Talk Fantasy Football

straume wrote:1) It is rude. I am probably a bit to tender for the Interwebs, but I just don`t get how people assume it is okay to act like a jackass on a forum.

It's a GIFT.

Vanguard wrote:

straume wrote:2) Any sound argument which might be there is diminished/drowned because of 1.

This.

That is how I felt. I stopped reading VM posts because of it.

Hi Dode,
just so we're absolutely clear, which data are you looking for?
Straight PCRP+ data or NTBB (which includes PCRP+)?
Cheers
Martin

So long as it's clear which is which, all of it.

legowarrior wrote:

Vanguard wrote:

straume wrote:2) Any sound argument which might be there is diminished/drowned because of 1.

This.

That is how I felt. I stopped reading VM posts because of it.

I wonder how many people read VoodooMike's posts first.

So, Dode, what is your null hypothesis before this all gets started. What types of statistical tests will you be running. You are asking for a lot of data from Plasmoid, so the least you can do is lay out the methodology before he gives it to you. Otherwise you might nust be wasting all our time.

Also, piggybacking on all thats happened so far, what is the end goal of all this? I mean, it has been stated over and over and over again that blood bowl isnt balanced and that some teams are better than others (and that trying to balance them all will just lead them to becoming cookie cutter teams, like in starcraft or magic the gathering). So what are trying to prove? That the changes dont make a difference or that they do make a difference?

Are you going to take TV values into account? Inquiring minds want to know.

Also, I am curious about the statistical package you are going to use. I use to have access to SPSS, but since I no longer work on the statistics of my thesis, I have lost access to it.

Plasmoids CRP+’ changes to the CRP deal mainly with the current slant towards bash, in order to make other strategies and roster types more viable in a mid- to long term setting. Plasmoids CRP+ also introduce changes to the Human roster, which seems to struggle near the low end of tier 1.

That's from the NTBB website reference CRP+ specifically. It appears to have the aim of:
1. Making non-bash teams "more viable" in a mid to long term setting. Plasmoid would need to define which teams he means, what "more viable" means (win%?) and what mid to long term is. Once that is done the null is that this can not be said to have happened.
2. Improving Humans, presumably in terms of win%. The null would be that the human win% cannot be said to have improved.
The former requires clarification from plasmoid until the test type can be decided, but the latter can be a look at win percentages.

So what are trying to prove? That the changes dont make a difference or that they do make a difference?

I'm not the one trying to prove anything. Plasmoid has stated his aims and until we can say they have been met they cannot be said to work. That's not the same thing as saying they don't work.

Are you going to take TV values into account? Inquiring minds want to know.

NTBB appears to be the part which looks at TV ranges rather than CRP+. To that end it doesn't seem that TV values should be taken into account.

I am curious about the statistical package you are going to use.

PSPP is a good "version" of SPSS, but I will likely use Excel as it has some excellent modules available as well as the built-in stuff.

Thanks. I'm curious about what you find out, and the methods you use.

How many games does he have? How many teams play to 100 games?

For most standard statistics, 30 per category should be enough. The smaller the size of the change, the greater the sample size that is required.

I would test it with chi squared. Compare the win rate of single team at a tv range (probable break out the individual races played) and just a before and after test. So 2 ( with CRP+ and without) x 24 (each of the races).

legowarrior wrote:For most standard statistics, 30 per category should be enough. The smaller the size of the change, the greater the sample size that is required.

I would test it with chi squared. Compare the win rate of single team at a tv range (probable break out the individual races played) and just a before and after test. So 2 ( with CRP+ and without) x 24 (each of the races).

You're talking to the 'willfully ignorant' here.

Is that 2 x 30 x 24 x number of TV ranges?

Don't we also want to know how teams do over the course of their lives. e.g. at least 50 games?

Do we really care what they will do over careers? There is no balance to the teams or to the skills at all, so just look at have the win ratios look at a particular tv point. Do the new win ratios met the criteria? KISS

legowarrior wrote:Do we really care what they will do over careers? There is no balance to the teams or to the skills at all, so just look at have the win ratios look at a particular tv point.

Look, the game lack any semblance of balance as it stands today so I am not sure what statistical analysis you want to use. Just pick a tv value range and compare the win ratios and see what plasmoids changes did. Than pick another range and do it again. Comparing matches with similar tv ranges only to minimize the impact of inducement and use enough games that you have a good range and do it both games with better rule set and games with the archaic rule set.

Everything else has so many variables that you might as well ignore it and hop that you have a good set of data to balance it all for you.

Step 3 is profit.