Recommend
 
 Thumb up
 Hide
15 Posts

BoardGameGeek» Forums » Board Game Design » Board Game Design

Subject: Testing bluffing mechanics solo rss

Your Tags: Add tags
Popular Tags: [View All]
Greg
United Kingdom
flag msg tools
designer
Avatar
mbmbmbmbmb
I'm still developing my process for design and testing, but one of the early stages is to play the game against myself. I generally find this really helpful; I can try out all sorts of different strategies (especially those that are legal but not intended plays) and see how the rules hold up. Normally this leads to a lot of improvement before I show the first prototype to playtesters, which is good since playtesters are a limited resource and you only get their first impression once.

However, now I'm working on a game that involves an element of bluffing. I'm finding it almost impossible to do this sort of self testing. I've been messing with systems of having a player choose randomly, developing different algorithms for them to use and things like that, but nothing really seems to be working. I think part of the problem is that I'm not sure how it feels to play, if I run two algorithms against each other I don't get a handle on that part of the game and if I play against an algorithm I always know what it will do.

Has anyone else run into this problem? Are there any clever solutions out there?

Edit: SPAG
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Original Dibbler
Germany
Aachen
NRW
flag msg tools
designer
What's one damage?
badge
It's na-fing!
Avatar
mbmbmbmbmb
x_equals_speed wrote:
Are there any clever solutions out there?


Yes. Testplayers.
3 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Daniel Howard
United Kingdom
Gateshead
Tyne and Wear
flag msg tools
Avatar
mbmbmbmbmb
I think you'll have to resign yourself to the fact that solo testing a bluffing game is going to be a challenge. Algorithms aren't going to work. Perhaps giving different players 'personalities' so one could be an aggressive bluffer or one could always back down easily will allow you to at least see the gameplay effects of successful or unsuccessful bluffing. None of this will tell you if its fun however. You're going to have to use real humans for that.
1 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Greg
United Kingdom
flag msg tools
designer
Avatar
mbmbmbmbmb
Is that just a recommendation to skip this stage of development and go straight to playtesters or is "testplayers" the name of something specific?
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Greg
United Kingdom
flag msg tools
designer
Avatar
mbmbmbmbmb
ChowYunBrent wrote:
I think you'll have to resign yourself to the fact that solo testing a bluffing game is going to be a challenge. Algorithms aren't going to work. Perhaps giving different players 'personalities' so one could be an aggressive bluffer or one could always back down easily will allow you to at least see the gameplay effects of successful or unsuccessful bluffing. None of this will tell you if its fun however. You're going to have to use real humans for that.


Aye, you always have to use real humans for that bit

I just tend to view playtesters as a limited resource since I can usually pull together a few hundred hours of testing in a few weeks but it gets harder to test full time the longer I continue doing it. There's a definate benefit to working out as many things as possible in advance.
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Original Dibbler
Germany
Aachen
NRW
flag msg tools
designer
What's one damage?
badge
It's na-fing!
Avatar
mbmbmbmbmb
x_equals_speed wrote:
Is that just a recommendation to skip this stage of development and go straight to playtesters or is "testplayers" the name of something specific?


I meant playtesters.
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Jessey
Canada
flag msg tools
designer
I also purchased this and do not know what to do with it!
badge
I purchased this and do not know what to do with it!
Avatar
mbmbmbmbmb
You can, as mentioned, test degenerate strategies solo. But the nature of the bluffing beast will require real people at an earlier stage than other kinds of games.

Degenerate strategies such as "always lie" and "always tell the truth" and "always predict opponent lies" and "always trust opponent" and the worst of all "random with confidence"
1 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Darrell Hanning
United States
Jacksonville
Florida
flag msg tools
badge
We will meet at the Hour of Scampering.
Avatar
mbmbmbmbmb
When I have decided to play solo a game with bidding/bluffing, I have usually used a die-based escalation system, where the first round of bidding passes on a 6, the second round on a 5 or 6, etc. Your mileage may vary.
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Sturv Tafvherd
United States
North Carolina
flag msg tools
mbmbmbmbmb
Candi and Darrell covered the systems I use.

I've even entertained making an AI algorithm / decision flowchart that would be generic enough to use for more than one game. Like Darrell, I like using a threshold coupled with a random generator ... like this:


... Let's say the game is Poker.

Threshold (T) is set from 3 to 11, depending on the best hand the AI can make:

3 - High Card
4 - One Pair
5 - Two Pair
6 - Trio
7 - Straight
8 - Flush
9 - Full House
10 - Four of a Kind
11 - Straight Flush

Modifier (M) is set to 0 initially, changes during play.

... so, let's say it's the AI' turn to bet...

roll 1 six-sided die.
if result is less than T+M, then AI will "bluff" and place a bet.
... bet can also be a calculated amount...
...... like bet = round( (1.2)^(T+M-result) * 5) + (d6) -3
otherwise (result greater or equal to T+M) the AI folds.

player (you) then responds by either fold or call

if the AI wins a hand, his Modifier (M) increases by 1 (maximum +3)
if he loses a hand, his Modifier goes down by 1 (minimum -3)


... You could use two AI's using the same basic algorithm like this:

AI1 = round( (1.2)^(T+M-result) * 5) + (d6) -3
and
AI2 = round( (1.25)^(T+M-result) * 5) + (d6) -3

The two AIs would calculate their bets independently.

Then one of them bets first,
and the other will compare the bet against what he would bet. If his bet is equal or larger, he calls. If his bet is smaller, he folds.
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Michael Barlow
Canada
Stratford
Ontario
flag msg tools
designer
badge
No one takes the time to read anymore.
Avatar
mb
Break down the possibilities into percentages convertible to rolls of 2d6.



My math is old and scratchy, but I think a roll of 6-8 is a 44.42% likelihood.
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Greg
United Kingdom
flag msg tools
designer
Avatar
mbmbmbmbmb
I'm considering combining the personality and probability options. It should only take me an hour or so to whip up an AI with a few personalities that make decisions differently and have it randomly select one to play the game with, each decision is discreet enough that it could be fed a small amount of information and produce some sort of outcome. That way I wouldn't know which personality the AI was using in a given game, due to the noise the random element adds, but could adapt a little over the course of the game. After each game I could get it to tell me which one it used and see if that shows a dominant personality.

With longer to write the AI I could make it adapt to me, but I'm not sure I'd be using my time efficiently by then
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Sturv Tafvherd
United States
North Carolina
flag msg tools
mbmbmbmbmb
x_equals_speed wrote:

With longer to write the AI I could make it adapt to me, but I'm not sure I'd be using my time efficiently by then


Oh definitely. Games should be played by people, not AI.

There was an AI discussion a while back (Candid was there too). We spoke about simple adaptive AI... like, if Rock Paper Scissors were played by using cards (like War), then the AI adaptation would be to either add or replace a card in their deck with a card that beats what the opponent played. So if you play paper, the AI adds a scissor to his deck.
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Mike L.
United States
West Virginia
flag msg tools
mbmbmbmbmb
Might I propose a different opinion. You don't need to artificially test bluffing as long as knowledge is symmetrical.

Think about texas hold 'em, a huge part is bluffing and knowing the player who is bluffing. Can you simulate nervous twitches? Or their play patterns? You can know the probability of someone having a good hand with the cards on the table and what you have, but you can never really know if they have it, unless your cards say otherwise. Therefore bluffing and determining bluffs is a practiced trait thats effectiveness changes with the people playing and if people play your game enough and understand how others play it gets balanced out. In a bluffing game, it is required that the people figure each other out to keep the game fair.

I tried this with a deck building war game (put on a long term hiatus), during which positioning and combat were heavily influenced by bluffing. Whenever people fought, you would play cards from hand to give a corresponding attack or defense bonus to units. It always balanced out, because everyone always knew what cards people had bought and the general build of people's decks, but you never really knew until you took the plunge and called their bluff.
1 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Sturv Tafvherd
United States
North Carolina
flag msg tools
mbmbmbmbmb
To summ up Mike's great point:

When it concerns bluffing, you game the player, more than play the game.
1 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Mike L.
United States
West Virginia
flag msg tools
mbmbmbmbmb
Stormtower wrote:
To summ up Mike's great point:

When it concerns bluffing, you game the player, more than play the game.


That's way too long of a summary, pull it together man.
2 
 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.