Pocket-Monkey(tm) Discussion Forum

[Home] | [Games : Backgammon] | % of winning[Forum Rules] [Login]

% of winning

Thread #314 - Messages: 5   (some may be hidden) **This thread is closed** 
 % of winning Message #1304    Replies: 1
posted by hrqls (Matthieu Mosch) on 06/29 at 11:29
i cube problems i always see people talk about % to win/lose

(i am new to cube matches)

how does one (simply ;)) calculate ones chance to win a game ?
 
 Re: % of winning Message #1305    Replies: 1
posted by tjrm (Jim) on 07/02 at 13:42
Some thoughts:

In certain situations, one can compute probabilities fairly precisely. For example, if a player has four checkers on the one spot in his home board, while bearing off, that player knows that their are 6 of 36 possible rolls in which the checkers will be borne off in one roll, and 30 of 36 possible rolls in which the checkers will be borne off in two rolls. If the opponent has three checkers left (let's say one on the 2 and two on the one), you know that player has only 6 of 36 possible rolls to get the remaining checkers off in the cases when you don't roll doubles. Six of 36 is about a 17% chance. So to summarize, in this case one might double because if accepted, one has a) about a 17% chance of winning before the opponent rolls, b) a 69% chance of winning (83% times 83% in this case) if the opponent doesn't roll a double, and c) only a 14% chance (17% times 83%) of losing from the opponent rolling a double after you don't. These percentages can be modified, for example, when you know that only 4 or 5 of the possible doubles are sufficient, etc.

Similarly, earlier in the game, you may know that a critical roll will determine the outcome of the game with total or near certainty, and can reasonably judge the odds. For example, your opponent is substantially ahead, but you have a pair of checkers in the opponent's home board on the one spot. Your own home board is sealed up, and the critical play will come soon if at all. If not, you will have to break up your coverage on your home board. In such a case, it is likely that you will have to get a specific die -- e.g., a "5" or a "4". This is an 11 of 36 chance (about 30%), and even that chance depends on the the less-than-certain chance that your opponent will have to leave a blot for you to hit.

For a wide variety of other cases, you have to rely on experience. This comes from observing what happens in games (a good reason to play individual games as well as matches). Observation is helped, I believe, if a player resists the temptation to think about his or her opponent's luck and makes observations in a more detached way.
   
 Re: Re: % of winning Message #1306    Replies: 1
posted by hrqls (Matthieu Mosch) on 07/04 at 08:17
*nod* thanks .. i know the chances for dice rolls (at least i know them right away when i dont think about the doubles .. i should make a paper with the chances on them ;))

i guess i just need experience to judge on situations earlier in the game .. i started backgammon about half a year ago, and only played single point matches (individual games) so far :)

thanks for your answer!
     
 Re: Re: Re: % of winning Message #1307    Replies: 1
posted by tjrm (Jim) on 07/04 at 13:47
One further thought: In a thread about computers in the CatchAll forum, Will (Red Baron) mentioned that he had created a program to run monte carlo simulations for backgammong positions. Though unfamiliar with Will's implementation, I could well imagine having the capability to set up positions -- e.g., more or less identical for the two sides except that every checker is one position further away from being borne off (for a total difference of 15 pips) -- and set the program to run 100 trials to see how often the side that was behind was able to surpass the leader to win. Creating and analyzing the summaries for a designed sample of cases would condense, even improve upon, years of experience. Probably, some of the backgammon books even provide such tables.

Hrqls reports himself a software engineer, so he might try his hand at his own implementation of such simulations. He could also check out the books. Do any of our avid readers -- TJ? -- know of a good reference?

       
 Re: Re: Re: Re: % of winning Message #1308    Replies: 0
posted by hrqls (Matthieu Mosch) on 07/05 at 16:02
thanks! thats a nice idea :)
(i studied physics .. but i have to bring back some knowledge which has sunk deep .. i havent used monte carlo in quite some years :))

i am making a reversi program at the moment (so far only human vs human .. i am thinking about the tactics which the computer should use at first :))

i havent read anything in books about bg yet .. just started playing it about half a year ago .. and read what i know now in the forums on the sites where i played

Forum software by Crowder Software
Pocket-Monkey and the Pocket-Monkey logo are trademarks of T.J. Crowder and Jock Murphy. All other trademarks are the property of their respective owners.