Recommend
8 
 Thumb up
 Hide
4 Posts

Onitama» Forums » Strategy

Subject: Does the game have a tendency to stall amongst certain players? rss

Your Tags: Add tags
Popular Tags: [View All]
Ben Stanley
United States
flag msg tools
designer
badge
Avatar
mbmbmbmbmb
I played my first game (US version) tonight: both of us strong chess players, and the game seemed to be heading toward a stalemate at one point.

Our actions included Ox, Horse, Cobra, Rooster, and Monkey. There was a point when both of us advanced our lines one space and then both struggled for ways to position in the middle line without a gambit sacrifice or an obvious loss of a piece, so both players were advancing and retreating over and over.

I broke the stalling pattern with a push for a positional advantage and a victory via the way of the stream (rather than the way of the stone), and was able to win by stone as the pressure mounted, but I wonder:

* How many games (among good players) end by stream and how many by stone?
* Are stalemates a realistic or real problem? Endless loops of repeating movement? Does it happen amongst great players? Is retreating a weakness more often than not, or a viable strategy? Do strong players tend to push forward and offer sacrifices?

Thanks for the thoughts: it seems many of you have heavily analyzed the game and even coded AIs. I'd love some deeper statistical evaluation of the common game outcomes (stone, stream, or stalemate).
6 
 Thumb up
1.00
 tip
 Hide
  • [+] Dice rolls
tysen
United States
San Francisco
California
flag msg tools
badge
Avatar
mbmbmbmbmb
I was curious about this, so since you asked so nicely I ran some experiments on my AI. I ran a series of matches between what I will call a "weak" AI which is about as strong as a human player who's played a couple times. I ran another set between two much stronger AI - a level that's not a perfect player but is vastly superhuman. The results of games vary quite a bit between the two types of AI.

First of all, what kind of win was achieved?
Weak Strong
Capture 57% 12%
Goal 43% 88%


The strong AI is very good at evading capture and so must often reach the opposing goal space in order to win.

Next I looked at how many moves did the game last?
Weak Strong
0-9 14% 0%
10-19 26% 1%
20-29 22% 10%
30-39 21% 31%
40-49 11% 30%
50-59 4% 14%
60-69 1% 8%
70-79 0% 2%
80-89 0% 2%
90-99 0% 1%
100+ 0% 2%


The weak AI sometimes makes moves that hands the opponent a win, just like people do. The strong AI can avoid that. I stop the games after 100 turns, so if you want to use that as a definition of stalemate, that's fine.

Lastly I looked at if certain cards made the game tend to be short or long. Below is a table of average game length if a certain card is in the game. Essentially all the cards are the same except for Tiger which shortens the game length.
Weak Strong
Boar 25 47
Cobra 26 45
Crab 25 45
Crane 26 46
Dragon 26 41
Eel 27 45
Elephant 25 43
Frog 27 44
Goose 27 48
Horse 26 45
Mantis 26 44
Monkey 26 45
Ox 26 46
Rabbit 27 46
Rooster 27 47
Tiger 18 35


19 
 Thumb up
4.00
 tip
 Hide
  • [+] Dice rolls
Ben Stanley
United States
flag msg tools
designer
badge
Avatar
mbmbmbmbmb
Thanks, tysen. That's exactly what I wanted to know. One final question: is there a place where someone could download and challenge your AI?
6 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Kris Rhodes
United States
Indianapolis
Indiana
flag msg tools
I am interested in downloading this AI as well.
4 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Front Page | Welcome | Contact | Privacy Policy | Terms of Service | Advertise | Support BGG | Feeds RSS
Geekdo, BoardGameGeek, the Geekdo logo, and the BoardGameGeek logo are trademarks of BoardGameGeek, LLC.