What is the Best Faction and Why?
Long overdue results from my faction survey, and more data science fun
Note: If you are interested in watching me in a game of Twilight Imperium I’ll be playing on SpaceCatsPeaceTurtles Twitch twice in the next week. The first will be tomorrow (5/20) in a tournament game at 11:00 UTC / 7 am EST. Then I’ll be playing again next Friday (5/27) in another game at 16:00 UTC / 12 pm EST. I obviously won’t be reading it during the game but it’d be cool to have fans in chat!
So this article is long overdue. Way back in December of last year I put out a faction ranking survey, as a follow-up to the one I’d previously run to write this article about ranking every technology in the game. However, then school and life got in the way so I never wrote up the results. However, with big changes now arriving for Yin and Naalu, along with a new faction in the mix, I thought it was now or never, so I dusted off the old data science cap and got to work.
Overview of Results
As a quick explanation, the survey used what was called pair-wise match-ups. Essentially the user was shown two random factions, and asked to rank which one was better (whatever that meant to them). With over 6,000 results of individual match-ups you can then calculate how often each faction “won” a match-up to create a ranked list. The results were as follows. As a reminder these answers pre-date Codex 3 and the SCPT tournament (both of which I suspect will impact people’s perception).
Overall, I think these results are not really surprising. The only one that perhaps sticks out to me is Winnu being exactly middle of the pack, despite their powerful Hero and Commander. That being said, they do still have a weak start, and are known for drawing table heat even when behind the pack, leading many players to prefer not to play them.
Next, I wanted to compare this “perception” based approach of asking people to rank factions to empirical data from games. So using data from ParsleySage’s Google dashboard I plotted how the community ranking compared to win-rates on TTS. The TTS data is all from PoK games, but I split it into games with only 6 players, and then also all player counts to get a larger sample size.
As you can see, for the most part the community’s rankings (along the x-axis) mostly match up with the faction’s win-rate. Notably though, Winnu in both cases is an outlier, winning far more often than any other faction, while still maintaining a low community ranking.
I think there are a few potential explanations for this. First, certainly in the all player count games, I think Winnu has an advantage. Even in base game Winnu performed much better in four player than six. Second, I think this backs-up the idea of Winnu as a “beat them up or they’ll win” faction. People don’t percieve them as good, because often they are subject to intense knee-capping (on top of an already weak start). But, if they are left alone too much they have an incredibly high win rate.
But Why Are They Good?
In addition to this ranking though, I wanted to figure out why factions performed well or were ranked high. To do that, I made a huge spreadsheet of quantifiable variables of a faction. Then I narrowed it down to the following seven that I thought represented diverse parts of a faction.1
Does a faction start with blue tech? This is a binary variable, and was chosen due to blue tech’s overall strength, as well as importance in claiming custodians.
Does a faction have 2 carriers & 4 infantry? The classic “2c4i” metric, but specifically designated to not include agent abilities2, but to include factions with special units. Basically, if you can take two systems, each with two planets, on your first two actions then you count as having it.
How much “Move 2 Capacity” does a faction have? This refers to how many infantry a faction could land on a system 2 tiles away on their first action (so no tech or action cards). This once again is hugely important for claiming equidistants first, as well as getting custodians.
How many “Tech Levels” does a faction start with? This is just a total of the techs with level two techs counting double (to represent their higher utility). So for example, Xxcha with Graviton Laser systems has 2 “Tech Levels”, Jol-Nar has 4, and Sardakk has 0.
Total Resource and Influence at Home
Commodity Value
Using a Multiple-Linear Regression model, I can estimate which of these variables is most indicative of a strong win rate or community ranking. However, do take these with a grain of salt. They’re just some patterns in the data, nothing proving causality, and of course they gloss over the dozens of abilities and faction techs that add the nuance to the game. However, below are each variable ranked in terms of their effect on both win rate and community ranking.
I’m not really sure what to make of these, so I’m honestly curious for your thoughts? I think it makes sense that tech and a strong start in terms of capacity and movement are positive indicators of success. But are they really that causal, or just a correlation. For example, obviously having higher commodities is better, even if some of the best factions (Nekro, Titans, Saar) have lower commodity values.
Either way, let me know your thoughts in the comments below and thanks for reading!
This was to avoid an issue in data science called collinearity, which I won’t get into here
I went back and forth on this, but I decided since those agents trade-off with having some other powerful agent ability that factions such as Arborec or Muaat don’t really have 2c4i, even though using their agents they’ll often get there.