file Sciencing V:tES

03 Mar 2016 22:18 - 03 Mar 2016 22:40 #75706 by 1muflon1
Sciencing V:tES was created by 1muflon1
The exciting possibility of relaunching v:tes started a very interesting discussion on the balancing issues of dominate (look at this excellent blogpostblogpost ), and certain clans or archetypes. However, since I am by profession econometrician I always get a bit skeptical when someone jumps from some simple statistical patterns to causality (not because I think that these ideas are nonsense. I actually think that some of those observations are right, but in my professional life I already saw so many anecdotes and common wisdom disproved that I always get paranoid).

For example, brettscho sparked interesting discussion on the prevalence of dominate in tournaments. But the argument that this is because dominate is OP discipline is just one hypothesis and there are other possibilities. Maybe this is because dominate is the main bleeding discipline and since bleeding is the main ousting mechanism, it is this what drives the prevalence of dominate.

But luckily such hypotheses can be tested against each other. For example, If adding bleeding deck dummy to econometric model would be significant, and dummy for having a deck with majority of dominate vamps suddenly loses significance, it would be good evidence that it might be actually bleed itself which is being overpowered, and dominate just shows because its bleeding discipline or otherwise it will simply show that brettscho was right.

There are also many other interesting things that can be mined out of the data. For instance, If the database would include the length of play, I could build a model which would show which decks or cards are most likely to be slowing down the play.

I could also build some probit or logit models to determine how much at margin certain cards increase your chance of winning. Or the same can be done for individual clans or groups like camarila or just archetypes.

Although the best way how to settle the issue would be to do about dozen double-blind randomized experiments, and then do a meta-analysis of them, this would be probably too costly and time-consuming, so really any panel data would do, although the results would be less strong and subject to some caveats (However, we econometricians are masters of estimating relationships from non-experimental data).

Let me guys know if anyone has a good database or if there are some volunteers who would like to help me build such database from official tournament results that are on the net, and the deck compositions from the secret library, or just by playing and recording stats. I would publish the results here for everyone to see and benefit from it, and maybe it can even help vekn with the relaunch.

PS: If there is some large and representative (and crazy enough) group of players I could even design one of the aforementioned randomized experiments, although that would take some time to pull off.
Last edit: 03 Mar 2016 22:40 by 1muflon1.
The following user(s) said Thank You: brettscho

Please Log in or Create an account to join the conversation.

More
03 Mar 2016 23:13 #75708 by brettscho
Replied by brettscho on topic Sciencing V:tES
Awesome idea. I also lament the lack of a proper data set. The TWDA archive provides some very nice data, but it's such an incomplete set! Very difficult to draw conclusions for. Plus I think that it's inherently more difficult to balance cards in a multiplayer CCG because the players themselves help to provide a check on decks that have more powerful cards, which might make an unbalanced card seem more reasonable.

Anyway, good luck with your concept, it sounds great, I'm just not sure how to pick up more data. I'm happy to participate, but my two local play groups are certainly not representative of the community at large. Let me know how I can help!

Check out my VTES blog: Gaming with BS

I also host a google doc which separates the TWDA into clans . That means I track how often clans win, which crypt groups get used, and how many people attend events. You can access all of that info here:

Please Log in or Create an account to join the conversation.

More
04 Mar 2016 09:55 - 04 Mar 2016 09:57 #75713 by 1muflon1
Replied by 1muflon1 on topic Sciencing V:tES
I am glad that you like the idea. I will have at least look at TWDA database, although it will take me while to put all data in a single file and code it so statistical software can read it.

Also, even if your playgroup is not representative maybe analysis fo the games you guys play would provide some interesting insights (however, the question would be whether these insights then apply also to other playgroups or are just specific to yours).

For meaningful analysis, you need data at least for 30 games, and more data the more precise will the analysis be. In my playgroup we meet only each second week for two games so it would take half a year to do enough games, but if you have two groups, and if you meet more often it would go much faster.

If you are up to it, it will be best if you would record data in a way as I do in an example herehere (because it will save me a lot of the time if data are properly structured from the get go).

The more details you include the more interesting questions I can answer, but then I can understand that you want to also actually play not just record stuff so something close to my example would be enough. Maybe also add a time of play or other interesting things that I might have missed.

Also to really have good unbiased results it would be best if players were assigned to tables completely at random, so by dice or something similiar (but I guess that should be fine since at least in my playgroup that's already common practice).
Last edit: 04 Mar 2016 09:57 by 1muflon1.

Please Log in or Create an account to join the conversation.

More
04 Mar 2016 10:24 #75717 by elotar
Replied by elotar on topic Sciencing V:tES
Willingness of econometricians to make assumptions based on bad data never will stop to amuse me.

As I've said to Bret - TWDA data cannot be used as a basis of analisis of deck balance, even leaving out the question that decks for analisis are grooped by arbitrary criterion (clans, representation of one discipline e t.s.).

There are too many factors, influencing the result of this games - local metagame and conventions, seating, card availability, player skill e t.s.

I greatly suggest that only EC day 2 games can be used as a basis for any meaningful analisis of real balance of deckbuilding components.

:splat: NC Russia
:DEM::san::nec::cap4:

Please Log in or Create an account to join the conversation.

More
04 Mar 2016 10:36 #75721 by 1muflon1
Replied by 1muflon1 on topic Sciencing V:tES
But you can control for many of these factors. For example, you can control for the skill of player by including vekn ranking or years of active play in your model.

The seating might have a big influence on game granted, but if the seating is randomized, and you will record enough games it will not matter. Alternatively, you can again include seating in the model and control for it.

Also, I am not saying that what comes out of regression analysis is the word of god. However, it will give us much more information than just saying that this and that clan wins x% of time or anecdotes like I see a lot of deflection or governs being played, which is frankly meaningles.

Furthermore, I am well aware of what such analysis can answer and what it can not. If I would do it and make post about it I would be extremely careful about mentioning all possible problems or alternative interpretation so people can draw their own conclusion.

Please Log in or Create an account to join the conversation.

More
04 Mar 2016 11:09 #75726 by elotar
Replied by elotar on topic Sciencing V:tES

But you can control for many of these factors. For example, you can control for the skill of player by including vekn ranking or years of active play in your model.


Adding more meaningless data to the model will not make it better. Years of active play? Lol

Why you don't like to work with EC d2 data?

:splat: NC Russia
:DEM::san::nec::cap4:

Please Log in or Create an account to join the conversation.

More
Moderators: AnkhaKraus
Time to create page: 0.086 seconds
Powered by Kunena Forum