Blasting to 3N with 4333
KWSchneider
9pt 4333 hands vs a 15-17 pt notrump, differentiated by opener's
count.
Tricks 15Count 16Count 17Count
0 0 0 0
1 0 0 0
2 0 0 0
3 1 0 0
4 1 0 0
5 8 3 1
6 46 18 7
7 187 107 36
8 427 357 235
9 268 366 455
10 55 136 222
11 6 12 40
12 1 1 4
13 0 0 0
Avg 8.088 8.526 8.982
SD 1.017 0.982 0.936
Freq 436 330 234
So, opposite the 43.6% likelihood of a 15count, you can expect to make
3N 33% of the time [DD - likely slightly higher in real life],
opposite a 33% 16count, you can expect to make 3N 51.5%, and opposite
23.4% 17count, the likelihood is 72.1%.
Each result is based on 1000 deal simulation with NO constraints put
on opposition hands and opener limited to 4333, 4432, 5m332 hands,
typical of most NA strong notrump partnerships.
Cheers,
Kurt
Tugger
Good stuff, Kurt. Could you run the same simulation for 4432 and 5332
shaped dummies?
Regards, Theo
OldPalooka
In particular the 'no constraints on opps cards' condition adds to the
success rate because they will bid on some number of hands and that
makes it more likely you will make better whether to bid 3NT and how
to play decisions.
Another point, at IMPs if you stop in 2NT with 15 opposite 9, you will
gain 5 or 6 43% and lose 6 or 10 33%. The other quarter are 2 and 3
IMP wins. This means yuo gain ~.65 IMPs by inviting NV and break even
V. Since the information is worth somewhere between a 10th and a
quarter of a trick to the defense at single dummy (more if you stayman
to invite), and because having 2NT as a winning option instead of a
break even option almost certainly tips the scales to blasting.
Thanks
Derek Broughton
And having done that - does it change your odds significantly when responder
only bids 3N on a "Rule of 16" hand
--
derek
Nick France
Seems about normal results. Blasting gives you about a 48% vs those
that pass while inviting gives you just over 50% vs those that just
pass. The interesting thing is those that invite get a 57% vs those
that blast. I think there is a good case that an invite bid is
needed.
Nick France
KWSchneider
Not much different for 4432...
Tricks 15Count 16Count 17Count
0 0 0 0
1 0 0 0
2 0 1 0
3 1 0 0
4 1 0 1
5 13 5 3
6 52 22 6
7 201 109 42
8 395 308 208
9 273 373 411
10 60 164 273
11 4 18 52
12 0 0 4
13 0 0 0
Avg 8.052 8.581 9.062
SD 1.056 1.064 1.009
Freq 427 352 221
Major observation is that they are virtually identical to 4333
results. 5332 later.
Cheers,
Kurt
paul
I haven't figured out how to post a link to an old thread, but if
you'll search this group for "alex martelli data on invitational 2NT"
you'll find a number of threads discussing this. One I started dates
from May 13, 2006.
Double-dummy analysis of thousands of hands generally shows 4333 to be
the best shape for notrump, followed by 5332 then 4432. This again
illustrates problems with DD analysis: with perfect information, the
defense always attacks a weak suit, and that's more likely when you
have a doubleton; in real life they simply don't, and 5332 (with a
source of tricks) is more valuable than 4333. There was a thread last
fall where someone reported the French Bridge Federation estimated the
real-life value of five card suit at notrump to be 0.4 tricks, about
the same as a ten.
1000 hands isn't enough to distinguish the different shapes, even
ignoring the DD bias; your margin of error is almost 3% (1 over square
root of sample size is a good estimator) which exceeds the differences
between the shapes.
The advantage of pass-or-bash rests on low information; no way to
determine that double-dummy, but I can say the advantage of using 2NT
as an invitational bid is pretty small. I suspect the advantage of
using 2NT for any other purpose is also quite small, so I think it's a
matter of taste.
KWSchneider
>
> - Show quoted text -
I plan on running SD [100 only due to runtime] for one or two of these
based on 1N 3N auction and 1N 2C 2D 3N auction [obviously I need to
eliminate any 4card majors in opener]. This would show the differences
[if any] for the information presented to defenders. We should also
consider the impllications of a 1N 2N invite - since it tends to deny
a 4card major in responder.
Kurt
paul
>
> - Show quoted text -
Keep in mind margin of error: with a sample size of only 100, the
margin of error will be about +/- 10%. So if the 1NT-3NT auction makes
3NT 52% of the time, you can be 95% confident that an exhaustive study
of all hands would yeild a result in the range of about 42% to 62%. I
doubt you can conclude anything with such a large MOE, but maybe the
informational edge is larger than I expect.
KWSchneider
>
> - Show quoted text -
We can do what we can do - 1000 hands would probably take 2 days to
run per simulation, since we need to run two SD [and 2 DD] sims on the
same hands.
Kurt
jogs
I convinced a ten in a secondary suit has no value. Even unsupported
jacks have little value.
On pass-or-bash, don't think it's possible to calculate the value of
withholding info from opponents. Bashing should be unnecessary
against weak opponents. Why make this board the make or break
board of the match?
paul
Double-dummy analysis convinces me your statements about tens and
jacks are simply wrong. Virtually any ten adds 8-10% to your chances
of making 3NT, and virtually any Jack adds 16-20%. Ignoring unlikely
cases such as AKQJ10, or I seriously doubt the real-life values are
significantly lower.
Eric Leong
There are some players who open more often with 14 hcp hand when the
card says 15 to 17. Perhaps, opposite such players it is less
wonderful to blast to game with 9 hcp.
Eric Leong
jogs
Not speaking about tens in long suits. Short suit tens are
nearly worthless. Tx. That ten only has value when partner
has length and value in the suit. Also doubt that Jx or Jxx
is worth anywhere near one point.
henry...@yahoo.com
Back when I first learned to play bridge, I learned to deduct one
point for
Qx
Jx
Jxx
holdings.
Even in my approaching alzheimerian senility, I find this a sensible
rule.
Henrysun909
paul
wrote:
I learned that as well -- when opening a suit. Goren included that
adjustment as well as deducting one point for 4333 shape. Goren
specifically excluded such adjustments when bidding notrump; double-
dummy analysis supports that. Very few adjustments are justified for
notrump bidding, the primary one being extra tens, which Goren
originally valued (correctly) as 1/2 point but later dropped for
simplicity. I'm not of the "Goren knew it all school" but it seems the
more we learn, the more we realize what a lot of work he put into his
analysis of valuation. And that after all was his primary contribution
-- he made very few changes to Culbertson's bidding system.
I strongly expect that single dummy analysis would show, opposite a
15-17 notrump, J9x Jxx K10x AJxx is worth bidding 3NT, with odds of
success well over 50%. I know double dummy analysis supports that.
I agree 10x isn't very impressive, but since an average hand has one
ten and I'm not upgrading unless I have three tens or two tens and a 5-
card suit I'm not apt to treat 10x any different than any other ten.
Kieran Dyke
"Nick France" <gand...@att.net> wrote in message
news:4936bcd6-2030-4629...@l26g2000yqd.googlegroups.com...
Nick France
Except, of course, that inviting will often gain the defense a trick, as
they (a) judge whether to go active or passive, or (b) work out which
holdings declarer can have which make the contract beatable, having narrowed
his range from accepting/declining the invite. And it's much worse if you
have to use Stayman to invite, as the defense know which risky lead not to
make, or count out declarer's exact shape at trick three.
Tiggrr
Nick France
> "Nick France" <gandal...@att.net> wrote in message
>
> - Show quoted text -
None of the things you mention will add much to the results when
comparing blasting to inviting. Now using Stayman and then inviting
does tell a little about declarer's hand but I find it hard to believe
it will make more than a minor difference.
Nick France
jogs
None of us know what concealing declarer's hand is
worth. Someone needs to examine those 1/2 million
hand records from okbridge to find out.
Eric Leong
> "Nick France" <gandal...@att.net> wrote in message
>
> - Show quoted text -
Another purpose for an invitational bid is to explore for the right
game.
For example, on the auction:
1NT 2NT (invitational)
the choice is not between pass and 3NT. What you really want to be
sure is to reach the right game when opener wants to accept. The 2NT
bid serves to:
1. Reach four of a major on a 5-3 fit when 3NT goes down.
2. Reach four of a major on a 4-3 or a 5-2 fit when 3NT goes down.
3. Reach five of a minor when 3NT goes down.
I have give a number of specific examples on rgb in the past with
bidding so I am not going to do so again.
Blasting to 3NT gives up on being lucky and finding a better
alternative game.
Eric Leong
Frances
>
> - Show quoted text -
The problem with this type of analysis is that it treats all 9 point
hands the same.
Even looking at purely 3=3=3=4 9-counts there is a world of difference
between
QJ2
Q32
QJ2
J432
and
AJ10
1098
1098
A1098
(the second is a 3NT bid, the first is probably a pass)
The same is true of opener's hands. While it's true that (as Eric
says) I will open 1NT on some 14-counts, I'll also accept an
invitation and some 14 and 15s, and reject on some 16s and possibly
the odd 17.
The ability for real life players to use judgement makes the existence
of an invitation more useful than a pure simulation indicates.
OldPalooka
Just as the ability of real life defenders to use the information
makes the invitation potentially costly when accepted (~55%). I
believe Kurt's numbers are an excellent indication that inviting and
pass or blast are close enough to equal in the hands of real life
players with judgement, that it is a matter of personality rather than
technique, almost certainly within the margin of error.
-- Bill Shutts
castigamatti
Andrew
It's a problem with this specific analysis. Using Thomas Andrew's Deal
program you could easily generate the hands by using 5ths count or
some other better hand evaluation metric so that responder is bashing
on hands actually worth a bash.
Andrew
KWSchneider
>
> - Show quoted text -
FWIW I can generate hands using whatever hand evaluation metric anyone
would want - unless specifically generating for HCP count, I normally
generate [for notrump] using the a 4.25, 3, 1.75, 1, 0.5 scale with
minor adjustments for length/shortness. But if someone wants the same
anlysis done based on BumRap, 5ths, KR or whatever, it's relatively
easy. In the end however, your success will be ultimately based on the
location of the cards in the defender's hands. The comparison of
methods of go/no go will primarily be in the comparison of std
deviations - a lower SD should show a better "tool".
Kurt
OldPalooka
One valid way to reduce the analytical variance [as opposed to the
real world variance] is to swap defender's hands and take the mean. I
suspect better still is to generate multiple sets of defender's hands
per offense generation and swap defender's hands for each trial. This
in theory should not affect the mean number of tricks or raw
distribution of results.
KWSchneider
>
> - Show quoted text -
Fascinating concepts - both...but wouldn't twice as many randomly
dealt hands yield the same result as the first option? I'll let Paul
step in on this.
Kurt