Designer: Alan Wallat

Year: Unpublished, in development @ 2023



Introduction

This report and analysis was generated on a iteration in the middle of the design phase. Hence the reference to previous versions of the game at points in the text.

This set of experiments implements the rules as defined after the playtest sessions up to 3rd May 2023. This includes the changes:

Executive Summary

The previous issue with players having no need to hire smugglers has been resolved. Smugglers are now hired (and betrayed). However the game-end triggers are now vastly different - games almost always end due to the Corruption track or deck exhaustion. This may suggest that the tracks (and/or deck sizes) need to be adjusted to obtain the desired balance of end-game conditions.

Experiment set up

Only one set of agents is reported here, with 5 seconds of thinking time per decision given that this has been found to be equivalent to good human play in previous iterations. Except where otherwise stated, the results are the average over 100 games.

Games were run with each of 2, 3 and 4 players.

Results

How long do games last, and why do they end?

Plotting the number of rounds that games last, it is just about possible for a 4-player game to finish in Round 4; but on average 3/4-player games last for 8 rounds, and 2-player games for 10 rounds.

An early finish occurs when Thugs appear early in the Smuggler deck, and a player in second place uses these to accelerate the game given the new rule that if the game ends due to Corruption, then the `winner’ actually gets zero points. The rule change certainly has an effect on the way the game is played.

Average length of game in rounds
players median, 5s mean, 5s
2P 10 10.0
3P 8 8.1
4P 8 7.6

With the new rule changes, games end either because of deck exhaustion; or because of the corruption track. The Ammonia and Contraband victory conditions that were previously dominant now trigger very rarely (and even more so once you move beyond 2 players). A tentative conclusion is that the Corruption track needs to be extended; or the Ammonia/Contraband tracks shortened. (Finding an optimum ratio of track lengths given a specific target of how often the game should end one way or the other is possible using the framework.)

This supports observations in human play-testing that the Corruption track rapidly gets to the end, and good players need to manage it down by betraying smugglers. The previous behavior observed of jockeying for rank position to sell Ammonia and Contraband immediately after other players to maximise the medal score is now less obvious. Instead players try to accelerate the Corruption track to the end if they are second; or try to be first when one of the moon decks is exhausted.

Percentage of games by end condition
players Deck%, 5s Ammonia%, 5s Contraband%, 5s Corruption%, 5s
2P 58 1 7 43
3P 44 0 0 64
4P 33 0 0 70

This is reinforced by the next set of graphs show the position of the Ammonia, Contraband and Corruption tracks at game-end. The faint vertical lines show the terminal track positions of each (25, 36 and 16 respectively).

It is immediately clear that the Corruption end-track is regularly hit, with Ammonia and Contraband tracks often not budging from zero, especially beyond 2 players.

Winning and Runner-up scores

The next graph shows the scores of players according to their ordinal position.

Note the peak at zero due to the frequency of the Corruption game-end trigger. There remains a very wide spread in scores; with zero now very common given the rules that trigger if the game ends due to corruption.

Mean winning score
playerCount 5s
2P 26.2
3P 15.8
4P 12.4
Mean score difference between winner and runner up
budget playerCount Difference
5s 2P 14.2
5s 3P 4.0
5s 4P 2.2

Medal distribution

This table summarises how many medals have been acquired by game end.

Due to the game ending prematurely (compared to previous runs), the higher-level medals are very rarely seen. Whereas previously the value 5 medals were regularly acquired, now this never happens beyond 2-players, and is very rare even then.

Average number of medals of each type taken by game end
budget PlayerCount MEDAL_2 MEDAL_3 MEDAL_4 MEDAL_5 MEDAL_6
5s 2 1.7 1.3 0.6 0.1 0
5s 3 1.3 0.8 0.3 0.0 0
5s 4 1.2 0.5 0.0 0.0 0

Card Locations

The following graphs shows how cards are distributed between their different locations in 5 sample games for each player count. Thiis is more informative than just looking at the average, as each game can be very different in the details.

There is one graph per card type (Ammonia, Contraband, Smugglers, Favours). In each case cards may be:

  • In Deck (where they all start)
  • On Moon (when they are drawn from the Deck at the start of the turn)
  • In Hand (when they are picked up by a player visiting the moon)
  • In Crew (Smugglers only, when they are hired from hand)
  • Sold (Either sold on Sirius, or used to hire a smuggler or place a Cartel - each of these is worth 1 point)
  • Discard (betrayed smugglers)

Note that Smuggler Cards are now picked up avidly and do not get left on the Moon. However Favour cards are not very active; note the sample games with 2-players in which it is possible that no player thinks it worthwhile to go the Favour moon to pick up the cards even as the number available gets up to 8 to 10.

Actions

The graphs below show the usage of different actions as a game progresses. These have been standardised to the number of times each player takes the action on average in each turn or game (depending on context).

Since every player must Move each turn, the `MoveToMoon’ is uniformly 1.0 as a baseline.

There is now (unlike the previous version of the rules), significant hiring and betraying of Smugglers.

Base Actions taken in one game per player on average
playerCount budget BetraySmuggler FavourForCartel FavourForRank HireSmuggler MoveToMoon SellCards UseCardToHireSmuggler
2P 5s 4.0 2.4 1.1 5.9 10.6 1.1 5.1
3P 5s 1.7 1.8 0.9 3.2 8.6 0.6 2.9
4P 5s 1.3 1.3 1.0 2.5 8.0 0.4 2.3

The next set of graphs show the hiring/betrayal rates of the five different smuggler types over time, and then overall in each game.

There is some indication that Thug and Cargo Runner are the most used. Thugs sin particular can be used to end the game very early if one is in second position (and hence win). Cargo Runners are useful to tactically exhaust a deck near the game end while one is in the lead position.

The next graphs show what Cartel was chosen, or the rank selected. Cartels remain useful, but are not the dominant strategy that they were in past design iterations.

Average percentage occuption of moon with cartel (over whole game)
Players Metropolis, 5s Mining Colony, 5s Outpost, 5s Processing Station, 5s
2 28 60 88 66
3 38 42 44 58
4 36 23 28 40

There is now a tendency in a 2-player game to actively seek to be 2nd at the start to avoid the Police Raid penalties, and then 1st at the end of the game to be able to end it at a point of their choosing; although the total usage of Favour cards to change Rank is quite low.

mean number of Rank choices per game
Players Rank 1, 5s Rank 2, 5s Rank 3, 5s Rank 4, 5s
2 0.44 0.64 0.00 0.00
3 0.38 0.25 0.26 0.00
4 0.38 0.22 0.18 0.18

The locations of players by Round now avoids the previous oscillation between the Ammonia and Contraband moons. The Outpost (to pick up Smugglers) is now equally popular.

Note the first-turn emphasis on the Metropolis (Favour cards) and Processing Station (Contraband) - the two required cards to set up a Cartel. After this the Metropolis is relatively little used (especially in a 2-player game, when there are plenty of rich pickings to share across three moons).