Skill normalized card winning rate vs play rate

This graph shows data by analyzing end game tableaus.

Strong cards have high skill normalized winning rates and tend to be played more often. You can click on a card's icon to see its name.

Cards played as homeworlds are excluded from the data, so that they don't totally skew the play rate.

The absolute play rate is divided by the number of instances of the card in the deck, so investment credits is divided by 2, and contact specialist is divided by 3. By doing so, cheap developments do not dominate the play frequency.

From 115637 Gathering Storm games.

Skill Normalized Winning Rate
	Probability instance of card appears on tableau

Sub-analysis

Here is a graph of the number of points each six cost development scores when it both when a 6 dev is played and when not played.

Here is an animated graph of the winning rate/play rate as a function of the number of players.

Here is an animated graph on the win rate play/rate as a function of the inclusion of goals.

Goal Influence

Goal Influence Graph

Influence of goal on winning rate of homeworld.

The baseline winning rate of each homeworld is the fat dot. The winning rate with the goal is the end of the segment without the dot. Hence, you can tell the absolute rate of winning by the end of the line, and the relative change by the magnitude of the line.

Goal Influence Table

Homeworld	Baseline Winning Rate	Frequency	Most Prod worlds	Most Developments	Most Military	Most Rares or Novelties	First 5 vps	First 6 pt dev	First all phase powers	First discard	First all worlds	First 3 aliens
Alpha Centauri	1.180	16024	-0.047	-0.002	-0.048	0.092	0.007	0.016	-0.007	-0.005	0.007	-0.005
New Sparta	1.157	16538	-0.026	0.004	0.091	-0.076	-0.036	0.006	0.001	0.014	0.007	-0.004
Ancient Race	1.061	15971	0.004	0.028	-0.001	-0.027	-0.005	0.012	0.005	0.000	0.003	-0.012
Damaged Alien Factory	1.003	16000	0.074	0.009	-0.019	-0.061	-0.009	-0.013	-0.012	0.016	0.004	0.016
Separatist Colony	1.010	16342	-0.049	0.019	0.068	-0.044	-0.030	-0.002	0.023	-0.006	-0.005	-0.002
Epsilon Eridani	0.998	16194	-0.028	-0.011	0.027	0.011	0.001	-0.002	0.008	-0.011	0.014	-0.005
Earths Lost Colony	0.970	16209	0.062	-0.042	-0.046	0.027	0.034	-0.012	0.005	-0.009	-0.019	-0.007
Old Earth	0.875	16286	0.011	-0.016	-0.014	0.020	0.013	-0.004	-0.011	-0.001	-0.005	0.004
Doomed World	0.865	15749	0.008	-0.013	-0.008	0.013	0.016	0.002	-0.002	0.003	-0.002	-0.006

Player rankings by game type

Rating Methodology

Each column comes from running an Elo rating algorithm on a filtered set of games. The first number is the rating and the second number is the percentile for that rating. To display a rating, 10 2-player games, 7 3-player, or 5 4-player games are required. These differ from the Genie rating in at least the following ways.

• The ratings are computed with an Elo system with K value 15. I am unsure about what the Genie server uses. Eventually, I'll play around with fitting some more sophisticated models to the data.
• Ties do not count.
• In multiplayer games, a second place is scored the same as a last place finish. Win or bust!
• The ratings are computed in game number order (which is ordered by game start time), rather than game end time, as Genie does. Since there are some players who game the Genie system, I suspect this method may be slightly more accurate simply because players do not have much of an incentive to game it.
• This includes games from flex, which are currently all assumed to occur after the last game on genie. Similarly, all games from keldon are assumed to occur after the last game on flex.