[Fluffy]

han, on Dec 11 2008, 05:54 AM, said:

What a difference 1 HCP can make.. as expected?

Double dummy analysys, so KJ9xx will play the trump suit so good catching the queen almost anywhere.

you also get perfect defence of course and opponents ruffing or removing dummy's trumps in time, but those are also present when you have QJ9xx.

[jdonn]

Fluffy has a very good point. Even though I agree with your point han, you have to admit a trump suit of KJ9xx will be overstated in the double dummy analysis, as its queen-catching abilities will be outstanding

[Winstonm]

han, on Dec 11 2008, 12:54 AM, said:

Following rogerclee's suggestion I reran the simulation with a similar but better hand: KJ9xx KQJx Axx x. We both think that this is a clear invite so hopefully the simulation would support that. If not, perhaps the results shouldn't be taken seriously.

Out of 400 hands:

6 tricks: 3
7 tricks: 20
8 tricks: 100
9 tricks: 166
10 tricks: 99
11 tricks: 9
12 tricks: 3

So now if we assume that our inviting is perfect and we get to game on all hands with 10+ tricks and stop in 3S on all others then our results are:

111 times win 11: +1221 IMPs
100 times lose 5: -500 IMPs
23 times lose 3: -69 IMPs

This gives a total of +660 IMPs, overwhelming odds in favor of trying for game.

Of course the assumptions here are a little too rosy but these same assumptions said we should pass with the previous hand.

What a difference 1 HCP can make.. as expected?

That's pretty eye-opening. I would not have thought that there was that big of difference between QJ9xx and KJ9xx, all else being equal.

I suppose, those, it should, as those times that partner does have the Ace the chances of escaping for zero losers goes from virtually nil to decent.

I don't know if it can be done, but it would be interesting to see if it isn't this ability to capture the missing Queen that creates the disparity.

Thanks again, Han.

[Winstonm]

jdonn, on Dec 11 2008, 11:04 AM, said:

Fluffy has a very good point. Even though I agree with your point han, you have to admit a trump suit of KJ9xx will be overstated in the double dummy analysis, as its queen-catching abilities will be outstanding

It would be interesting (to me, anyway) to see how must change there would be with these 3 suits: QJ9xx, KJ9xx, and KJ109x.

[Cascade]

cherdano, on Dec 11 2008, 05:51 PM, said:

Lets say we are vulnerable at IMPs. The gains for a game try are easy to compute:
50 hands make game for a gain of 50*11 = 550 IMPs.

123 hands make 8 tricks for a loss of 123*5=615 IMPs.
67 hands make less than 8 tricks for a loss of 67*3 = 201 IMPs.

I hope this s clarifies how wrong it is to ignore the hands where 2S is down...

... and sometimes (especially when 2♠ was already down) we might get doubled at a higher level and lose a bunch more.

[han]

jdonn, on Dec 11 2008, 11:04 AM, said:

Fluffy has a very good point. Even though I agree with your point han, you have to admit a trump suit of KJ9xx will be overstated in the double dummy analysis, as its queen-catching abilities will be outstanding

I didn't make a point, nor did I try to make a point.

[jdonn]

han, on Dec 11 2008, 01:57 PM, said:

jdonn, on Dec 11 2008, 11:04 AM, said:

Fluffy has a very good point. Even though I agree with your point han, you have to admit a trump suit of KJ9xx will be overstated in the double dummy analysis, as its queen-catching abilities will be outstanding

I didn't make a point, nor did I try to make a point.

han, on Dec 11 2008, 12:54 AM, said:

What a difference 1 HCP can make

That's not a point? Sorry for the apparant mislabeling, next time I'll choose between phrase, clause, or parenthetical aside.

[han]

Whatever Josh, the simulation results were very different after changing a queen to a king. I pointed out that there was a big difference, this does not mean that I was making a point. I did not claim that the first simulation said we should pass nor did I claim that the second simulation says that we should invite with that hand.

In fact, I pointed out some factors why these simulations should be regarded somewhat skeptically. Fluffy pointed out one factor of why the results of the second simulation might be so different from the first simulation, I agree that that was a very good observation.

Also, I didn't run the second simulation to make a point, I ran it because (a) Roger asked me to run the same simulation for that exact hand, and ( I thought that perhaps the first simulation results were more negative than they should be. Running the same search for a hand where my intuition says we should clearly invite can be a helpful: if the simulation results again suggested that one should pass then that would be reason to be even more skeptical about the first simulation. It turned out that this was not the case but that doesn't mean that either simulation should be accepted as enough evidence for the best call at the table, and I certainly didn't say they should.

It remains an interesting question how closely simulation results match real world results. I've read some work that was done in this direction but I never found what I was looking for. There is data suggesting that for large sets of hands (i.e. different hands dealt to us), the average number of tricks a double dummy simulation gives is quite close to the average number of tricks taken in real life, and that in real life, declarer tends to take more tricks in partscore contracts, fewer tricks in slams and about the same number of tricks in games, and for each of these situations the margin is not very large.

But that does not mean that for any specific hand the double dummy results will be close to the at-the-table results, it is easily possible that for some hands the real life results tend to be much more in favor of the defense while on others the results tend to be much more in favor of declarer. So I remain very skeptical about double dummy simulations.

I think perhaps the best use of simulations is as feedback for your intuition. I first try to guess how likely game or slam is and then I run a double dummy simulation to see if it agrees with my intuition. If not, I will look at few dozen hands manually to see if I can learn something. Sometimes the hands dealt to partner are very unrealistic and I will have to adjust the specifications. More often I notice some factors I didn't consider before, and hopefully I will have learned something.

[jdonn]

I think you are getting awfully nitpicky...

Of course no one believes double dummy will mirror real life, I was just agreeing with fluffy that a trump suit with a likely guess for a queen will in particular be very different between double dummy and real life. I think in fact it would be interesting to run it for trump suits of exactly Axxxx and Kxxxx and otherwise the same hand. I rather expect the difference between them to be smaller than the difference between KJ9xx and QJ9xx, but maybe I would become illuminated.

[Hanoi5]

This discussion went away from its original idea. If you invite with that hand or not wasn't exactly what I wanted to know (although I agree that it is a little too much). The is what try wouldyou use. My partner at the table used 2NT and I agree with whoever said that responder should/could bid the suit where he had 'help'. I failed to see this (I as responder would have bid 3♣ and my partner would have passed) but I think partner should have bid 3♦ (asking for help) or 3♥ (if it shows 4 cards, giving the choice between 4♥ and 4♠).

[CSGibson]

If I make a try, it is going to be a short suit game try of 3 clubs. That being said, this is a bad 6 loser hand, and assuming a reasonable partner, I'm not making a try.

[whereagles]

If pard couldn't Drury this, we probably have no game.