Recommend
3 
 Thumb up
 Hide
8 Posts

BoardGameGeek» Forums » Board Game Design » Board Game Design

Subject: Monte Carlo as a playtesting method? rss

Your Tags: Add tags
Popular Tags: [View All]
Jay Sekhon
United States
Wisconsin
flag msg tools
mbmb
I know that many a designer probably dislikes the extensive playtesting involved in making a game play well, be balanced, etc. I was wondering how effective people think a Monte Carlo simulation would be as a method of playtesting.

For those that may not be familiar, Monte Carlo simulations are frequently used in research and software testing. What it basically consists of is taking a domain of possible inputs (possible moves, in the case of a game) and randomly choosing one. More robust simulations can provide a weighted randomness (an AI, perhaps) for picking certain moves over others.

Obviously, using this has its challenges--one would need to program a computer version of the game as well as a Monte Carlo simulator, and have some way of recording the actions. However, the advantages I see are pretty tremendous--it'll allow a designer to see thousands of plays in much faster time than would otherwise be possible. Additionally, there's no upfront cost aside from time invested in programming.

So what do you guys think? Could something like this be useful as a partial substitute for regular, manual, exhaustive playtesting?
1 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
A. B. West
United States
Beech Grove
Indiana
flag msg tools
designer
publisher
Why aren't you PLAYING a game?
Avatar
mbmbmbmbmb
To see if the mechanics lead to some saddle point or feedback loop, it might work. To see if the game plays well, it arguably will not Play testing involves so much more than just seeing if a game 'works'. It involves the impressions of players - how easy it is to understand, what appears like a valid strategy, how well components feel and where they are placed, etc.
3 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Dan Keith
United States
Umatilla
Florida
flag msg tools
mb
I believe it misses one of the vital parts of play testing.

So guys, was it fun?
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
J C Lawrence
United States
Campbell
California
flag msg tools
designer
badge
Avatar
Zanshinj wrote:
So what do you guys think? Could something like this be useful as a partial substitute for regular, manual, exhaustive playtesting?


I do quite a bit of Monte Carlo work on my designs. It isn't an ultimate tool, but it works well for some rather intractable questions.
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
J C Lawrence
United States
Campbell
California
flag msg tools
designer
badge
Avatar
Dirg wrote:
So guys, was it fun?


I never ask that question. If I'm curious I'll watch the players and make my own conclusions. Tom Lehmann's first question for a new game is, Did I find the game? That's a great question and one I've added to my list. My most common question is, Was that interesting? What was and wasn't interesting?
1 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Dan Keith
United States
Umatilla
Florida
flag msg tools
mb
clearclaw wrote:
Dirg wrote:
So guys, was it fun?


I never ask that question. If I'm curious I'll watch the players and make my own conclusions. Tom Lehmann's first question for a new game is, Did I find the game? That's a great question and one I've added to my list. My most common question is, Was that interesting? What was and wasn't interesting?
However you do it a computer simulation lacks it.
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Russ Williams
Poland
Wrocław
Dolny Śląsk
flag msg tools
designer
badge
Avatar
mbmbmb
For all the people replying "testing with a computer AI player can't tell you if the game is fun", I say "DUH! I think the original poster knows that!" Please note that he asked "Could something like this be useful as a partial substitute for regular, manual, exhaustive playtesting?"

And the answer is certainly YES it can be useful as a partial substitute (not to mention an efficient method to perform zillions of tests that would be impossible by humans). It can be useful for finding unexpected glitches in the rules. The mere act of the programming the game rules will uncover ambiguities or rule holes you weren't aware of.

It would also be quite useful for finding how strategically deep the game actually is: if a simple computer program can beat human players, it's a sign the game is probably not very strategically deep and has broken easy strategy.

I know of at least one game designer who has used this technique (playing their game against simple AI players) to do testing.
1 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Russ Williams
Poland
Wrocław
Dolny Śląsk
flag msg tools
designer
badge
Avatar
mbmbmb
clearclaw wrote:
If I'm curious I'll watch the players and make my own conclusions. Tom Lehmann's first question for a new game is, Did I find the game?

Hm? And do you answer "Yes, I looked in the box, and there it was!" ?
 
 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.