[han]

Recent threads convinced me to buy a deal simulator. I've been playing around with it and decided to double dummy test a theory by an occasional partner of Frances:

http://forums.bridgebase.com/index.php?sho...4&hl=occasional

I couldn't quite remember the exact hand so I started with Qxxx Kxxx Axx xx and searched for a large number of deals where we are on lead with this hand and the bidding has gone 1NT-3NT. For any given hand the program marks all the leads that give the lowest number of tricks. Since Gnasher's problem was posted at IMPs I decided only to check the hands where 3NT is actually beaten. These are the results:

504 hands where 3NT can be beaten.

low spade: 370
low heart: 334
low diamond: 308
low club: 336

For example, on 334 of these hands a low heart does at least as well as any other lead.

OK, so it seems like there may be some truth to this rule, at least double dummy. But notice that the club (nicknamed gnome lead around here) also does better then the heart!

Next Arend found the thread for me and I searched with the original hand Q862 K862 J62 62. Now the results are as follows:

421 hands where 3NT goes down.

low spade: 281
low heart: 256 (6 actually only 255, the 6 didn't cost in any of the other suits)
low diamond: 235
low club: 258

So again queen wins, and again the gnome lead does just better than underleading the king.

I have the following questions:

- Has anybody else tried a similar analysis and what were the methods and results?

- What do you think the effects of double dummy play are?

- Will this change your future leads?

[hotShot]

Just to understand, the gnome rule says lead from the Q or lead ♠?

What I find interesting is, that leading ♦ seems unsuccessful.

[Fluffy]

I have been thinking on doing some double dummy analysis for ranking common holdings, I wanna know if its better to lead from AKxx or KQJ10x vs suit and such. But that would take ages to compute I guess.

[han]

hotShot, on May 10 2008, 03:36 AM, said:

Just to understand, the gnome rule says lead from the Q or lead ♠?

Gnome leads from xxx or xx more often than most mortals. Frances's occasional partner claims that it is better to lead from Qxxx than Kxxx when holding both and this search suggests that that is correct.

[hotShot]

Thanks! It seems logical.
Leading from Qxxx has a much better chance that the trick goes to A or K, who would get the trick anyway. playing from Kxxx the Q could win the trick.

[jdonn]

What was the hand where leading the 6 from K862 cost a trick over leading the 2? Just curious.

[ArtK78]

I don't think this analysis takes into account many relevant factors.

For example, in my experience, it is often true that the lead from the K gives away an irrelevant trick - either a virtually meaningless overtrick at IMPs or a trick that does not add to declarer's total trick count.

On the other hand a lead from an unsupported Q often gives declarer a game going trick or even more than one trick - the lead creates no losers for declarer when he has a suit combination consisting of everything except the Q and the J, or he is just missing the Q and the lead gives him a trick (possibly more than one) by locating the Q and also loses a tempo.

I leave it to your simulation to see if my experience in this has any merit.

[han]

Matmat pointed out some typos in my original post (fixed now) and also had some questions about the data which caused me to run the simulation with a larger number of hands. Here are the constraints:

Opener: 15-17 balanced.
Responder: 8-14 HCP, no 4-card major or 6-card minor.

This is a flaw but I don't see an easy way to include more hands without having to check manually if something is a 3NT bid. I would like to include hands with a 6-card minor or 3433 hands but I will have to think about this some more.

I dealt 4000 hands, giving myself Q862 K862 J62 62 as before. On 1066 of those hands 3NT can beaten and marking the best lead(s) gives:

spade 6: 679
spade 2: 681
heart 6: 661
heart 2: 662
diamond x: 574
club x: 643

So this time hearts did a bit better than clubs. Honors all did significantly worse as expected (heart king is the worst lead).

Out of the 1066 hands the best lead can come out of any of the suits on 337 hands and can only come from one suit on 393 hands.

Also note that Art's intuition does not correspond with these results as these are only the hands where 3NT can be beaten. It may even be that underleading the king is actually better if you include the hands where 3NT cannot be beaten (which would be good at MPs). I will run that simulation right now.

Josh, I don't know how to find the hands where leading one of the 6's instead of one of the 2's costs. I'm not going to search the 5000 hands manually.

[han]

OK I went through some hands manually, here is a funny hand:

Scoring: IMP

---

What do you lead here?

This also shows that the restrictions for responder aren't right because many would not bid 3NT with the west hand.

[matmat]

I find it a little bit frightening that when I hold the hand in the original post, opp's 3nt is ~75%...

[jdonn]

han, on May 10 2008, 04:07 PM, said:

OK I went through some hands manually, here is a funny hand:

Dealer: East Vul: ???? Scoring: IMP

♠ J109 ♥ AJ1053 ♦ 10754 ♣ 7 ♠ 743 ♥ 7 ♦ K983 ♣ AKJ104 ♠ AK5 ♥ Q94 ♦ AQ ♣ Q9853 ♠ Q862 ♥ K862 ♦ J62 ♣ 62

What do you lead here?

I knew the heart 6 was the right lead! sigh...

[Mbodell]

I think your sample size is too small, as evidenced by how close the results were and how the different second place deals are changing their position. Also, even at IMPs, it may be wise to consider both the hands where 3nt is beat and where various leads blow over/under tricks. As if some of these leads are very close (and these numbers suggest the ♥ versus ♠ setting difference of around 0.5% of the hands) then the over/under trick difference might well be larger than the contract setting numbers.

Also what deal program and DD solver are you using? Would you mind posting the code of the simulation you are using? Also do you think most players with no 6 card minor and no 4 card major and only 8HCP would bid 3nt over 15-17 1nt? That seems more aggressive than most to me as I don't think I would do that without a very good source of tricks 5 card minor with only 8 HCP.

[han]

Here are the numbers when all the hands are included (so also those where 3NT makes):

total 4000

spade 2: 2719
heart 2: 2583
diamond 2: 2422
club 2: 2586

On 1321 hands there is only one suit to lead that holds declarer to the minimal number of tricks, on 1510 any suit would do.