[Free]

nige1, on 2011-January-11, 05:37, said:

Wow! what a clever program! Is it free ?
Should you also specify:
diamonds (south) >= clubs (south)
and
not (clubs(south) = 3 and (diamonds (south) = 3)
or something like that

That's redundant. Overcaller has 5+♣, partner already has 6♣, so we can only have 0-2♣. It's also specified that opener has 3+♦.

[nige1]

Free, on 2011-January-11, 06:05, said:

That's redundant. Overcaller has 5+♣, partner already has 6♣, so we can only have 0-2♣. It's also specified that opener has 3+♦.

Thank you, Free. Is there a free version of the dealer program?

[whereagles]

nige1, on 2011-January-11, 05:37, said:

Wow! what a clever program! Is it free ?

It's free, yes. You just write a script and plug it into this BBO page: http://www.bridgebas...aler/dealer.php

[nige1]

whereagles, on 2011-January-11, 06:11, said:

It's free, yes. You just write a script and plug it into this BBO page: http://www.bridgebas...aler/dealer.php

Thank you, Whereagles. Fantastic and fun! I reran your script and got similar results for partner's hand....
0.084 4432
0.131 4342 or 3442
0.208 3451 or 4351
0.257 4441
0.093 4450
0.148 xx6x
13.148 av HCP
Generated 1074099 hands
Produced 1000 hands
Initial random seed 1294748464
Time needed 1.641 sec
I was surprised by the frequency of the 4432 hand. Thanks again

[655321]

This is interesting, I guess the reason partner is more likely than LHO to have any spare clubs is that we are (rightly) constraining partner and overcaller from having a 5 card major, therefore LHO is stuck with 3 spades and four hearts, leaving only 6 spots available for clubs. Partner only has 3 of his spaces reserved (for diamonds), so has 10 spaces available for clubs.

Don't know without looking at some hands how well the overcall is defined, i.e. whether overcaller would pass a lot of the simulation's 5 card club suits, and would in practice have a 6th club more often than this simulation suggests. My guess is that even though we are looking at 6 clubs, overcaller would also have 6 clubs more often than the simulation's 12% of the time.

[gnasher]

If I were doing this simulation, I would exclude hands where:
- South has a strong notrump
- West has an opening bid
- East has an opening bid
- West has a takeout double
- West has only five clubs and fewer than three of the top five clubs

This is a good example of why you should give details of the criteria, rather than just the results.

Edit: And also hands where East has a 3rd-seat weak two in diamonds, hearts or spades

This post has been edited by gnasher: 2011-January-11, 08:21

[whereagles]

655321, on 2011-January-11, 07:02, said:

Don't know without looking at some hands how well the overcall is defined, i.e. whether overcaller would pass a lot of the simulation's 5 card club suits, and would in practice have a 6th club more often than this simulation suggests.

Copy-paste the script to a text editor, write "produce 10" instead of 1000 and add a line at the end "printall" (don't forget a comma after average(hcp, south)). Then plug the script into the webpage and you'll get 10 examples.

And yes, you might see some rather revolting 2♣ overcalls LOL. In my script I insisted on overcaller having KQ of clubs, but that was just about it for suit quality.

[whereagles]

gnasher, on 2011-January-11, 07:49, said:

If I were doing this simulation, I would exclude hands where:
- South has a strong notrump
- West has an opening bid
- East has an opening bid
- West has a takeout double
- West has only five clubs and fewer than three of the top five clubs

This is a good example of why you should give details of the criteria, rather than just the results.

Edit: And also hands where East has a 3rd-seat weak two in diamonds, hearts or spades

That's true, but I believe the results with the given constraints are sufficiently close to the real odds to dismiss the fine-tuning. Removing the strong NT hand is probably the most relevant factor, as it will probably shift the odds towards opener having 5 diamonds.

Factoring in all negative inferences would make the script rather boring to read too

[gwnn]

hcp (west)>6 really?

Anyway if I put in hcp(west)>10, it still gets 0.06 Interesting.

edit: nope, it was just because I had 100 hands. if I do 10000, it gets 8-9%, no matter what lower limit I'd put for west

This post has been edited by gwnn: 2011-January-11, 15:19

[whereagles]

gwnn, on 2011-January-11, 15:13, said:

1. hcp (west)>6 really?

2. Anyway if I put in hcp(west)>10, it still gets 0.06 Interesting.

1. Does it show I've been seeing some ridiculous overcalls?

2. Yeah, contrary to common sense, the 4432 isn't that unlikely. Let me rerun taking out the strong NT hand.

[whereagles]

Well, taking the strong NT out (~4% probability) and west having 10+ hcp still gives the 4432 like ~8%.

[gnasher]

whereagles, on 2011-January-11, 15:11, said:

That's true, but I believe the results with the given constraints are sufficiently close to the real odds to dismiss the fine-tuning. Removing the strong NT hand is probably the most relevant factor, as it will probably shift the odds towards opener having 5 diamonds.
Factoring in all negative inferences would make the script rather boring to read too

This sort of attitude is the reason that I treat most simulations with considerable scepticism.

If partner has two clubs, that gives LHO a club void. About 55% of hands with a void contain a six-card or longer suit. There are very few hands containing a six-card suit which would pass in third seat, non vulnerable. Of the remaning 45%, some fraction would also be worth a 3rd-seat weak two.

We have an eight-count, and RHO doesn't have an opening hand. On many hands where partner has a balanced 12-count and RHO doesn't have an opening bid, LHO will have enough for a 3rd seat opening. Therefore the chance that partner has a balanced 12-count is considerably less than your simulation suggests.

I'm running out of rant-juice, but I expect you get the idea.

[cherdano]

gnasher, on 2011-January-11, 16:39, said:

I'm running out of rant-juice, but I expect you get the idea.

I agree with the rest of your post, but I would bet against this.

[Gerben42]

I'm not going to bet on the 8% so just raisy-daisy. I will invite.

[Fluffy]

well since this has became dealer's thread now, does dealer have doubel dummy analysis?

[whereagles]

gnasher, on 2011-January-11, 16:39, said:

This sort of attitude is the reason that I treat most simulations with considerable scepticism.

Well, maybe it's just me but I for one prefer to get a rough idea of the odds rather than no idea at all. But I'll honor you. gimme a couple minutes.

[whereagles]

Ok I'm back. So I took out most weak 2 candidates for 3rd seat, together with it having an opening, and the strong NT hand from opener. It came out this:

0.078 <-- 4432
0.097 <-- 4342/3442
0.208
0.305
0.132
0.124
13.506
Generated 2405126 hands
Produced 1000 hands
Initial random seed 1294812425
Time needed 4.725 sec

As you can see, the odds for the balanced 12-14 hand do skeptically reduce from like 20% to 17,5%.

ZA CODE was

produce 1000 
dealer south 
predeal north SK2, H2, DQ432, CA98732 
 
ab1d = hcp(south)>10 and hcp(south)<21 and spades(south)<5 and hearts(south)<5 
       and diamonds(south)>2 
int  = hcp(west)>6 and clubs(west)>4 and hcp(west, clubs)>4 and  
       spades(west)<5 and hearts(west)<5 
n1st = not(hcp(south)>14 and hcp(south)<18 and 
       shape(south, any 4333 + any 4432 + any 5332))
nwk2s= not(spades(east)>5 and hcp(east, spades)>1)
nwk2h= not(hearts(east)>5 and hcp(east, hearts)>1)
nabe = not(hcp(east)>9)

condition ab1d and int and n1st and nwk2s and nwk2h and nabe

 
action 
average(shape(south, 4432)), 
average(shape(south, xx42)), 
average(shape(south, xx51)), 
average(shape(south, 4441)), 
average(shape(south, 4450)), 
average(shape(south, xx6x)), 
average(hcp(south))

and if you wanna test whether some of the east hands match your weak 2 requirements, plug in instead

produce 10
dealer south 
predeal north SK2, H2, DQ432, CA98732 
 
ab1d = hcp(south)>10 and hcp(south)<21 and spades(south)<5 and hearts(south)<5 
       and diamonds(south)>2 
int  = hcp(west)>6 and clubs(west)>4 and hcp(west, clubs)>4 and  
       spades(west)<5 and hearts(west)<5 
n1st = not(hcp(south)>14 and hcp(south)<18 and 
       shape(south, any 4333 + any 4432 + any 5332))
nwk2s= not(spades(east)>5 and hcp(east, spades)>1)
nwk2h= not(hearts(east)>5 and hcp(east, hearts)>1)
nabe = not(hcp(east)>9)

condition ab1d and int and n1st and nwk2s and nwk2h and nabe

 
action 
average(shape(south, 4432)), 
average(shape(south, xx42)), 
average(shape(south, xx51)), 
average(shape(south, 4441)), 
average(shape(south, 4450)), 
average(shape(south, xx6x)),
average(hcp(south)),
printall

here are some examples... perhaps I should remove some East hands as they are 3rd hand 1-of-a-suit openers? Or maybe the lolish hand 3? Or some overcalls worth a take out dbl instead? Well, I leave it to thee

  1.
K 2                 Q 8 5 3             A J T 6             9 7 4 
2                   Q 9 8 4 3           K J T 7             A 6 5 
Q 4 3 2             9 8 7               A K J 6 5           T 
A 9 8 7 3 2         5                   -                   K Q J T 6 4 

   2.
K 2                 A Q J 9 5           T 7 6 4             8 3 
2                   J 9 8 7 5 3         A K 6 4             Q T 
Q 4 3 2             7                   K J 8 6             A T 9 5 
A 9 8 7 3 2         6                   T                   K Q J 5 4 

   3.
K 2                 Q 7 4               A J 6 5             T 9 8 3 
2                   Q T 5 3             K 9 8 4             A J 7 6 
Q 4 3 2             K J T 9 7 6         A 8 5               - 
A 9 8 7 3 2         -                   J 5                 K Q T 6 4 

   4.
K 2                 Q 8 6 5 4           9 7 3               A J T 
2                   9 8 6 5 3           A Q J T             K 7 4 
Q 4 3 2             K T 8               A J 9 5             7 6 
A 9 8 7 3 2         -                   6 5                 K Q J T 4 

   5.
K 2                 T 9 5 4             Q 8 7 3             A J 6 
2                   Q 5 4 3             K 9 7 6             A J T 8 
Q 4 3 2             J T 9 6             A K 8 5             7 
A 9 8 7 3 2         T                   5                   K Q J 6 4 

   6.
K 2                 7 6 4               A J 8 5             Q T 9 3 
2                   K J 6 4 3           A Q 9 7             T 8 5 
Q 4 3 2             K 9 8 5             A J T 7 6           - 
A 9 8 7 3 2         J                   -                   K Q T 6 5 4 

   7.
K 2                 Q 8 5 4 3           T 7 6               A J 9 
2                   K J 8 5 3           A Q 9 7             T 6 4 
Q 4 3 2             J 7                 A K T 8 6           9 5 
A 9 8 7 3 2         4                   6                   K Q J T 5 

   8.
K 2                 J 8 7 4             A T 6 5             Q 9 3 
2                   A 9 8 6 4           Q J 7               K T 5 3 
Q 4 3 2             9 7 5               A K J T 6           8 
A 9 8 7 3 2         J                   T                   K Q 6 5 4 

   9.
K 2                 Q 8 6 4             A T 7 5             J 9 3 
2                   Q J 9 3             A T 6 5             K 8 7 4 
Q 4 3 2             K T 8 6             A J 9 5             7 
A 9 8 7 3 2         4                   5                   K Q J T 6 

  10.
K 2                 9 8 6 5             A Q T 4             J 7 3 
2                   J T 9 8 6 4 3       A K Q 7             5 
Q 4 3 2             6                   J T 8 5             A K 9 7 
A 9 8 7 3 2         J                   6                   K Q T 5 4 

[JLOGIC]

Whereagles, the 2C overcaller is a passed hand. Your simulation does not account for that. It is HUGELY likely when a passed hand overcalls at the 2 level they have a 6 card suit as they don't have a very good hand.

Even more importantly, your simulation has not accounted for the fact that east would make a takeout double rather than overcall with (34)15. It is absolutely normal to double with that shape, come on.

Given our diamond fit AND 6 card club suit, it is hugely likely that when RHO does not have 6 clubs, he has a 3 suited hand with short diamonds, or a 2 suited hand with a 5 card major. In fact almost every single one of your example hands that has only 5 clubs is a takeout X rather than a 2C overcall. Even 3325 would often double.

Your 2245 hand is not realistic because it includes one hand passing in 3rd seat with 6-5 in the majors and a pretty reasonable hand. But even on these 3325 hands, and hands where 3rd seat was 6-5, partner was STILL not 4432. The real odds of partner being 4432 is so astronomical when we have this hand that I would bet on it being more likely that you get struck by lightning. This is obvious without a simulation.

IMO you should stop trying to do simulations, you do not do them very well to the point that they are useless. Even after gnasher told you what to exclude and why, you somehow still included takeout doubles and opening hands, and some 3rd seat openers.

[whereagles]

I'm trying to do something constructive here. If you don't like the conditions I put in, whose main purpose was to show game was odds-on (which got sort of proven), then I suggest you start doing some work of your own. A bright guy like you should be able to add lines to remove East's opener (which doesn't even affect the 4432 odds much) and take-out dbl shape (that one does affect the shape, though not by much, I'd bet).

Go on, try it. I'm sure you can do it...

[gnasher]

I'm not having a go at you in particular. My main point is that carrying out a good simulation is difficult.

I agree that the problems I described probably don't affect your main qualitative conclusion, which was that partner usually has five diamonds. However, you did also say "Yeah I also thought 8% was too much, but that's what the simulation yield." and "Yeah, contrary to common sense, the 4432 isn't that unlikely."

In my opinion, simulators should:
- Post their criteria
- Post their code, unless they're very sure it's free of errors
- Inspect some individual hands and revisit the criteria if necessary
- Include details of any subjective decisions they made (such as excluding hands manually)
- Be sceptical about any surprising results