I would think the best way to determine ranking wold be to use the formula

((n+1)-p) / (n+1)

n = number of participants
p = position in tournament

You could then set the solution equal to zero if the unit wasn't used in a particular force.

Alternatively, in order to weight based on number of formations taken as well, you could simply use the formula

q * ((n+1)-p) / (n+1)

q = quantity of formation taken

My worry is that this would weight the scales on a) favorite units of perenially high ranking players, and b) 'fad' units, which are seen as the best units, but may not be. Different data sets would be required for every army list and ideally every revision of each list. This would likely return statistically insignificant results in all but the most stable lists i.e. those that don't need changes. It also papers over the combined arms factor; taking an army entirely of the best unit is not as good as a balanced army list.

_________________
SG

Ghost's Paint Blog, where everything goes that isn't something else.

This is an issue that requires the touch of an artist, not the number-cruncing of a computer.

GI-GO.

_________________
Currently doing a plastic scenery kickstarter

I thinks we gots some smart folk talk here...

OK, so who's the engineer and who's the physicist? Sounds a lot like my offiice!

Wow. This has the potential to be worse than my WhES

_________________
We are returned!
http://www.epic-wargaming.de/

Big bang theory quote:
Dr Gablehauser: Hello boys!
Raj: Dr Gablehauser!
Dr Gablehauser: Dr Koothrapali!
Leonard: Dr Gablehauser!
Dr Gablehauser: Dr Hofstadter!
Sheldon: Dr Gablehauser!
Dr Gablehauser: Dr Cooper!
Howard: Dr Gablehauser!
Dr Gablehauser: *Mr* Wolowitz!
Howard: I have a Master's.
Dr Gablehauser: Who doesn't?

Could have potential if someone can work out a way to do it and put the time in.

I agree that army lists are as much art as science, but people in discussions here often do refer to how units are used in tournaments and accurate statistics per unit could make such assessments more valid as well as potentially revealing subtler patterns we've not picked up.

Good luck to who ever decides to do this, It's not going to be me though!

Isn't this going to generate goofy stats like Rhinos being the best unit ever?
I'd say go for it. Until the numbers are crunched, no whacky fundamental changes to the list.

_________________
author of Syncing Forward and other stories...

It's a dog-eat-dog world, and I've got my Milkbone underwear on.

Breaking down the years stats into games we get the following.

There have been 149 games involving marine lists that can take warhounds. Of these they break down into 52 wins, 34 draws and 63 loses.

Of the wins T'bolts were in 45 of the games, T'Hawks in 49 and Warhounds in 40.
Of the loses T'bolts were in 50 of the games, T'Hawks in 58 and Warhounds in 24.

And of the 48 games played with steel legion no one played without T'bolts.

Only Warhounds are statistically significant I'd say.

_________________
Currently doing a plastic scenery kickstarter

The problem is, there's too many reasons why a formation is taken, coupled with a truly limited number (comparatively) of games in the sample pool. You can get trends, perhaps, but there's a lot of other issues that preclude a purely statistical approach from working.

Off the top of my head, these are some of the reasons a formation may be fielded.
- The formation is over-effective because of it's cost.
- The formation is required as a limited/singular Core Formation. (Black Legion Retinue, Necron Warriors)
- The formation is iconic or liked for fluffy reasons, and the player feels it should be there.
- The formation is painted (happens locally a fair amount. "Yeah, I know they're crap, but I already painted them.")
- The formation accomplishes something that nothing else in the army does. (AA, or ranged TK, or long BP are some examples, if the army has severely limited options).
- The local metagame forces certain options (aircraft tend to be the biggest variance here, some groups seem to play limited amounts, and so AA isn't as important. Others all but require 500+ pts of counter/defensive AA).
- Inexperienced players also tend to skew data. Even if they take an optimal build, their inability to use it to the fullest can effect the numbers. Warhounds that expose themselves to Shadowsword batteries, Terminators placed sub-optimally, etc. And even experienced players playing armies outside their comfort zone.

If done on a unit base, as Moscovian says, Rhinos, Thunderbolts and Warhounds will be up there, due to appearing multiple times, or in multiple armies.

Unless you can filter out all but the first bullet point, any raw statistical analysis is going to be skewed heavily into the "feelings" based area at least as much as the current system. I have faith in numbers, even when it screws me. So much so, it tends to drive FrogBear a little nutty at times. But there's a reason "Lies, damned lies, and statistics" is a common usage phrase. Substitution of poor data, for thought, can be just as bad, if not worse, than no thought at all.

Morgan Vening

What's to suggest the 'irrational' biases are unevenly distributed over tournament-position?

E.g.


Are better players more/less likely to do this than worse players? If not, I'm thinking it won't affect the 'utility/effectiveness' measurement.

_________________
Uti possidetis, ita possideatis.
May your beer be laid under an enchantment of surpassing excellence for seven years!
An online epic force creator: Armyforge

I know I play certain unit types because I like how the paint job came out or -in the case of Banshees- because they're chicks.

_________________
author of Syncing Forward and other stories...

It's a dog-eat-dog world, and I've got my Milkbone underwear on.