Regression on Game data

[sasin]

Hey all,

There are always these long conversations about why this or that is OP in the forums and the server. I was wondering if the games played data is in an easy to read form and if we've ever run any regressions on it or something (I guess this is directed primarily towards Zep).

The whole point of a regression is to let the actual evidence support or reject a claim. Everyone can make arguments about whether this or that is OP, but the actual results of matches will show if races are or aren't OP/UP. Zep, I think this would be beneficial for you as you wouldn't have to spend as much time arguing with people.

I do stats professionally. I'd be happy to do this. What I'd need is the data of games played. Some sort of dataset where you can see the rating and race of both players and who won for 1v1s. At the very least it'd be interesting to see the results.

[SeraphimLeftNut]

Great idea, except that the player base reacts to the current balance and there are no top players playing aeon, except in a few GW matches.
Arguing is part of the battle in supcom FA.

[FunkOff]

I do stats professionally. I'd be happy to do this. What I'd need is the data of games played. Some sort of dataset where you can see the rating and race of both players and who won for 1v1s. At the very least it'd be interesting to see the results.

That's interesting.

I think FA might be a bit too complicated to effectively analyze numerically, but you're welcome to give it a shot and I'm genuinely interested in what you produce, if anything.

Zep and zep alone has access to records of games. You should ask him.

[sasin]

Great idea, except that the player base reacts to the current balance and there are no top players playing aeon, except in a few GW matches.
Arguing is part of the battle in supcom FA.

Haha, I get that. And yeah I'm not saying to totally supplant it. But it'd at least be interesting to see the results.

Sparse datasets are not ideal, obviously, but there's ways of working with whatever data you have . It'd obviously be best if we had a full range of ratings and even amount of games for all factions, but given that we don't we can still look at the data and get a sense of what we can figure out. It could be fun and at least somewhat informative.

I assume the balance patch data isn't complete enough, but we could at least assess if aeon was UP before.

One big problem is that players' ratings are based on if they win or lose. It'll work a lot better if we have players who play games with different factions. I don't know the extent to which that happens. If everyone just palys as the same faction all the time, then it won't work so well, because you cant' adjust for the plyaers skill vs. the faction.

Anyways, I'm a lot better at regressions than sup com, so I think we could at least figure out some stuff...

[IceDreamer]

This is a great idea. Nothing to too complex to run the numbers Funkoff, you just need more numbers! :p

Regarding the whole way balance debating works, what we really need is an objective observer. Every single player suggesting balance changes, myself included, has personal feelings and bias getting mixed in. Perhaps you love the Aeon, perhaps you don't like the Seraphim and Cybran (I myself fall victim to loving UEF too much). I think it would be useful to find out which faction(s) certain high-profile players favour so that when we seen them argue for a change in chat or on the forums, we can get a better idea of how much of what they are saying is biased and based more on sentiment than on balancing for the good of the game.

[sasin]

I do stats professionally. I'd be happy to do this. What I'd need is the data of games played. Some sort of dataset where you can see the rating and race of both players and who won for 1v1s. At the very least it'd be interesting to see the results.

That's interesting.

I think FA might be a bit too complicated to effectively analyze numerically, but you're welcome to give it a shot and I'm genuinely interested in what you produce, if anything.

Zep and zep alone has access to records of games. You should ask him.

Thanks! I'm not intending to like break down the game with stats or anything, merely to look and see if there is anything stats can help us say about OP factions or whatever. The richer the data the better (we could see who won at what tech level or whatever) but the basic data couId let us say something about factions being OP or not.

I think Zep might have foed me when I was PMing him to ask about it... miscommunication? If you're reading this now, zep, what do you think?

[sasin]

This is a great idea. Nothing to too complex to run the numbers Funkoff, you just need more numbers! :p

Regarding the whole way balance debating works, what we really need is an objective observer. Every single player suggesting balance changes, myself included, has personal feelings and bias getting mixed in. Perhaps you love the Aeon, perhaps you don't like the Seraphim and Cybran (I myself fall victim to loving UEF too much). I think it would be useful to find out which faction(s) certain high-profile players favour so that when we seen them argue for a change in chat or on the forums, we can get a better idea of how much of what they are saying is biased and based more on sentiment than on balancing for the good of the game.

Thanks for the support Shadow! There are definitely a lot of issues with balancing etc. and I may have some other thoughts, but I certainly have no preconceived notion of who is OP or UP and I think the numbers may help!

[Ze_PilOt]

They are no such stats in the DB, and I'm not planning to make some.

[sasin]

There's no data on the rating, faction, and winners of each game?

[sasin]

So I was going through my old forum posts for nostalgia and stumbled across this one. With the new wave of activity and development, I was wondering if anyone might be able to open up this data? I could still run some data analysis on different factions and maps and it could be pretty cool! Icedreamer, you sounded enthused before, anyway we can make this happen?