplied by Koadah...
Uhm... before you go any further.. didn't those files include games going back as far as 2005? You realize that FUMBBL only switched to CRP in early 2012, right? Also, I'm pretty sure box, ranked, and league didn't switch to CRP at the same time, so you'd need to filter down that data to only include games that were actually played under the rules that we're discussing.
Pluisje wrote:So it looks that Amazons don't come above the 55% in Box and Ranked, statistically speaking.
Across all TV levels they don't, no. The issue (some people feel exists) with Amazons is that they start overly strong and then get weaker with TV.. it averages out to be closer to 50%, but this is the reason I made that hyperbolic example of a team that has a 100% win rate for half the TV ranges played, and 0% win rate for the other half. Is that team "balanced"? It is by the originally stated criteria, which is why I question that criteria's utility.

Reason: ''
Image
plasmoid
Legend
Legend
Posts: 5334
Joined: Sun May 05, 2002 8:55 am
Location: Copenhagen
Contact:

Re: NTBB: Stats

Post by plasmoid »

Ach, I really wanted to reply tonight.
But I've been grading papers till 2 on the morning for several days now, and the kids still get up at 7.
I think I'll be able to squeeze in my reply tomorrow. Just need to get that pesky birthday party over with first :orc:

Mike - as for the BBRC data, I believe the answer has been given in this very thread. More on that tomorrow.

As for amazons. I think I've got the smoking gun on those. Girls are going down... (and not in a dirty way :oops:)

Cheers
Martin

Reason: ''
Narrow Tier BB? http://www.plasmoids.dk/bbowl/NTBB.htm
Or just visit http://www.plasmoids.dk instead
User avatar
Darkson
Da Spammer
Posts: 24047
Joined: Mon Aug 12, 2002 9:04 pm
Location: The frozen ruins of Felstad
Contact:

Re: NTBB: Stats

Post by Darkson »

VoodooMike wrote:
Darkson wrote:I'll be sitting in the corner playing a game with people who "feel" something., and don't need reams of data to suggest changes that work for them.
I bet neither you nor those people would be in a big hurry to use a medical system built on "feel", or drive cars designed by "feel", or live in houses that were built on "feeling-based" principles, etc. Your entire life is data-driven, and just because statistics and data analysis confuse you does not mean that they're some hard and confusing thing that should be reserved for what you perceive to be "critical" portions of life.
Oh sure, cos a game is as important as any of those things. :roll:

I game for fun, you troll for fun. I couldn't give a damn if the maths are "right", if it's not fun because it doesn't feel right, then I change it. I assume you feel the same way about your trolling?

Reason: ''
Currently an ex-Blood Bowl coach, most likely to be found dying to Armoured Skeletons in the frozen ruins of Felstad, or bleeding into the arena sands of Rome or burning rubber for Mars' entertainment.
User avatar
Pluisje
Rookie
Rookie
Posts: 24
Joined: Fri Oct 09, 2009 2:08 pm

Re: NTBB: Stats

Post by Pluisje »

VoodooMike wrote:Uhm... before you go any further.. didn't those files include games going back as far as 2005? You realize that FUMBBL only switched to CRP in early 2012, right? Also, I'm pretty sure box, ranked, and league didn't switch to CRP at the same time, so you'd need to filter down that data to only include games that were actually played under the rules that we're discussing.
Thanks, I didn't know the exact date. The matches date from 16 jan 2011 to 10 feb 2013. I did notice a (not jet calculated) rise in the win% in the matches from the beginning of 2012. When I've got the dates of the CRP switch, I'll recalculate.
VoodooMike wrote: Across all TV levels they don't, no. The issue (some people feel exists) with Amazons is that they start overly strong and then get weaker with TV.. it averages out to be closer to 50%, but this is the reason I made that hyperbolic example of a team that has a 100% win rate for half the TV ranges played, and 0% win rate for the other half. Is that team "balanced"? It is by the originally stated criteria, which is why I question that criteria's utility.
I'll do two more calculations, sorted by TV and (for Plasmoid) number of matches played.


To show how strong the demands for significance are, I calculated the results of the 1,500 matches of Amazons vs Dwarf and Chaos Dwarves. Average was 0,4101, Sd 0,0365, Av+2*Sd=0,4831. Not significantly outside the 45-55%. So there is no evidence Amazons lose more against Dwarves :roll:.

Reason: ''
koadah
Emerging Star
Emerging Star
Posts: 335
Joined: Fri Mar 25, 2005 5:26 pm
Location: London, UK

Re: NTBB: Stats

Post by koadah »

Pluisje wrote:
VoodooMike wrote:Uhm... before you go any further.. didn't those files include games going back as far as 2005? You realize that FUMBBL only switched to CRP in early 2012, right? Also, I'm pretty sure box, ranked, and league didn't switch to CRP at the same time, so you'd need to filter down that data to only include games that were actually played under the rules that we're discussing.
Thanks, I didn't know the exact date. The matches date from 16 jan 2011 to 10 feb 2013. I did notice a (not jet calculated) rise in the win% in the matches from the beginning of 2012. When I've got the dates of the CRP switch, I'll recalculate.
Switch over dates I have are:
16/01/2011 Box
12/02/2011 Ranked
12/05/2011 League

Edit: check with Christer to be sure.

As mentioned earlier in the thread League uses all manner of house rules. possibly odd formats e.g. 1v1, 4v4, 7v7. Box & Ranked tend to be mainly TV matched.

It's great to hear that "there is no evidence Amazons lose more against Dwarves".

Reason: ''
dode74
Ex-Cyanide/Focus toadie
Posts: 2565
Joined: Fri Jul 24, 2009 4:55 pm
Location: Near Reading, UK

Re: NTBB: Stats

Post by dode74 »

To show how strong the demands for significance are, I calculated the results of the 1,500 matches of Amazons vs Dwarf and Chaos Dwarves. Average was 0,4101, Sd 0,0365, Av+2*Sd=0,4831. Not significantly outside the 45-55%. So there is no evidence Amazons lose more against Dwarves .
Really? On the previous page you gave the results for Zons vs everyone. Taking your 200-200 figures:

Code: Select all

             Box      Ranked
Mean         0.59140  0.57938
Mean - SD*2  0.50433  0.52426
You "Zons vs Dwarf types" figures were a mean of 0.4101 and a mean + SD * 2 of 0.4831.

Given that the top end of the "Zons vs Dwarf types" range is lower than the bottom end of the "Zons vs everyone" range then we can say that there is evidence that Zons lose more to Dwarf types than to other teams.

Even this evidence is a little flawed because for a proper comparison you would need to remove dwarven teams from the overall data so you were comparing "Zons vs Dwarven teams" with "Zons vs non-Dwarven teams". Such a change would increase the gap between the means, although whether the gap between the ranges would be increased would depend on the change to the SD - I suspect that the gap between the ranges would also increase.

Reason: ''
User avatar
Pluisje
Rookie
Rookie
Posts: 24
Joined: Fri Oct 09, 2009 2:08 pm

Re: NTBB: Stats

Post by Pluisje »

dode74 wrote:Really? On the previous page you gave the results for Zons vs everyone. Taking your 200-200 figures:

Code: Select all

             Box      Ranked
Mean         0.59140  0.57938
Mean - SD*2  0.50433  0.52426
Your "Zons vs Dwarf types" figures were a mean of 0.4101 and a mean + SD * 2 of 0.4831.

Given that the top end of the "Zons vs Dwarf types" range is lower than the bottom end of the "Zons vs everyone" range then we can say that there is evidence that Zons lose more to Dwarf types than to other teams.

Even this evidence is a little flawed because for a proper comparison you would need to remove dwarven teams from the overall data so you were comparing "Zons vs Dwarven teams" with "Zons vs non-Dwarven teams". Such a change would increase the gap between the means, although whether the gap between the ranges would be increased would depend on the change to the SD - I suspect that the gap between the ranges would also increase.
Oops, my bad. Correcting the flaw (Box only): mean (no Dw/CD) 0,6268, SD 0,0524, mean-2*SD= 0,5219. So as you said, there is a significant difference in Win%, but as I said, neither falls outside the 45-55%. I was a bit too quick to call this "not losing more against Dwarves". A better statement would be "the win% against Dwarves is not a ground to change the Amazons, as it does not fall outside the 45-55%.".

Reason: ''
dode74
Ex-Cyanide/Focus toadie
Posts: 2565
Joined: Fri Jul 24, 2009 4:55 pm
Location: Near Reading, UK

Re: NTBB: Stats

Post by dode74 »

Agreed :D

Reason: ''
User avatar
Lunchab1es
Star Player
Star Player
Posts: 613
Joined: Tue Jun 01, 2010 7:21 pm
Location: Georgia, USA

Re: NTBB: Stats

Post by Lunchab1es »

All bickering aside, I am actually enjoying reading the discussions from the sidelines on the statistics being thrown around.

If there ever was a thorough, meaningful statistical analysis on BB (taking match ups, TV levels etc into account) it's the kind of thing I would love to read through.

Being a non-math major, I'll leave it to y'all to hash out the details of BB Statistical Mission: Impossible. Let me know when you're done. :P

Reason: ''
Looking for: 5th ed Human Thrower #2
User avatar
garion
Ex-Mega Star, now just a Super Star
Ex-Mega Star, now just a Super Star
Posts: 1687
Joined: Mon Jun 07, 2010 12:59 pm

Re: NTBB: Stats

Post by garion »

Darkson wrote:I didn't "miss" the second point, I don't care about the second point.
You want to break the World down into statistics and data, go ahead. You already killed one discussion with your demands that people quantify their feelings with cold hard facts.
I'll be sitting in the corner playing a game with people who "feel" something., and don't need reams of data to suggest changes that work for them.
:lol:

Never a truer word spoken!

Oh and Voodoo mike you clearly missed the whole point of my post. I have written some rules for me and some people that will play them, that is all. Plasmoid has written some rules for him and a group of people to play them. Darkson has done the same. This statistical analysis is all pointless, it will not change how any of the 3 people who have created house rules approaches their changes. I makle my changes based on what I find fun. Darkson too. Plasmoid is being a little more restrained etc... but untimately its still just his personal preference. If Plasmoid thinks Orcs at low TV need a nerf then he is free to do so, if people like the ideas they may give them a go. Suck it and see, if people think it the changes make the game more FUN then great. Now go find stats for what different groups of people find fun.

FYI I do understand statistics, they are just clearly bollocks in a game with as many variables, especially when one of the most important of which is coaching ability and others include house ruled leagues to draw conclusion from and tiny sample sizes from other sources.

But meh, fill your boots. Lets see how many pages we can fill with Lies, Damned Lies and Statistics.

Reason: ''
dode74
Ex-Cyanide/Focus toadie
Posts: 2565
Joined: Fri Jul 24, 2009 4:55 pm
Location: Near Reading, UK

Re: NTBB: Stats

Post by dode74 »

but untimately [sic] its still just his personal preference. If Plasmoid thinks Orcs at low TV need a nerf then he is free to do so, if people like the ideas they may give them a go.
Absolutely no issue with that. Just don't claim to have statsitical backing for it if it's not there. Which it isn't.

Is that so hard a concept?
FYI I do understand statistics
Really? Forgive me for not believing you but I'm going to put that one down as a "lie". Possibly a "damned lie".

Reason: ''
User avatar
garion
Ex-Mega Star, now just a Super Star
Ex-Mega Star, now just a Super Star
Posts: 1687
Joined: Mon Jun 07, 2010 12:59 pm

Re: NTBB: Stats

Post by garion »

dode74 wrote:
but untimately [sic] its still just his personal preference. If Plasmoid thinks Orcs at low TV need a nerf then he is free to do so, if people like the ideas they may give them a go.
Absolutely no issue with that. Just don't claim to have statsitical backing for it if it's not there. Which it isn't.

Is that so hard a concept?
I agree with you, its not there, clearly not. So fair point, he shouldn't claim that, if he even is? He is using the method of suck it and see. So again.... whats the point in this thread?

Reason: ''
dode74
Ex-Cyanide/Focus toadie
Posts: 2565
Joined: Fri Jul 24, 2009 4:55 pm
Location: Near Reading, UK

Re: NTBB: Stats

Post by dode74 »

I believe Martin started it to talk about the statistical basis on which NTBB is "based" and which was questioned by several of us. That's what we're discussing.

He can do what he wants with his rules, of course. I don't think anyone who has questioned the statistics has said otherwise. The changes made just can't be justified on a statistical basis, is all, so claiming that they are gives them a false provenance.

Reason: ''
User avatar
VoodooMike
Emerging Star
Emerging Star
Posts: 434
Joined: Thu Oct 07, 2010 8:03 am

Re: NTBB: Stats

Post by VoodooMike »

plasmoid wrote:Mike - as for the BBRC data, I believe the answer has been given in this very thread. More on that tomorrow.
If so, then it suggests that the BBRC was working from low-meaning data, which would certainly explain why their resulting decisions created a ruleset you, and others, feel the need to "fix", though, again, I point out that using the same methods they did is likely to result in something just as flawed.
Darkson wrote:Oh sure, cos a game is as important as any of those things.
Not as important, no, but since using proper statistical analysis is not significantly tougher than just making shit up, there's no reason not to use those same methods. Your repeated objection to that fact suggests that you think its like mountain climbing, or tiger dentistry, which is really just a result of you not understanding it.

Summary - you're taking the standpoint that anything you don't understand is obviously unimportant. That takes it from ignorance to beligerant ignorance.
Darkson wrote:I game for fun, you troll for fun. I couldn't give a damn if the maths are "right", if it's not fun because it doesn't feel right, then I change it. I assume you feel the same way about your trolling?
My postings are straight on topic, unlike your typical contributions to these threads, or any on Cyanide's forums, etc. Trolling is just making comments to stir up ill-will and get angry responses and/or taking opposing viewpoints to someone not because you hold those opposing beliefs but because you want to fight with them. The only person who is "trolling" in this thread is Koadah.
Pluisje wrote:To show how strong the demands for significance are, I calculated the results of the 1,500 matches of Amazons vs Dwarf and Chaos Dwarves. Average was 0,4101, Sd 0,0365, Av+2*Sd=0,4831. Not significantly outside the 45-55%. So there is no evidence Amazons lose more against Dwarves .
If you're using a computer to do your calculations, you may want to go past "SD x 2" for your CI. We use that in casual conversation because nobody is going to bust out a calculator or charts to give the exact numbers... but you can get considerably more accurate.

Also, you don't have evidence that "Amazons don't lose more against Dwarves". First off, the numbers you gave are, in fact, significantly outside the range... the CI is the specific measure of significance, you don't find the upper limit and then eyeball whether its "close enough" to something else. What your numbers suggest is that the win% versus dwarves falls OUTSIDE the expected T1 range with 96% confidence.

This also doesn't mean "Amazons lose 'more' against dwarves", that sort'v thing would require an ANOVA and follow-up tests. It just tells us that in whatever dataset you were using, the amazon win% versus dwarves falls under 45%, and we're 96% certain that can be generalized to the population the sample represents.

You should get yourself a copy of PSPP (an open-source, though less powerful clone of SPSS) and work with that... or get a copy of SPSS by whatever means you're comfortable with. Certainly SPSS will do all the necessary calculations for you, and PSPP does quite a few of them... I've only used it a little bit, though.
Lunchab1es wrote:If there ever was a thorough, meaningful statistical analysis on BB (taking match ups, TV levels etc into account) it's the kind of thing I would love to read through.
There has been some analysis done on the fairly decent-sized datasets from FUMBBL and FOL, across the past 2-3 years. Analysis is pretty specific, though, so... only specific questions have been looked at.
garion wrote:This statistical analysis is all pointless, it will not change how any of the 3 people who have created house rules approaches their changes.
No, it is you who continues to miss the point. You're not required to engage in statistical analysis to make your houserules. You will, however, be called to task on your lack of analysis if you try to claim that your houserules are data driven, and serve to address issues in specifically statistical imbalances in the game, as "NTBB" has.
garion wrote:FYI I do understand statistics, they are just clearly bollocks in a game with as many variables, especially when one of the most important of which is coaching ability and others include house ruled leagues to draw conclusion from and tiny sample sizes from other sources.
What a crock 'o crap. You and Darkson should found a "Who needs math when we have the Bible?" group. You can burn calculators and follow your hearts to the promised land.

Again, these are mathematical tools used to evaluate life or death issues, from drug effects to disaster planning to epidemic mapping and so on. But no, I'm sure this board game is outside its scope because it has WAY more variables and impossible to measure influences.

Reason: ''
Image
User avatar
Pluisje
Rookie
Rookie
Posts: 24
Joined: Fri Oct 09, 2009 2:08 pm

Re: NTBB: Stats

Post by Pluisje »

VoodooMike wrote:You should get yourself a copy of PSPP (an open-source, though less powerful clone of SPSS) and work with that... or get a copy of SPSS by whatever means you're comfortable with. Certainly SPSS will do all the necessary calculations for you, and PSPP does quite a few of them... I've only used it a little bit, though.
I tried PSPP, but I couldn't get it to work. Turns out I forgot to install it with admin rights :oops:. It should work now. Will try some calculations the coming week.

Reason: ''
Post Reply