• Stay out of game?

    From nrford100@21:1/5 to All on Thu Jun 25 03:35:48 2020
    Match Points, nobody vul., North is dealer.
    N: Q4-AK92-AKT3-AQ3 (22 HCP)
    S: J2-76-Q965-KJ654 (7 HCP)

    Double dummy analysis shows that on average (100 iterations of mixing E-W cards and recalculating DDA), the most N-S can take is 10 tricks in C/D, 8 in H, 5 in S, 7 in NT.

    Is there a normal way (i.e.: not inventing a convention for this specific deal) to open this 2C without going overboard?

    How about with a normal opening other than 2C?

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Co Wiersma@21:1/5 to All on Thu Jun 25 13:12:29 2020
    Op 25-6-2020 om 12:35 schreef nrford100:
    Match Points, nobody vul., North is dealer.
    N: Q4-AK92-AKT3-AQ3 (22 HCP)
    S: J2-76-Q965-KJ654 (7 HCP)

    Double dummy analysis shows that on average (100 iterations of mixing E-W cards and recalculating DDA), the most N-S can take is 10 tricks in C/D, 8 in H, 5 in S, 7 in NT.

    Is there a normal way (i.e.: not inventing a convention for this specific deal) to open this 2C without going overboard?

    How about with a normal opening other than 2C?


    I think for most the north hand is a 2NTopening bid.
    South will so bid 3NT and will often go down. Not always mind you.
    For example if one opponent has spades AK10 or AK109 you will make 3NT easy.

    Co Wiersma

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From ais523@21:1/5 to All on Thu Jun 25 11:43:27 2020
    nrford100 wrote:
    Match Points, nobody vul., North is dealer.
    N: Q4-AK92-AKT3-AQ3 (22 HCP)
    S: J2-76-Q965-KJ654 (7 HCP)

    Double dummy analysis shows that on average (100 iterations of mixing
    E-W cards and recalculating DDA), the most N-S can take is 10 tricks in
    C/D, 8 in H, 5 in S, 7 in NT.

    Is there a normal way (i.e.: not inventing a convention for this
    specific deal) to open this 2C without going overboard?

    How about with a normal opening other than 2C?

    North's hand has five losers and no long major, so most people won't be
    opening it 2C. Most likely they'll use a bid to show a strong balanced
    hand. The Qx in spades isn't worth 2 points, so the hand is a bit of a downgrade, but nonetheless will likely end up squarely in the common
    20-22 2NT range, and you won't be able to stop South raising to game at
    that point. Even if North downgrades to a 19-count (which would be
    extreme), South will raise to game.

    Note that 3NT will make unless the opponents lead spades; there are
    plenty of winners, so it would be a good contract if not for the
    unstopped suits. Most players choose not to look for unstopped suits, on
    the basis that telling the opponents what to lead probably hurts more
    than going down in game due to an unstopped suit does; there's no
    guarantee that the opponents will find the unstopped suit even if it
    exists.

    However, systems that look for unstopped suits can be created, and they
    will stay out of 3NT on this hand. I've been working on a somewhat
    scientific system that aims to find the best contract without worrying
    about giving away information (or about pre-empting the opponents
    using a 1NT opening bid), and it bids these hands like this:

    1C (ART, 13+ balanced or 13-15 one minor or strong 2 in a major)
    1S (ART, 7+ no 4-card major)
    3D (ART, 22-23 balanced; or might bid 3C showing 20-21 balanced)
    3H (try for 3NT, no heart stopper)
    3S (hearts stopped, no spade stopper)

    and now South needs to decide between 4C or 5C, both to play. (I
    haven't defined a pick-a-minor bid in this situation yet, because it's
    fairly rare, but 4NT is unused so that would make sense. So perhaps we
    would end up in 5D instead.)

    That said, I would personally pick the 5-level contract. I'm surprised
    that it normally goes down double-dummy; if diamonds split 3-2 and clubs
    at worst 4-1, there are 11 top tricks and sufficient entries to take
    them, so the only other risk would be an early ruff before you gain the
    lead. I would have thought that it would therefore be above 50% to make
    (thus worth bidding both at matchpoints and at IMPs), but maybe not?

    --
    ais523

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Peter Smulders@21:1/5 to Co Wiersma on Thu Jun 25 16:57:41 2020
    On 25-6-2020 13:12, Co Wiersma wrote:
    Op 25-6-2020 om 12:35 schreef nrford100:
    Match Points, nobody vul., North is dealer.
    N: Q4-AK92-AKT3-AQ3  (22 HCP)
    S: J2-76-Q965-KJ654  (7 HCP)

    Double dummy analysis shows that on average (100 iterations of mixing
    E-W cards and recalculating DDA), the most N-S can take is 10 tricks
    in C/D, 8 in H, 5 in S, 7 in NT.

    Is there a normal way (i.e.: not inventing a convention for this
    specific deal) to open this 2C without going overboard?

    How about with a normal opening other than 2C?


    I think for most the north hand is a 2NTopening bid.
    South will so bid 3NT and will often go down. Not always mind you.
    For example if one opponent has spades AK10 or AK109 you will make 3NT
    easy.

    Or one opponent has 6 spades with AK but his partner is on lead ...

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Barry Margolin@21:1/5 to cshearts@gmail.com on Thu Jun 25 11:15:00 2020
    In article <360866e2-7a39-4011-9dfd-fa9b14a3c345o@googlegroups.com>,
    nrford100 <cshearts@gmail.com> wrote:

    Match Points, nobody vul., North is dealer.
    N: Q4-AK92-AKT3-AQ3 (22 HCP)
    S: J2-76-Q965-KJ654 (7 HCP)

    Double dummy analysis shows that on average (100 iterations of mixing E-W cards and recalculating DDA), the most N-S can take is 10 tricks in C/D, 8 in H, 5 in S, 7 in NT.

    Is there a normal way (i.e.: not inventing a convention for this specific deal) to open this 2C without going overboard?

    How about with a normal opening other than 2C?

    It's virtually impossible to stay out of game with combined 29 HCP. Even
    if you downgrade both hands because of Qx and Jx, you still have more
    than enough to bid game. Few pairs have any way to discover that the
    flawed suits are opposite each other.

    But you'll have plenty of company, so it shouldn't be a disaster.

    --
    Barry Margolin
    Arlington, MA

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From John Hall@21:1/5 to cshearts@gmail.com on Thu Jun 25 16:45:20 2020
    In message <360866e2-7a39-4011-9dfd-fa9b14a3c345o@googlegroups.com>,
    nrford100 <cshearts@gmail.com> writes
    Match Points, nobody vul., North is dealer.
    N: Q4-AK92-AKT3-AQ3 (22 HCP)
    S: J2-76-Q965-KJ654 (7 HCP)

    Double dummy analysis shows that on average (100 iterations of mixing
    E-W cards and recalculating DDA), the most N-S can take is 10 tricks in
    C/D, 8 in H, 5 in S, 7 in NT.

    I'm surprised, as I'd expect 5C and 5D to make far more often than not.
    (And if spades divide 5-4, then you should normally make 8 tricks in
    NT.) Even with a 4-1 diamond break, double dummy you can surely always
    make a minor suit game unless you suffer a spade overruff on the third
    round of the suit, a first round heart ruff or the clubs are 5-0.


    Is there a normal way (i.e.: not inventing a convention for this
    specific deal) to open this 2C without going overboard?

    I doubt it, as with these values not being able to make game is
    something of a freak.


    How about with a normal opening other than 2C?

    I confess that if I was playing both hands the bidding would probably go 2NT-3NT. Looking on the bright side, if by any chance the opposition
    don't lead a spade or manage to block the suit, I'll outscore anyone
    playing in a minor suit game.
    --
    John Hall
    "It is a very sad thing that nowadays there is so little useless
    information."
    Oscar Wilde (1854-1900)

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From nrford100@21:1/5 to All on Thu Jun 25 09:38:58 2020
    On Thursday, June 25, 2020 at 6:43:29 AM UTC-5, ais523 wrote:
    ... I would personally pick the 5-level contract. I'm surprised
    that it normally goes down double-dummy; if diamonds split 3-2 and clubs
    at worst 4-1, there are 11 top tricks and sufficient entries to take
    them, so the only other risk would be an early ruff before you gain the
    lead. I would have thought that it would therefore be above 50% to make
    (thus worth bidding both at matchpoints and at IMPs), but maybe not?

    Well, you've ruined my day! :-)

    For a number of weeks I've been using average DDA optimum contracts to help my program make bids beyond what's in my bidding database (www.aeyec.com/BidBase), now I find out that I've screwed up.

    Based on your surprise about going down double-dummy, I single-stepped through my DDA averaging code to show you a thing or two.

    Following are my results. With N-S holding the big hands, I mix up the E-W cards and recalculate N-S's DDA optimums, doing this 10 times and averaging the results.

    I originally decided not to round the results up, thinking that playing MPs, you don't want to bid game if the average DDA is less than game, even by a small fraction. That just seemed logical to me, so I didn't verify it by making a chart like this one:

    N: Q4-AK92-AKT3-AQ3
    S: J2-76-Q965-KJ654

    E W
    1: K987-J5-J42-T872 AT653-QT843-87-9 C:11 D:11 Avg.: 11 11
    2: K9865-85-J742-82 AT73-QJT43-8-T97 C:11 D:11 Avg.: 11 11
    3: AKT876-J4-7-T972 953-QT853-J842-8 C:11 D:11 Avg.: 11 11
    4: K9875-T85-87-872 AT63-QJ43-J42-T9 C:11 D:11 Avg.: 11 11
    5: A65-QJT3-874-972 KT9873-854-J2-T8 C:11 D:11 Avg.: 11 11
    6: A6-QT54-J8742-T7 KT98753-J83--982 C:10 D:10 Avg.: 10.833 10.833
    7: K9853-QJ8-8742-9 AT76-T543-J-T872 C:11 D:11 Avg.: 10.857 10.857
    8: 9876-QT83-J74-92 AKT53-J54-82-T87 C:11 D:11 Avg.: 10.875 10.875
    9: AKT98-QJ54-8-872 7653-T83-J742-T9 C:11 D:11 Avg.: 10.889 10.889
    10: K75-JT543-842-72 AT9863-Q8-J7-T98 C:11 D:11 Avg.: 10.9 10.9

    As you can see, N-S can make 5C/D 9 times out of 10, so using the average was not good thinking.

    In my own defense, I'm a total idiot when it comes to statistics, barely passing an intro course when working on an MBA.

    Now my guess would be that all I have to do is determine the greatest number of tricks which could be taken at least half the time. So out of 10 mixes of E-W's cards, if N-W can take 10 tricks in a suit 6 out of 10 times, or 10 tricks 4 times and more
    than 10 tricks 2 other times, then the optimum for that suit would be 10 tricks.

    I'm not sure if, in MPs, I would want to go to game if N-S can only make it 5 times out of 10 (50-50). I usually just do 10 iterations because doing 100 takes too long, but I could do 11 to get fewer ties (6-5 instead of 5-5).

    So am I on the right track now?

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From dfm@21:1/5 to John Hall on Thu Jun 25 10:50:46 2020
    On Thursday, June 25, 2020 at 11:49:27 AM UTC-4, John Hall wrote:
    In message <360866e2-7a39-4011-9dfd-fa9b14a3c345o@googlegroups.com>, nrford100 <cshearts@gmail.com> writes
    Match Points, nobody vul., North is dealer.
    N: Q4-AK92-AKT3-AQ3 (22 HCP)
    S: J2-76-Q965-KJ654 (7 HCP)

    Double dummy analysis shows that on average (100 iterations of mixing
    E-W cards and recalculating DDA), the most N-S can take is 10 tricks in >C/D, 8 in H, 5 in S, 7 in NT.

    I'm surprised, as I'd expect 5C and 5D to make far more often than not.
    (And if spades divide 5-4, then you should normally make 8 tricks in
    NT.) Even with a 4-1 diamond break, double dummy you can surely always
    make a minor suit game unless you suffer a spade overruff on the third
    round of the suit, a first round heart ruff or the clubs are 5-0.

    I'm also surprised. What exactly does "on average ... the most N-S can take" mean? Is that literally the mean of how many tricks N-S can take double dummy? If so, the median would probably be more useful, or just the probability of making game in each
    strain, or a frequency chart of the various par contracts.

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From ais523@21:1/5 to All on Thu Jun 25 21:04:13 2020
    nrford100 wrote:
    I'm not sure if, in MPs, I would want to go to game if N-S can only
    make it 5 times out of 10 (50-50). I usually just do 10 iterations
    because doing 100 takes too long, but I could do 11 to get fewer ties
    (6-5 instead of 5-5).

    The way to think about matchpoint strategy is that you're deciding
    between two contracts; in many cases, this will be a usual one (bid
    by most of the field) and an unusual one. The usual contract will score
    an average (if everyone plays it equally well); the unusual contract
    will score a top or a bottom, depending on whether it scores more or
    less. An average scores 50%. With a top-or-bottom situation, your score,
    on average, will be equal to the proportion of the time you get a top.
    So if the unusual contract is more than 50% to make, you should bid it; otherwise, bid the usual one.

    Of course, not everyone will play it equally well, which has some
    influence on the strategy. If you're a better declarer than most, that
    should increase your tendency to play in the usual contract. Likewise,
    if you're a worse declarer than your typical opponents, playing in an
    unusual contract looks more attractive (because you'll probably score
    well below 50% playing in the usual contract).

    Game versus partscore doesn't come into it; the situation is
    mostly symmetrical between bidding a partscore when most people are in
    game, and bidding game when most people are in a partscore. (There is a
    small amount of difference related to people who are in /really/ weird contracts, e.g. sacrifices, bidding mistakes, and overly speculative
    penalty doubles, but that influence is fairly random and hard to work
    into a strategy.)

    (Meanwhile, at IMPs, games are bid even if a little less likely than 50%
    to make, because the nature of IMP scoring is that boards where game is available have more influence on the result than partscore boards. So if
    you bid game and are wrong, the resulting bad score doesn't matter much, whereas if you bid partscore and are wrong, the bad result has somewhat
    more of an effect because a gme was available.)

    --
    ais523

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Travis Crump@21:1/5 to All on Thu Jun 25 22:32:42 2020
    On 06/25/2020 12:38 PM, nrford100 wrote:
    On Thursday, June 25, 2020 at 6:43:29 AM UTC-5, ais523 wrote:
    ... I would personally pick the 5-level contract. I'm surprised
    that it normally goes down double-dummy; if diamonds split 3-2 and clubs
    at worst 4-1, there are 11 top tricks and sufficient entries to take
    them, so the only other risk would be an early ruff before you gain the
    lead. I would have thought that it would therefore be above 50% to make
    (thus worth bidding both at matchpoints and at IMPs), but maybe not?

    Well, you've ruined my day! :-)

    For a number of weeks I've been using average DDA optimum contracts to help my program make bids beyond what's in my bidding database (www.aeyec.com/BidBase), now I find out that I've screwed up.

    Based on your surprise about going down double-dummy, I single-stepped through my DDA averaging code to show you a thing or two.

    Following are my results. With N-S holding the big hands, I mix up the E-W cards and recalculate N-S's DDA optimums, doing this 10 times and averaging the results.

    I originally decided not to round the results up, thinking that playing MPs, you don't want to bid game if the average DDA is less than game, even by a small fraction. That just seemed logical to me, so I didn't verify it by making a chart like this
    one:

    N: Q4-AK92-AKT3-AQ3
    S: J2-76-Q965-KJ654

    E W
    1: K987-J5-J42-T872 AT653-QT843-87-9 C:11 D:11 Avg.: 11 11
    2: K9865-85-J742-82 AT73-QJT43-8-T97 C:11 D:11 Avg.: 11 11
    3: AKT876-J4-7-T972 953-QT853-J842-8 C:11 D:11 Avg.: 11 11
    4: K9875-T85-87-872 AT63-QJ43-J42-T9 C:11 D:11 Avg.: 11 11
    5: A65-QJT3-874-972 KT9873-854-J2-T8 C:11 D:11 Avg.: 11 11
    6: A6-QT54-J8742-T7 KT98753-J83--982 C:10 D:10 Avg.: 10.833 10.833
    7: K9853-QJ8-8742-9 AT76-T543-J-T872 C:11 D:11 Avg.: 10.857 10.857
    8: 9876-QT83-J74-92 AKT53-J54-82-T87 C:11 D:11 Avg.: 10.875 10.875
    9: AKT98-QJ54-8-872 7653-T83-J742-T9 C:11 D:11 Avg.: 10.889 10.889 10: K75-JT543-842-72 AT9863-Q8-J7-T98 C:11 D:11 Avg.: 10.9 10.9

    As you can see, N-S can make 5C/D 9 times out of 10, so using the average was not good thinking.

    In my own defense, I'm a total idiot when it comes to statistics, barely passing an intro course when working on an MBA.

    Now my guess would be that all I have to do is determine the greatest number of tricks which could be taken at least half the time. So out of 10 mixes of E-W's cards, if N-W can take 10 tricks in a suit 6 out of 10 times, or 10 tricks 4 times and more
    than 10 tricks 2 other times, then the optimum for that suit would be 10 tricks.

    I'm not sure if, in MPs, I would want to go to game if N-S can only make it 5 times out of 10 (50-50). I usually just do 10 iterations because doing 100 takes too long, but I could do 11 to get fewer ties (6-5 instead of 5-5).

    So am I on the right track now?

    ~/bridge/simulations$ time deal -i 112.tcl 10000
    5C makes 0.9274 out of 10000
    5D makes 0.9301 out of 10000

    real 1m51.182s
    user 1m51.124s
    sys 0m0.016s

    No idea what you are doing that 10 times takes a long time. This is
    probably even a fairly slow computer. You aren't going to get anything
    remotely useful from 10 runs in most simulations. (I wouldn't draw any conclusions from those results as to which contract is better.)

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From nrford100@21:1/5 to All on Fri Jun 26 03:16:45 2020
    No idea what you are doing that 10 times takes a long time.

    You misread my post. What I said was:
    "I usually just do 10 iterations because doing 100 takes too long."

    My bidding practice program deals a hand, uses Bo Haglund's DDA routine to calculate the optimum contracts for each suit for each player and displays that along with the hand.

    Because DDA optimums can be unrealistic when based on just one specific lay of the cards, the program will optionally keep the N-S cards, mix the E-W cards, and get the optimums for that lay of the cards and assign them to an array.

    This is repeated a number of times and the optimums in the arrays are averaged and displayed. Then this is repeated keeping the original E-W cards and mixing the N-S cards to get the average optimum for E-W for a total of 20 iterations (10 for each pair).

    The 10|20 iterations takes about 9 seconds on average. Doing 100|200 iterations takes about 90 seconds. If Bo's module allowed doing all this with a single call, I'm sure it would only take a small fraction of the time, but that is not an option.

    Another difference between what I was doing and what your simulation did was that mine calculates optimums in all strains, including NT, for each player as declarer since optimums can vary depending on who declares. Also, I have to record all results for
    each strain and player, not just whether or not a particular contract makes or not.

    As to how many iterations it takes to be "useful" - the things which *most* affect the optimums which are based on a single lay of the cards are whether or not finesses are on and whether or not suits split favorably. A sample size of 10 is usually
    enough to smooth out those things and it is fast enough to do for each deal automatically. At the very least, a 10|20-iteration run is frequently more accurate than optimums based on a single lay of the cards while a 100|200 run is often no different
    than a 10|20 run, so I only call for a 100|200 run when I want to verify the results of a 10|20 run.

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Douglas@21:1/5 to All on Fri Jun 26 15:04:35 2020
    nrford100 posted 6/25/2020:

    "The 10|20 iterations takes about 9 seconds on average. Doing 100|200 iterations takes about 90 seconds.”

    "As to how many iterations it takes to be "useful" - the things which *most* affect the optimums which are based on a single lay of the cards are whether or not finesses are on and whether or not suits split favorably. A sample size of 10 is usually
    enough to smooth out those things and it is fast enough to do for each deal automatically. At the very least, a 10|20-iteration run is frequently more accurate than optimums based on a single lay of the cards while a 100|200 run is often no different
    than a 10|20 run, so I only call for a 100|200 run when I want to verify the results of a 10|20 run.”

    Some very basic stat reliability numbers that are on point:

    100 iterations = 95% confidence with 3% error max.
    75 = same confidence = 4% error max.
    38 = same confidence = 8% error max.
    19 = same confidence = 16% error max.
    10 = same confidence = 32% error max.

    These values are rounded conservatively for readability.

    Douglas

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From nrford100@21:1/5 to All on Fri Jun 26 15:47:05 2020
    Some very basic stat reliability numbers that are on point:

    100 iterations = 95% confidence with 3% error max.
    75 = same confidence = 4% error max.
    38 = same confidence = 8% error max.
    19 = same confidence = 16% error max.
    10 = same confidence = 32% error max.

    Can I assume that these stats are based on the same methods/constraints I was using? To recap:
    1. For each iteration, keep the N-S cards and mix the E-W cards between those two hands.
    2. Use Bo's DDA (or equivalent) to compute the optimum contracts for each suit for each player.
    3. Repeat steps 1 & 2 for the specified number of iterations.
    4. Where I was screwing up the process was averaging the results of the iterations to get average optimums when, from what I now gather, I should have just been using the median.

    I'm now in the process of changing my code to use the medians and when that is done, I will try various sample sizes and post the results here.

    I'll be very surprised if the difference in reliability between 10 and 100 iterations will be as great as your stats show, but historically, the more certain I am about something, the wronger I tend to be.

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From nrford100@21:1/5 to Douglas on Sat Jun 27 09:22:20 2020
    On Friday, June 26, 2020 at 5:04:36 PM UTC-5, Douglas wrote:
    nrford100 posted 6/25/2020:

    "The 10|20 iterations takes about 9 seconds on average. Doing 100|200 iterations takes about 90 seconds.”

    "As to how many iterations it takes to be "useful" - the things which *most* affect the optimums which are based on a single lay of the cards are whether or not finesses are on and whether or not suits split favorably. A sample size of 10 is usually
    enough to smooth out those things and it is fast enough to do for each deal automatically. At the very least, a 10|20-iteration run is frequently more accurate than optimums based on a single lay of the cards while a 100|200 run is often no different
    than a 10|20 run, so I only call for a 100|200 run when I want to verify the results of a 10|20 run.”

    Some very basic stat reliability numbers that are on point:

    100 iterations = 95% confidence with 3% error max.
    75 = same confidence = 4% error max.
    38 = same confidence = 8% error max.
    19 = same confidence = 16% error max.
    10 = same confidence = 32% error max.

    I've changed my code to use the average median instead of the average optimum. Here is a screen shot showing the results for the deal in my original post: http://www.aeyec.com/bidbase/Bidbase%20Practice%20-%20DDA%209.jpg
    As you can see, there is very little variation between the 11 and 101 runs.
    I think that is because N-S's winners and losers are too fixed to be swayed by mixing E-W's cards. There are no finesses and very few bad distributions possible. (I changed to odd numbers of iterations to avoid 5-5 and 50-50 results.)

    Here is another deal in which there is more variance: http://www.aeyec.com/bidbase/Bidbase%20Practice%20-%20DDA%209b.jpg
    There's still nothing like a 3% error for 100 vs 32% for 10.
    However, in MPs, the question of whether to bid a major or NT can swing on a very small variance resulting in a top or bottom.

    One benefit of redoing my code is that using the median only takes 1/3rd as long as using the average optimum, so instead of 90 seconds for 100 runs, it only takes about 30, but that's still too long to use 101 runs as the default, especially since there
    are other factors at work distorting the results which are beyond my control.

    Below is part of the 101-iteration record which shows the totals after the 101st iteration. The first column is the player number, then the suit number, then the number of tricks taken when that suit is trump, then a running total for that number of
    tricks. So "1 1 11: 92" means that 92 times out of 101 iterations, south took 11 tricks when clubs were trumps.

    In hearts, south took 9 tricks 42 times and 8 tricks 42 times, so its optimum contract in hearts is 8 tricks, or 2H, because 8 is the number of tricks at which south can take the median of at least 51 total tricks.

    South:
    1 1 9: 1
    1 1 10: 8
    1 1 11: 92 < 11 C

    1 2 9: 4
    1 2 10: 5
    1 2 11: 92 < 11 D

    1 3 5: 3
    1 3 6: 1
    1 3 7: 13
    1 3 8: 42 <- 8 H
    1 3 9: 42 |

    1 4 3: 1
    1 4 4: 8
    1 4 5: 46 <- 5 S (5 tricks if spades are trump, not a bid of 5S)
    1 4 6: 46 |

    1 5 5: 1
    1 5 6: 14
    1 5 7: 32
    1 5 8: 52 <- 8 N
    1 5 9: 1 |
    1 5 10: 1 |


    West:
    2 1 1: 4
    2 1 2: 92 <- 2 C (2 tricks if clubs are trump)
    2 1 3: 5 |

    2 2 1: 20 0 C (0 tricks on average if diamonds are trump))
    2 2 2: 4 |
    2 2 3: 4 |

    2 3 3: 2
    2 3 4: 67 <- 4 H
    2 3 5: 28 |
    2 3 6: 4 |

    2 4 5: 24
    2 4 6: 34 <- 6 S
    2 4 7: 28 |
    2 4 8: 13 |
    2 4 9: 2 |

    2 5 1: 2
    2 5 2: 94 <- 2 N (Can take 2 tricks in Notrump)
    2 5 3: 5 |


    North:
    3 1 9: 3
    3 1 10: 7
    3 1 11: 91 < 11 C

    3 2 8: 1
    3 2 9: 3
    3 2 10: 4
    3 2 11: 93 < 11 D

    3 3 5: 3
    3 3 6: 1
    3 3 7: 12
    3 3 8: 43 < 8 H
    3 3 9: 42

    3 4 3: 1
    3 4 4: 9
    3 4 5: 45 < 5 S
    3 4 6: 46

    3 5 5: 1
    3 5 6: 14
    3 5 7: 32
    3 5 8: 52 < 8 N
    3 5 9: 1
    3 5 10: 1

    East:
    4 1 1: 4
    4 1 2: 92
    4 1 3: 5

    4 2 1: 19
    4 2 2: 5
    4 2 3: 4

    4 3 3: 2
    4 3 4: 67
    4 3 5: 28
    4 3 6: 4

    4 4 5: 24
    4 4 6: 34
    4 4 7: 28
    4 4 8: 13
    4 4 9: 2

    4 5 1: 2
    4 5 2: 94
    4 5 3: 5

    If anyone sees any flaws in my logic or calculations, I would appreciate hearing about it.

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