• When should computer bid game?

    From cshearts@gmail.com@21:1/5 to All on Sat Jan 2 05:53:20 2021
    I have written a bridge bidding program which gets its bids from a database.

    When there is not a DB entry with matching specifications for a hand, it can compute a bid using Bo Haglund's double-dummy analysis code.

    Since DDA does not provide accurate average long-term results due to using the exact card distribution for just one deal, my program performs 11 iterations of mixing the opponents' cards and computing the average DDA for the 11 deals.

    (While a larger number of iterations would provide a slightly more accurate result, please accept that even 100 iterations takes too long for my purposes and that even 11 iterations is far more accurate than just using the original deal's DDA.)

    I have some questions about this process:

    (1.) Right now I am saying that if a bid makes 50+% of the time and no
    other contract scores higher using average DDA, then that is the optimum contract and the computer should bid up to that contract if possible (and
    if necessary: if 5H is the optimum contract, it won't bid 5H if 4H is being passed out).

    The question is whether 50% success is sufficient for MPs.

    Say 4H+ makes "exactly" 50% of the time. Should 4H or 3H be the optimum
    heart contract be for MPs? ("Exactly" is in quotes because there's a lot of inexactness in DDA.)

    If not the 50% contract, then where do you draw the line - 55%, 60%, ...?

    (2.) Right now, the player (actually, the program bidding all 4 hands)
    knows if an opponent can make his bid and if so, the player will sacrifice
    if he can profitably do so.

    This often means that if south bids his optimum of 2S, east will sacrifice
    with 3H when his optimum is 2H.

    The question is whether or not east can sac profitably based only on
    knowing that south can make his optimum 50% (or whatever) of the time and that his own sac will only be profitable 50% of the time.

    That is, 50% of the time, east may make his bid even though it is at a
    higher level than his optimum due to the specific lay of the cards in the current deal.

    Similarly, south may go set unprofitably 50% of the time; i.e.: he
    goes set 2 at 3H even though 2H was his optimum, and not because he played poorly but just because the lay of the cards will be unfavorable 50% of the time.

    So this question is whether or not south should double 3H (in MPs) for
    what will be a set of 1+ trick(s) 50% of the time.

    This is the flip of question 1, since the higher the % is required for a contract to be optimum, the less likely south can set it. That is, if a contract makes 60% of the time, it will be harder to set than if it only
    makes 50% of the time, obviously.

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Fred.@21:1/5 to cshe...@gmail.com on Tue Jan 5 09:00:08 2021
    On Saturday, January 2, 2021 at 8:53:22 AM UTC-5, cshe...@gmail.com wrote:
    I have written a bridge bidding program which gets its bids from a database.

    When there is not a DB entry with matching specifications for a hand, it can compute a bid using Bo Haglund's double-dummy analysis code.

    Since DDA does not provide accurate average long-term results due to using the exact card distribution for just one deal, my program performs 11 iterations of mixing the opponents' cards and computing the average DDA for the 11 deals.

    (While a larger number of iterations would provide a slightly more accurate result, please accept that even 100 iterations takes too long for my purposes and that even 11 iterations is far more accurate than just using the original deal's DDA.)

    I have some questions about this process:

    (1.) Right now I am saying that if a bid makes 50+% of the time and no
    other contract scores higher using average DDA, then that is the optimum contract and the computer should bid up to that contract if possible (and
    if necessary: if 5H is the optimum contract, it won't bid 5H if 4H is being passed out).

    The question is whether 50% success is sufficient for MPs.

    Say 4H+ makes "exactly" 50% of the time. Should 4H or 3H be the optimum
    heart contract be for MPs? ("Exactly" is in quotes because there's a lot of inexactness in DDA.)

    If not the 50% contract, then where do you draw the line - 55%, 60%, ...?

    (2.) Right now, the player (actually, the program bidding all 4 hands)
    knows if an opponent can make his bid and if so, the player will sacrifice
    if he can profitably do so.

    This often means that if south bids his optimum of 2S, east will sacrifice with 3H when his optimum is 2H.

    The question is whether or not east can sac profitably based only on
    knowing that south can make his optimum 50% (or whatever) of the time and that his own sac will only be profitable 50% of the time.

    That is, 50% of the time, east may make his bid even though it is at a
    higher level than his optimum due to the specific lay of the cards in the current deal.

    Similarly, south may go set unprofitably 50% of the time; i.e.: he
    goes set 2 at 3H even though 2H was his optimum, and not because he played poorly but just because the lay of the cards will be unfavorable 50% of the time.

    So this question is whether or not south should double 3H (in MPs) for
    what will be a set of 1+ trick(s) 50% of the time.

    This is the flip of question 1, since the higher the % is required for a contract to be optimum, the less likely south can set it. That is, if a contract makes 60% of the time, it will be harder to set than if it only makes 50% of the time, obviously.

    I'd think your would need more than 50% chance of making since you
    will normally be in a non-ideal field where *any* plus score has an
    expectation of some match points.

    This is particularly true of minor suit games where they usually win
    only if they make and 3NT goes down. When 3NT goes down, 5m
    usually scores only a little bit better than 3m, and, of course, only
    if 5m makes.

    Fred.

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)