Sciencing V:tES

1muflon1 wrote: ...science...

The only thing that can be said about science with sertainty is that 99% of it's theories will be proven wrong in some time.

If we are talking about econometrics, than, I think, we can rise this estimate to 99,9% - ease of instruments has led to videspread "analisis" of white noize.

1muflon1 wrote: You dont need to take control of all factors, just the main variables you are interested in and a bunch of important controls.

This one is a main offender - if you are leaving some factors outside the model than you need to be sure that they are not important, and how you can be sure if you are not analizing them?

1muflon1 wrote: It is then better to do balancing analysis on basis ofanecdotes and hunches?

Yea, it'll be better - than you'll know that it's analisis based on anecdotes and will make desitions appropriately.

...

My point is not that the VtES should not be analized, just that you need to chose reliable sample of data to do it.

Even the opposite - if somebody provides data about several last EC d2 meta and archons, I'll be happy to join the effort to dig them for any meaningful information about game balance.

Although as a doctoral soc-sci student I also have to admit some scepticism as to controlling all of the significant variables in VTES (given the amount of 'luck' sometimes in drawing the right card at exactly the right moment), I am intrigued by this proposal. I don't know what exact variables you are looking for, but I don't think any existing data would be available containing them all: better to design your data collection instruments and methods from scratch.

Perhaps it would be possible to run a series of experiments with several players using LackeyCCG , assuming it logs the full game? Using the pre-constructed decks in the earlier releases for each player would keep things simple. Or alternatively, if you can tweak the rules, make all unique cards non-unique between players and allow people to play the same deck. That's a lot of time and effort to collect and crunch the data though....

chrisn101 wrote: Perhaps it would be possible to run a series of experiments with several players using LackeyCCG , assuming it logs the full game? Using the pre-constructed decks in the earlier releases for each player would keep things simple.

Yes, I was thinking about something like this, the only problem is that it would take a lot of time. The less reliable, but still interesting thing would be to analyze achrons from EC. I was thinking some simple logit probit model to see how several variables affect your chances of winning a match. I was thinking making dummy winning the match as dependent variable, and then on RHS to include dummies like camarilla, clan, or variables like the number of a certain card in a deck, or number of vamps with certain disciplines, but thats just first idea I am oppen to suggestions. True there would be still external validity issues, but it would still provide at least some starting point for further discussion.

Also to elotar. If your argument is only that I should use only EC d2 data, I dont mind starting there, I just dont see a reason why not look also at causal games.

Great idea. I'd love to see the results

Another 'proxy sample' could be to set up, and then analyse, a series of 'online tournament' games (thereby making the data collection relatively easier for all players at the table).

Kevin Mergen used to be involved in some online tournaments work.
LackeyCCG could have some organised to do similar.

I'd be surprised if the minimum baseline you suggested of 30 games, couldn't be gotten in 3 months from both systems.

It will also help 'randomise out' the player in respect of the cards, since there would potentially be wider combinations of experienced players and inexperienced players playing 'optimal' and 'non-optimal' decks. (i.e. The sample may get to the point where there is enough variation between players using optimal/sub-optimal/poor decks to 'average out the player' in the statistical method.)

While I am not against someone making such analysis, I am not fond of the idea of using it as a basis of game design decisions. The data it produces is only as good as the person interpreting it.