Re: On the Limitations of DD Simulations

On Mar 22, 10:34 pm, Tim DeLaney <delaney.timo...@xxxxxxxxxxx> wrote:
On Mar 22, 3:16 pm, thg <thg0...@xxxxxxxxx> wrote:



On Mar 22, 2:41 pm, Tim DeLaney <delaney.timo...@xxxxxxxxxxx> wrote:

On Mar 22, 12:43 pm, Nick France <gandal...@xxxxxxx> wrote:

On Mar 21, 6:05 pm, Tim DeLaney <delaney.timo...@xxxxxxxxxxx> wrote:

My partner went down at 3NT on the following layout:
Matchpoints  club game

AJ6
KT2
K54
QT74

KT9
AJ9
AJT
J965

I did a simulation of 5000 deals, and -- to partner's chagrin -- all
5000 deals produce 11 tricks.  What did he do wrong?

South  North
1NT*  3NT    * 15-17

This was the full deal:

       AJ6
       KT2
       K54
       JT87
Q42           8753
6543          Q87
Q762          893
42             AK3
       KT9
       AJ9
       AJT
       Q965

West, not liking his chances of setting up the fourth winner in a red
suit, led a club.  East cashed the top clubs and got off play with a
third club, West discarding a diamond.  Partner, of course, cashed the
fourth club.  West threw a heart and East a spade.

My partner led the DJ and West ducked smoothly.  So South went up with
dummy's king and finessed the DT on the way back.  Rats!  West won the
DQ and pushed back a diamond.  When both followed, When East followed,
South was no better off than before.  With precious little to go on,
he led the HA and ran the HJ, losing to the king.  East, of course,
won the HQ and returned a heart.

At this point, 13 cards had been played in every suit other than
spades, so in the 3-card ending everybody held three spades.  Pard
remembered that East had discarded a spade, which meant that East was
a 4-3 favorite to hold the SQ.  Alas!  West's SQ was the setting
trick.

Note that while the simulation seemed to show that 11 tricks were
cold, partner without making a serious error took only eight.  Later
at the bar, he remarked that playing HA, HK, HT he would be a favorite
to make 9 tricks, but he felt this would be a very poor matchpoint
score.

Tim

Double Dummy is not going to help you where a hand depends on guessing
where a queen is.

On this hand, the opening lead is what I would have concentrated on..
The lead of a doubleton seems highly unusual to me and suggest
holdings he didn't want to lead away from.  I'd play West for every
queen and get two out of three right.

Another way is to cash the AK of diamonds as west discarded a diamond
and then play the Jack of diamonds and let west win.

In any case I'd guess 10% make 5, 45% make 4 and 45% make 3.  Do you
know how each result scored?

Nick France- Hide quoted text -

- Show quoted text -

This was an artificial construct to illustrate a point.  That I was
unsuccessful in constructing a deal in which every guess is a pure 50%
proposition serves to illustrate an unrelated point:  There is rarely
a pure 50% guess in bridge.

Your point seems to be that DD solvers are better than real life
declarers.  I don't think that is the case.  Studies have found that
real life declarers generally perform better than double dummy (or
more appropriately: real life defenders perform worse than double
dummy).  So long as you can quantify this declarer's advantage and
remember the limitations of double dummy simulations, they should be a
useful tool.

Also note that in your example hand I would expect a vast majority of
players to lead a non-club.  That likely gets declarer a trick closer
to the double dummy result.

Tim- Hide quoted text -

- Show quoted text -

No, my point was that the results of DD solvers depend very much on
the types of problems that the deal in question poses.  If much
depends on guesswork, the DD solver is unbeatable.  (Note that all the
objectors to my construction concede that real life declarers will
usually score less than 11 tricks.)  Surely, as you point out, there
are a family of hands real life declarers (if they are talented
enough) will score better than DD against real life defenders.

It was not my intention to *debunk* DD simulations, but rather to
point out that they have limitations.  Specifically, one cannot rely
on a simulation to give a realistic probability of success.

Tim

Of course true and it is not too difficult for me to know what caused
you to construct this. :-)
Nevertheless analyzing a declarer problem in the abstract while
assuming the defense will find the best opening lead and defense will
not give you real life probabilities either.
But I admit of course that as usual you have a point
Hard to argue with you

Rainer Herrmann



.

Relevant Pages

  • Re: On the Limitations of DD Simulations
    ... third club, West discarding a diamond. ...  I'd play West for every ... Your point seems to be that DD solvers are better than real life ... real life declarers generally perform better than double dummy (or ...
    (rec.games.bridge)
  • Re: On the Limitations of DD Simulations
    ... third club, West discarding a diamond. ... real life declarers generally perform better than double dummy (or ...
    (rec.games.bridge)
  • Re: On the Limitations of DD Simulations
    ... third club, West discarding a diamond. ... real life declarers generally perform better than double dummy (or ...
    (rec.games.bridge)
  • Re: (OT) Saying Goodbye Is Never Easy
    ... > This NG is no different from real life. ... > This is a noisy club for those with a common interest. ... > Nobody ever arrives at our portal without some baggage. ...
    (sci.astro.amateur)