Bonus post! To fully test a few of the observations in the last post related to the Kings, I coded up a four-player version of Ninety-Eight that changed the value of Kings to 0 instead of making them set the total discard pile value to 98. It had a striking effect in three areas!
First, it looks like the games are longer, and they are! Here’s a bar chart comparison of the average game lengths. The version with no King superpowers is about three to four times as long, and the all AI games are slightly longer with the players playing more carefully to get to 98 slower.
With longer games, should we worry that the cards will run out? Some other games shuffle the discard pile, but here every card in the discard contributes in some way to the total value, so we should hesitate to do that. Luckily, if we sum up all the point values for individual number cards, and subtract 40 for the 4 Tens, we get 104, which will just barely put us over the needed 98 before we run out of cards. Thus, the rule where a player can only draw if there are still cards will be fine.
Finally, while most of the players still believe they will win the game, our quick game ends are now missing. Only small blips appear in the middle of the game, most likely due to some noise in the random simulations that the AI players use. In the latter quarter of the game, we start to see the discard pile total rise close to 98, with players expecting others to fail with their random simulations, but these other players actually recovering with intelligent play. The game is not fully decided until the last few turns of the game.
Compare this to the previous lead history graph for four players, and we can see how the Kings bring an unexpected element to the game.
With the ability of the Kings to completely change the effect of the game, and the amount of instability and fun they add, I’ll look later into more variants to see how other superpowered cards like you see in UNO (Skip, Reverse, etc) can start to counteract the King.
Up Next - Stealing Bundles
Our next game will be Stealing Bundles, a simple fishing game related to more popular and complicated games like Casino and Scopa, which has recently been studied with Monte Carlo Tree Search AI techniques.
Give the rules a try, and I’ll walk through the RECYCLE coding for Stealing Bundles next!
(Image courtesy of BoardGameGeek)
comments powered by Disqus