[jdonn]

mikeh, on Mar 2 2009, 12:25 PM, said:

Obviously there is more to gain at imps by responding with josh's hand.. but it is (to me) a clear pass even red at imps.

Well this is the crux. Is this hand obviously better than the original one?

[mikeh]

jdonn, on Mar 2 2009, 12:28 PM, said:

mikeh, on Mar 2 2009, 12:25 PM, said:

Obviously there is more to gain at imps by responding with josh's hand.. but it is (to me) a clear pass even red at imps.

Well this is the crux. Is this hand obviously better than the original one?

1. the op said mps
2. yes, the second hand is 'obviously' better than the 1st
3. They are both, for me, clear passes even red at imps

[skjaeran]

jdonn, on Mar 2 2009, 06:28 PM, said:

mikeh, on Mar 2 2009, 12:25 PM, said:

Obviously there is more to gain at imps by responding with josh's hand.. but it is (to me) a clear pass even red at imps.

Well this is the crux. Is this hand obviously better than the original one?

I'd say the original hand is the better one.

♥QJ9x ♦T8xx beats ♥QJxx ♦Jxxx IMO.

I'd pass both hands, normally. But I can see myself inviting with the OP hand - if I need a swing.

[hanp]

Gwynn asked me to do a double dummy simulation. We used the following hand:

A73
QJ92
10964
72

Notice that the diamond 8 has become a 9, I blame Csaba.

For partner I didn't allow 18-counts, we did allow 5-card majors (15-17) and 6-card minors (14-16) but only when the suit has at most two of the top 5 cards. 5422 shape is allowed when the 5-card suit is a minor and the 4-card suit is not clubs. We generated 100 deals.

The strategy was as follows: pass 2H or 2S. Over 2D bid 2NT. Partner will expect the invitation with 16- and 17-counts, but not with a 4333 16-count.

These were the results, assuming that the rest of the field plays 1NT:

2H giving a bottom: 2 times.
2H giving a top: 20 times.

2S giving a bottom: 16 times.
2S giving a top: 14 times.

2NT giving a bottom: 16 times.
2NT giving a draw: 14 times.

3NT giving a bottom: 9 times.
3NT giving a top: 9 times.

The total is 43 wins, 43 losses and 14 draws. Whether this double dummy simulation is useful or not I'll leave to you. I did notice how often 2H does well when partner has 3 hearts, maybe bidding 2H over 2D and pulling to 2NT if partner then bids 2S would improve the results.

[hanp]

It occurred to Roger and me that the 22 hands where partner responded 2H was a little low. I ran 200 more hands and on those hnds, partner responds 2H 55 times. So indeed it seems that 22 was quite low but it seems that was just a coincidence, not because I used incorrect specifications.

It also seems that partner bidding game 18 out of 48 times is on the low side.

[hanp]

Some more data with the same specifications: on 400 deals partner bids 2H 114 times (102 times you win, 12 times you lose) and 2S 116 times (54 times you win and 62 times you lose).

If we interpolate the earlier results we get about 88 losses, 50 draws and 32 wins on the remaining 170 hands where partner bids 2D. (I don't want to redo this part because I'm lazy.)

The total is then 188 wins, 162 losses and 50 draws, and stayman is a double dummy winner.

By the way, passing 2D is bad compared to bidding 2NT. It wins little over 16% of the time, and draws about 17% of the time.

[hanp]

mikeh, on Mar 2 2009, 12:25 PM, said:

BTW, I really, really don't like the idea of ever downgrading an 18 count into a 15-17 range. If it happened, and we missed game because partner passed a flat soft 7 or 8 count, I know where I would assign the blame.

Which player would you give credit for the matchpoint top when game has no play?

Is that an unlikely scenario when one player has a hand that fits the worst 5% of 18-counts, and the other has a flat soft 7- or 8-count?

[jdonn]

mikeh, on Mar 2 2009, 12:47 PM, said:

yes, the second hand is 'obviously' better than the 1st

skaeran, on Mar 2 2009, 02:16 PM, said:

I'd say the original hand is the better one.

Money!

[mikeh]

hanp, on Mar 2 2009, 02:54 PM, said:

Gwynn asked me to do a double dummy simulation. We used the following hand:

A73
QJ92
10964
72

Notice that the diamond 8 has become a 9, I blame Csaba.

For partner I didn't allow 18-counts, we did allow 5-card majors (15-17) and 6-card minors (14-16) but only when the suit has at most two of the top 5 cards. 5422 shape is allowed when the 5-card suit is a minor and the 4-card suit is not clubs. We generated 100 deals.

The strategy was as follows: pass 2H or 2S. Over 2D bid 2NT. Partner will expect the invitation with 16- and 17-counts, but not with a 4333 16-count.

These were the results, assuming that the rest of the field plays 1NT:

2H giving a bottom: 2 times.
2H giving a top: 20 times.

2S giving a bottom: 16 times.
2S giving a top: 14 times.

2NT giving a bottom: 16 times.
2NT giving a draw: 14 times.

3NT giving a bottom: 9 times.
3NT giving a top: 9 times.

The total is 43 wins, 43 losses and 14 draws. Whether this double dummy simulation is useful or not I'll leave to you. I did notice how often 2H does well when partner has 3 hearts, maybe bidding 2H over 2D and pulling to 2NT if partner then bids 2S would improve the results.

I doubt that a large scale sample would reveal that 3N broke even. While I accept, and would always have expected, that 3N will make a significant amount of the time, I doubt that it would be 50-50. I mean, 2N did less well, and the dividing line between these two contracts is 1 hcp... it seems to me unlikely that the 1 hcp would have the equivalent of a full trick difference in outcome. Opposite a 15 count or a 4333 16, we make 8+ tricks less often than not while but opposite an acceptance, we make 9 tricks as often as we make 8 or fewer?

Since it is obvious that Stayman will be a big winner when partner has hearts, we need to be careful not to understate the costs when he doesn't, and I think your figures do that.

I also have concerns, that I have expressed in other threads, about the utility of double dummy analysis. No-one plays double dummy, so it is a very artificial means of assessment.

Having said that, I am impressed that it is closer than I thought.

Edit: I don't know what effect this will have on your figures, but I think that few experts would open 1N with a lot of 4=2=5=2 or 4=2=2=5 hands.. there is no rebid problem and these shapes lend themselves well to suit play.

[hanp]

Mike, I think 100 deals is too small a sample. The problem is that I was counting the outcomes manually, so I wouldn't like to do that for a large number of hands.

Let me think about a way to do a large number of hands without having to check them by hand, I have some ideas. I agree with you that the numbers look off, but 9 out of 18 are such small numbers that this isn't unexpected.

[hanp]

Alright, I did some computations with a much larger number of deals. On the hands where partner bids 2D and we invite with 2NT, partner will accept the invitation on about 45.1% of the hands. These are the 16- and 17- counts minus the 4333 16-counts but plus the 15-counts with a six-card minor (this I added since it seemed right to me).

On the hands where partner passes 2NT he makes it 51% of the time. On the hands where partner bids 3NT it makes 32% of the time. I think mikeh would agree that this is closer to what one would expect.

So out of 400 deals (I actually dealt about 2000 hands to get these latest numbers) we get the following results:

2H wins = 102
2H loses = 12

2S wins = 54
2S loses = 62

2NT draws = 47
2NT loses = 46

3NT wins = 25
3NT loses = 52

Total number of wins = 181
Total number of losses = 172
Total number of draws = 47

The most remarkable finding is that the total numer is still 400. The double dummy simulation still suggests Stayman is a winner. For the record, I think it will be a considerable loser in real life, especially when my expert partner is playing the hand.

[H_KARLUK]

2♣ Stayman
6.80 KnR, 19 Zar, 6+ Kleinman points, 1½ Honor tricks, no opening count.

I can live with these :
2♦ > pass ( I know maybe not good when he opened with 3-3-2-5 but we can live with 3334)
2♥ > tx pal for heaven sent rebid pass again
2♠ > 4/3 trump fit with ♣ ruff chance in a good day, pass again

[hanp]

mikeh, on Mar 2 2009, 05:05 PM, said:

Edit: I don't know what effect this will have on your figures, but I think that few experts would open 1N with a lot of 4=2=5=2 or 4=2=2=5 hands.. there is no rebid problem and these shapes lend themselves well to suit play.

I agree and I considered making this exception. The problem is that whether an expert would open 1m or 1NT probably depends on (1) the expert, (2) the honor location and (3) the strength of the hand (with 17 I would almost always open 1m while with 15 or 16 I would be more inclined to open 1NT). Since the effect of these hands is probably small I decided to ignore the problem.

However, now that you are complaining about it...

Out of 50 hands where opener is 4225 or 4225, 2S does better 27 times and 1NT does better 23 times. So if you do not allow these hands then the double dummy results would favor stayman slightly less.

Now hopefully it is clear why the effect of removing these hands will be quite small.

[hanp]

H_KARLUK, on Mar 2 2009, 05:59 PM, said:

I can live with these :
2♦ > pass

As I posted earlier, this is bad. With the new numbers bidding 2NT gives you about 28.4% of the matchpoints on average while passing gives you about 24.5%.