Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[Not a Cylon]

So, I'm checking out my Popular Buys page for what info I can find, and I come across a few rather puzzling entries: Namely, Gold and Platinum. My “Effect With” for gold is -0.23 and my “Effect Without” is 0.76. And it's not just because I skip to Platinum in Colony games: Platinum's “Effect with” and ”without” are -0.00 and 0.58.

In other words, it looks like, for me, buying Gold is as strongly associated with losing as skipping Fishing Village is, and skipping Gold is as strongly associated with winning as buying Fishing Village. And Platinum scores similarly to crappy old Saboteur.

WTF? I know correlation isn't causation and all that good crap. And certainly if I don't buy Gold, it's usually because I have some sort of draw engine going strongly enough that I don't feel the need for it — if my 2 Silvers and 4 Coppers can buy a Province every turn, why bother? But it doesn't seem like it's that often that I go for that sort of gambit and win … or does it? And it's not like I can write this off as a fluke; Gold is in every game …

So I guess what I'm asking is, what does this tell me about my play style? Anything? Something I do well? Something I do wrong? (I'm at level 22, so I'm not doing /everything/ wrong, but I ain't no Theory either :-) ) More generally, what can one glean from one's Popular Buys page? There's so much data there, gathered from every game you play; it seems like there must be some good way to leverage it. But it also seems remarkably easy just to get confused by it …

[DStu]

rappy old Saboteur.
But it doesn't seem like it's that often that I go for that sort of gambit and win … or does it?

Doesn't matter how often you do this. The "win without" only counts the cases where you didn't buy it, and gives you the win chance given that you didn't bought. And as you told, if you didn't buy gold, you have a good reason for it, so you expect to have a higher win chance.

So what does it tell you? If you assume there is some continuity (which must not really be given) you could think that you are a little to "conservative" in your playstyle and might more often consider skipping gold, also insetups which "a little less" benefits from it.

On the other hand one could argue that there might be a phase transition in setups which clearly favour skipping gold to the ones which need gold. In this case you either
a) are better in spotting to which of the two one a given setup belongs...
or b) are better in playing these setups ...
... than your "average" playing skill suggests. In this case you should probably not care about these "gold-issue" but improve other parts of the game, to improve your skill in other parts of the game.

tldr: I don't know.

[theory]

"Effect with" and "Win rate with" are rather different things, and we should probably explain this better.  When I wrote the documentation, I didn't realize that it was actually more complex than I thought.

The way "effect with" works is this.  My personal "leverage" for, say, Mountebank, is my win rate when I buy/gain Mountebank minus my overall win rate.  The general population's "leverage" for Mountebank is their win rate when they buy/gain Mountebank minus 1.00 (the global population's win rate).  "Effect with" is how much those two leverages vary, measured in terms of stdev.  (Effect without is derived in a similar way.)

So you might have a much higher win rate when you buy/gain Mountebank than your baseline win rate, but your effect with could still be negative if that difference isn't bigger than the general population's difference.

[theory]

Specifically for your purposes:

Your baseline win rate: 1.16 ± 0.08 (this is displayed as the win rate given avail for Silver/Gold/etc.)
Your win rate when you buy/gain Gold: 1.14 ± 0.09
Your win rate when you do not buy/gain Gold: 1.25 ± 0.21

When you buy/gain Silver: 1.14 ± 0.09
When you do not buy/gain Silver: 1.32 ± 0.23

When you buy/gain Platinum: 1.23 ± 0.18
When you do not buy/gain Platinum: 1.10 ± 0.36

What this suggests to me is that you do well in sets where you do not buy Treasure, or when you skip Silver and jump to Gold immediately, since in both of those situations you do quite a bit above your baseline.  Compared to the general population, though, this means your "leverage" when you do buy Gold is less, since most people do better when they are able to buy Gold, and you actually do a little worse, because of all those games where you don't buy Treasure and win.

That effect isn't there for Platinum, where you are more like the general population: you do better when you are able to buy it, and worse when you can't buy it.

[DG]

Lies, damn lies, and statistics. You've got to be careful what you read into these things. My effect without platinum is 2.94 and without colonies 2.04. Theory's effect without provinces is 3.2 and duchies is 2.73. So how do we win any games? It's probably ending the game on emptying three piles.

[Not a Cylon]

"Effect with" and "Win rate with" are rather different things, and we should probably explain this better.  When I wrote the documentation, I didn't realize that it was actually more complex than I thought.

The way "effect with" works is this.  My personal "leverage" for, say, Mountebank, is my win rate when I buy/gain Mountebank minus my overall win rate.  The general population's "leverage" for Mountebank is their win rate when they buy/gain Mountebank minus 1.00 (the global population's win rate).  "Effect with" is how much those two leverages vary, measured in terms of stdev.  (Effect without is derived in a similar way.)

So you might have a much higher win rate when you buy/gain Mountebank than your baseline win rate, but your effect with could still be negative if that difference isn't bigger than the general population's difference.

Ah — so it's not just about how strongly Gold correlates to winning for me, but how much more (or less) strongly than for most people?

This does make more sense — I noticed I have big negative effects for Bank, which means not necessarily that I lose when I buy Banks, but that, compared to most people, I don't exploit Bank very well.

Thanks!

[Not a Cylon]

Lies, damn lies, and statistics. You've got to be careful what you read into these things. My effect without platinum is 2.94 and without colonies 2.04. Theory's effect without provinces is 3.2 and duchies is 2.73. So how do we win any games? It's probably ending the game on emptying three piles.

Which makes even more sense given Theory's explanation here: You're particularly good at low-VP (or Gardens-type) three-pile endings.

[papaHav]

If I were to look at 1 metric, it would be most negative effect without.

It controls for "your broad win-rate" because most of your strategies are without any particular card, rather than with.
Whilst still comparing the population win-rate delta to control for "powered-ness" of a card.

e.g. *To say that theory is not "buffed by mountebank" as the average person is a given... its OP!
*To say theory uses library well doesn't help him improve. (other than by encouragement!)

*But to say, 95% of my games there is no councilroom/monument/baron.
In the other 5% of games, I play-on according to standard strategies
i.e. without the councilroom! ignoring it and going with any broad range of the other 95% of strats instead.

Now I begin to lose... and yet I'm doing the same thing as before... why...


n.b. or scientifically controlling for the same population of strats as before... imho this is the most telling statistic of bias