Print

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

[ednever]

There has been a lot of qualitative discussion on what makes a player at different levels different (Someone I'm sure can link to all of the threads below. Difference between 20 and 30, between 30 and 40 between 35 and 45, etc.)

Given how quantitative this group is, I haven't seen any quantitative looks at answering the question. The good news is there is a ton of interesting data on councilroom. I'm more an analyst than a coder, so pulling it out isn't easy, but I'll bet some of you could.

In the meantime, I'll ask some questions, and let you know what I've found with some ugly manual examinations.

First some results:

Which cards do the top players excel with vs the cards that level 30 players excel with?


I looked at the popular buy results for the current top 5 players on iso. I compared that to the top buys for the top five "Level 30" players on iso. Here's what I found:

(1) Obviously lots of variety. Out of the 25 cards on each list, only 3 were doubles (both lists had three doubles, but they were different doubles) (So each list had 22 unique cards)
(2) The doubles for the top five were: Goons, Possession and Ambassador
(3) The doubles for the Level 30 Players were: Jack of All Trades, King's Court, and Walled Village
(4) Of the 22 unique cards the Top Five had 10 attacks. The Level 30's had 6 Attacks
(5) Cards in common between Top 5s and Level 30s: Wharf, Ill Gotten Gains, Scrying Pool, Golem, Goons (2 on top 5, 1 on lvl 30)


Obviously this surface level analysis is just the tip of the iceberg. Here's is the real analysis that is screaming out to be done:

What is the a cards characteristics by player level. The top metrics I could this of are win rate on available, win rate on gain, win rate when not gained, and number gained.

We likely want to control for the win rate of a given level (as I expect all cards will go up)


I'll bet we see some interesting patterns. Cards that do very poorly for low level players, then get very good, then start to go down again (Salvager, Bank, Jack, Bishop all come to mind), while others hit a plateau early (Wharf, Torturer, Sea Hag), and others just keep getting better (Ambassador, Goons, Highway)


Anyway, thought I would tee it up and see if anyone bites.

-=-=-=-

Here is the full lists I pulled together in case people can see any other patterns I missed:

The Top 5 List:

Ambassador
Ambassador
City
Fishing Village
Fool's Gold
Golem
Goons
Goons
Highway
Ill-Gotten Gains
Inn
Ironworks
Masquerade
Minion
Native Village
Nobles
Possession
Possession
Remodel
Scrying Pool
Silk Road
Steward
Torturer
Upgrade
Wharf

The Level 30 List:

Apprentice
Bank
Bishop
Contraband
Courtyard
Develop
Golem
Goons
Haggler
Ill-Gotten Gains
Jack of All Trades
Jack of All Trades
King's Court
King's Court
Margrave
Oracle
Pawn
Salvager
Scrying Pool
Sea Hag
Stables
Walled Village
Walled Village
Watchtower
Wharf


Ed

[Captain_Frisk]

<SNIP>

The Top 5 List:
<SNIP>
Ironworks

Council room is kindof slow, so I'm not going to go hunting, but I'm assujming you're looking at % purchased?  Or is this win rate given avail?  Given the presence of IGG, I'm assuming its % purchased.

I would love to know who has Ironworks in their top 5 and to hear an explanation of why.

[jsh357]

Definitely not surprised by Ambassador being #1--I see it misplayed a lot by lower level players and pretty much always lose in games with it against high level players.  I'd be interested to see more detailed/automated results on this.

[paulbaxter]

<SNIP>

The Top 5 List:
<SNIP>
Ironworks

Council room is kindof slow, so I'm not going to go hunting, but I'm assujming you're looking at % purchased?  Or is this win rate given avail?  Given the presence of IGG, I'm assuming its % purchased.

I would love to know who has Ironworks in their top 5 and to hear an explanation of why.

Someone who always ends up with gardens on their boards???

[WanderingWinder]

@OP: You need to explain exactly how these cards landed on these lists. Because other than it being something with popular buys... it's extremely unclear.

[rrenaud]

What kind of data and what format do you want?  Getting access to structured data for further analysis should be easy.  Popular buys per player in csv?

Councilroom is super slow ;(.  Sorry.

[ednever]

@OP: You need to explain exactly how these cards landed on these lists. Because other than it being something with popular buys... it's extremely unclear.

Sorry: It's just a top five list of cards for the top 5 players (and "top 5" Level 30 Players), ranked based on Win Given Availability.

Similar lists could obviously be generated based on Win Given Purchase, or something around number gained. But I like Win given Availability best as a single metric (even if it sometimes includes things like "Not buying Pirate Ship")

Ed

[DG]

Something to consider is that the top players can't all have the same successful cards if they play each other regularly.

[HiveMindEmulator]

More useful might be to look at biggest differences in gain rate, not win rate, since as DG said, people tend to play others around their level. Gain rate, however, still indicates the cards that tend to be more important (in at least some sense).

[WanderingWinder]

What kind of data and what format do you want?  Getting access to structured data for further analysis should be easy.  Popular buys per player in csv?

Councilroom is super slow ;(.  Sorry.

Actually, I'd like two things. One, for each game, you must already be taking buys and gains, since they're up there. I just want you to sum those, to give you total buys/gains. That's the easy part. The harder, but hopefully still not too hard, part, is to run a correlation between this total number gained and win points.