[inquiry]

BBO member "free" has posted on his blog (free's blog entry on 5-4) a theory of leads against 3NT when holding 5-4 in two suits. In a nutshell, his theory is that iwth 5-4 it is best to lead the four card suit rather than the 5 card suit.

Such a blanket statement is a little hard to evaluate. Surely the bidding should have an influence on this decision, and the placement of honors in your two suits, and if the opponents have showed suit, in their suits.

Such claims, as a rule are difficult to evaluate, It turns out that one of many recent improvements to BridgeBrowser allow you to add the effectiveness of opening leads against any contract you want. The lead can be specified in a number of ways (an honor, a specific honor, 3rd best, 4th best, fifth best, 3/5, top of nothing, etc. This feature can be combined with other searches. For instance, on a specific auciton (say 1NT-3NT), or when holding anyone at the table holds any specific number of hcp, or specific districution, or a specific suit. To apply this to the type of question that Free proposed (which lead is best against 3NT)

If anyone can think of a useful way to frame this question, I am all ears. But here is a short test. For this, opening leader was (in theory) restricted to 5431 or 5422 distribution, and the auction began 1NT with no interference (which might have affected the choice of lead). The results are displayed as average imps or MP for the 3NT contract (I should have diplayed the average imps or matchpoints for the lead...so you will have to remember that the best results are the one with the lowest scores, since we want to examine the effectiveness of the leads). Also note, these leads are all on the same group of hands. The average ability of the various players is expected to cancel out over the very large number of hands examined. Here is the data looking at 38315 3NT contracts that meet these requirements.



The top chart is just the number of hands from which a lead was made from a singleton, doubleton, tripleton, or 4 or 5 card suit at both imps and matchpoints (the key thing is the number in the boxes). The bottom chart is the number of imps or matchpoints won when leading anycard from a suit of those legnths (round is matchponts, bars are imps).

It will be up to others to interpret this data, other than I will say...

1) The vast majority lead from the 5 card suit (31533 out of the 38315 times, or 82.3% of the time.

2) Leading the singleton looks pretty bad for imps and is really bad at matchpoints.

3) There appears to be a trend at imps, leading from long suits are better (see imps won for 3NT versus lead from 3 card suit (1.0), 4 card suit (0.65) and 5 card suit (0.37). ...lower imps for declare, the better the lead. so at imps, 5th best is best...

A poster to Free's blog claimed that

Quote

Paul Marston did a simulation 15-20 years ago comparing canapé leads to longest suit with 5-4. Condition was that auction went 1NT-3NT. The finding was, if COMPLETELY disregarding suit quality differences and whethe minor or major was longer, that long suit lead beat contract more often and 4-card suit lead collected more tricks

This data over 38 thousand leads seems to agree with this claim that the best chance to set the contract is the long suit lead. The matchpoint data, with 3 card lead being better than 4 or 5, and 4 card lead being better than 5 is suggestive of the second part of the Marston quote, that leading the shorter suit is more likely to win more tricks (hence better matchpoint score). However, the data set on the 3 card lead is small here, only 1369 or 3.6% of the leads, the lead from the 4 card suit was a little better, 5806 lead, or 15.2% of the hands.

On the otherhand, the higher imp score for making the unusual lead might be influenced by the fewer comparisons. The more normal the lead, the more closely to 0 imps it come.

These results are dirt easy to generate, the question is can the criteria getting to 3NT be set such that meaningful conclusions can be drawn (this was a relatively small data search.. BridgeBrowser online has more than 200 million hands you can search.

[joshs]

The matchpoint calculation is actually a little bit more complicated than more tricks on average is better. Consider
Case A: Every Other Table Also bid 1N-3N
Then letting declarer make an overtrick when most of the field was holding decalrer to making swings as many matchpoints as failing to beat a contract that everyone else beat. Both of these swings 1/2 matchpoint per pair your direction

Case B: No Other Table bid game and went plus: Since every other table went plus, giving up an overtrick swings 0 matchpoints, but letting a game make that you could have beat swings a full matchpoint per pair your direction.

Having said that, the 1N-3N auction is usually replicated in most tables (at least when its a strong NT) in which case more tricks is better.

Anyway, this is a very interesting topic. I always thought Paul Marsden was on crack with this claim, but it isn't sounded so unreasonable now. I do think suit quality is an issue. I think leading from Qxxx without exceptional spots is hopeless.

BTW, while you are at it, there was an old bridge world article tha clamied that leads from: ATxx K9xx Q8xx blow more tricks than any other holding with that top honor (basically having your second highest card 4 ranks below is the worst possible), maybe you can tell us if this is true in practice....

[P_Marlowe]

Hi,

while I dont buy it, one could do the follwoing.

Free claims, that the lead from the 5 card suit
is more likely to hit partners doubleton.

This is simple math, maybe someone can
takeover, altough I am a mathematician myself,
I wont do probability calculations, unless forced to.

If possible one should distiguish between mayor
and minor, since if responder holds a 4 card mayor
he will quite often use Stayman, i.e. if partner does
hold a doubleton, than only opener can hold the 4
card suit.

He also said, that if we hit openers 4 card suit,
we give tricks away, but we may develop length tricks,
while if lead our 4 card suit, we are less likely, but we
wont develop tricks for our side.
This argument is of the catgory going active / passive.
the 5 card suit being active, the 4 card suit being passive,
Paul Marsden simulation (mentioned in the blog) and
your analysis would support this claim.

With kind regards
Marlowe

[inquiry]

I don't put much faith in this "analysis" so far. It was quick and dirty. I had dealer open 1NT, the auction end in 3NT with no competition (it could be any auction as long as 3NT was final with no defensive bidding), and second hand had 5431 or 5422 (i left out 5440 on purpose). All the found hands fit this auction requirement, so while others may have ended in 1NT of 6NT or 3D or 4S, the comparision is between hands that ended in 3NT. Of course, the other contracts affect the scores somewhat, but the averages here are comparing 3NT with the different leads. Of course if only one hand reached 3NT and it went down for a top on any lead, that would average in as 100% as there was no second hand to compare it too. Oh, i also restricted the combined hcp to the opening side between 23 and 30. I will think about the test you are proposing.

[the hog]

The poster on Free's blog is correct. Paul

[Flame]

I remember commenting on this idea saying that it make sense at mp but not at imp. About IMP i beieve this search is true, leading the 5 cards will take the contract down more times. (i think if the normal contract was 4NT rather then 3NT the results would have been different)
About MP, i think the results of leading from 3 cards are enfluented by partner's bidding, most wouldnt lead this suit unless partner has bid it and maybe this bid is what gave them the edge (maybe their partner overcalled were its not a clear overcall)
The 4/5 lead on MP is intresting and i think ill buy it until i have new information, so on my next mp turney if i have equal 4 and 5 ill lead the 4 .
Another important thing is leading from short suit, maybe we should reconsider leading partner's suit when having a doublton in it.

[inquiry]

I am not sure don't lead partner's suit is a conclusion that you can draw from this data. First, partner didn't bid a suit on these examples. The weaker you hand is (say no hcp or very few), maybe leading short suit is better. I guess one could try 6421 hands to see what it shows, perhaps partitioned into strongish and weakest hands.

[Free]

Funny, just the other day I was playing with Jillybean, she held a 5-4 and lead the 5 card suit. She caught my doubleton, had no entry's to finish the suit, and the defense was over. However, if she'd lead her 4 card suit, she hit my 3 card support and we'd set the contract (because she still had her entry in the long suit). You can find it here: http://online.bridgebase.com/myhands/fetch...php?id=10313611. The score was adjusted by TD, but if you look, a ♠ lead turns out better than a ♥. A ♣ or ♦ works as well according to GIB, but these are unlogical leads imo.

I think it's true that the method works better at MP, because like I said on my blog, you seldom give away tricks, while leading your 5 card suit gives away a lot. However, at imps it's probably better to try and set the contract, and IF the 5 card suit is running, you better lead it. If you want to create a length trick in your 4 card suit, you'll still need some tricks at the side to defeat the contract, and this might explain why it's better to hope for the 5 card suit to come in.

[Flame]

inquiry, on Oct 11 2006, 06:39 AM, said:

I am not sure don't lead partner's suit is a conclusion that you can draw from this data. First, partner didn't bid a suit on these examples. The weaker you hand is (say no hcp or very few), maybe leading short suit is better. I guess one could try 6421 hands to see what it shows, perhaps partitioned into strongish and weakest hands.

I missed the part about us being quite, now this is very intresting. Is leading 3 cards is a winner in general at mp ?

[zasanya]

Just want to express my great appreciation of posters like this.Useful tip for players like me.Why cant we have more of this kind rather than some of those which degenrate into personal feuds?

[P_Marlowe]

<deleted>

Marlowe

[inquiry]

zasanya, on Oct 11 2006, 08:13 AM, said:

Just want to express my great appreciation of posters like this.Useful tip for players like me.Why cant we have more of this kind rather than some of those which degenrate into personal feuds?

You are certainly welcome, but I think eventually even this one might deteriate. For one thing, I am not sure I believe the data as present myself, as the criteria for selecting the hands is probably not in the correct depth. What if the your five card suit was bid on your right? ETC. Maybe the simple minded 1NT-3NT might be better to study, but then, you have to factor in leads in a minor (which responder clearly probably holds since no stayman), which complicates your lead choice if you are 5-4 one suit in a major another in a minor

The ability to study leads is new to BridgeBrowser, in fact, it is new as of two weeks ago, and this is the first time I have played with it. I wasn't all that excited when it was added, but it does raise interetting possibilities of which lead is best on average in certain carefully characterized situations. Like, leading a singleton against a suit slam or a part-score or game. Or leading an ace against a slam without the king and with no second ace.

I am looking for input to see if this type of data can be actually useful. See the my next post.

[inquiry]

Ok, I ran another test based upon the following assumptions. Dealer again opened 1NT and his side reached 3NT. Second hand had 5431 or 5422 and either 0 to 5 hcp or 8 to 17 hcp. I allowed second hand to bid, but not his partner, so no hint on what to lead. I also restricted openers side to 23 to 28 (two less than before). And I don't know if I mentioned it, I did not allow a double or redouble of 3NT in any of the studies.

LEADING WITH 5431 or 5422 with 0 to 4 hcp


This first pair of charts are when leader is weak. As was the case in the general condition posted earlier, Leadiing from the singleton was badish, and at mp, leading from the 3 card suit was best, followied by the 4 then the 4 cards suit. The imps were a bit odd, the two card suit lead was best, followed by the five card suit, then 3 then 5. Note, only 300 or so hands lead from a two cards uit, and even less (158 and 132) lead from a three card suit. Once again the five card s uit was the heavy favorite for the opening lead. But the trend of leading from the 4 card suit being better at imps and the shortert suits (particularily the three card suit) at mp continues.

LEADING WITH 5431 or 5422 with 0 to 4 hcp

Here for the first time we have to pause to look closely at the bars. The bar for imp contact when holding a 3 card suit (yellow bar) is actually going DOWN not up (not scale on right hand starts with zero in the middle of the page). At imps when holding a strong hand, leading the 3 card s uit average a minus score (-0.77) for delcarer whcih explains why it is downward. The same occurred for leads from a doubleton, but it was almost zero. The lead from a three card suit occurred only 60 times, however, out of 9793 chances, so I would not put too much faith in that lead. Also note that 50% matchpoint is also in the middle, so that lead of doubleton at matchpoints (green clynders) points downward. There were only 42 of those leads. With a strong hand, however, now the matchpoints favored five card suit over 4 card suit, and with 848 hands at MP choosing the 4 card suit, his data maybe believable. But, you can click on the matchpoint results (the cylinder) in bridgebrowser, and it will call up all the matchponts hands where the four card suit was lead for future analysis (any of these bars, or group of bars, can be choosen to create a new data set to study).

Anyway, that is what the data shows so far. I am not sure it is helpful, as with these restrictions, there are too few short suit leads to draw meaningful conclusions.

[Al_U_Card]

This is all somewhat mind-blowing. Apart from the illogic of some aspects of bridge (suit rank etc.) most anything to do with the mathematical components and restrictions have a logical component that allows for theorizing, analysis and conclusions to be drawn about the results from any action.

Bridgebrowser studies ACTUAL results of hands played by humans. It is NOT a double dummy analysis of the potential results, is it?

As far as the analysis of the human portion, many of us would tend to skew the results by our own in-grained prejudices learned as "good bridge practice".

The results available are certainly of value when applied to the real world play. (As the analysis gives insight into what "works".)

I would be interested in the analysis of double dummy solved values to see how closely they agree (if at all!) with human endeavor. Is this possible?

[inquiry]

Al_U_Card, on Oct 11 2006, 10:02 AM, said:

I would be interested in the analysis of double dummy solved values to see how closely they agree (if at all!) with human endeavor. Is this possible?

Peter Cheung investagated 18 million and 23 million OKbridge hands using double dummy software and actual results using bridge browser, see double/single dummy results

He found

Quote

The overall total number of tricks taken by the declarer is 9.21 (9.22 for imp and 9.20 for mp).  The double dummy analysis of the same deal produce 9.11 (9.12 for imp and 9.11 for mp).  So actual play by OKBridge player takes 0.1 tricks more then the double dummy analysis result. This is from 383,000 deals and over 25 million plays

A couple of issues here... double dummy often predicts two more tricks than actually taken and two less tricks than actually taken in the real world. When you look at 25 million plays, the extra tricks on one hand cancels out the extra tricks on anonther so the AVERAGE moves to zero difference.

The theory is if you look at enough hands under the same conditions, the results will match the double dummy results. This is why I worry about 56 leads of a singleton. The "n" is not nearly big enough to be of real value.

[kenberg]

Free, on Oct 11 2006, 07:16 AM, said:

Funny, just the other day I was playing with Jillybean, she held a 5-4 and lead the 5 card suit. She caught my doubleton, had no entry's to finish the suit, and the defense was over. However, if she'd lead her 4 card suit, she hit my 3 card support and we'd set the contract (because she still had her entry in the long suit). You can find it here: http://online.bridgebase.com/myhands/fetch...php?id=10313611. The score was adjusted by TD, but if you look, a ♠ lead turns out better than a ♥. A ♣ or ♦ works as well according to GIB, but these are unlogical leads imo.

I think it's true that the method works better at MP, because like I said on my blog, you seldom give away tricks, while leading your 5 card suit gives away a lot. However, at imps it's probably better to try and set the contract, and IF the 5 card suit is running, you better lead it. If you want to create a length trick in your 4 card suit, you'll still need some tricks at the side to defeat the contract, and this might explain why it's better to hope for the 5 card suit to come in.

Perhaps the right conclusion is that holding 5-4, the lead from the four card suit should be considered. Maybe chosen, maybe not. Like almost everything, hard to simulate.

In the free-jilly hand:

First, I would have led a heart also. So this is "after the fact" thinking, but it can be useful


Given the auction, partner opening 1D, 1N on the right, 3N on the left, opening leader is lucky to have an Axxxx in hearts. The ace will be a trick, led now or not, and the problem (at imps anyway) is to find four more. For the heart suit to make a contribution, there must be an entry. This can only be in hearts and requires partner to have three. It is possible to decide, perhaps, that it is better to try to build tricks in spades planning on the already available heart ace and four others elsewhere. (One of the elswheres being partner's heart king in this case).

This logic could change sharply if opening leader holds Axxxx of hearts and the ace of clubs, and hears an uncontested auction 1N-3N. Dummy can be expected to hold more minor suit cards than majors, probably they will need some clubs, and there is a decent shot at establishing hearts. Similarly with Kxxxx and the ace of clubs.

Simulating this sort of thing, where sometimes you lead from four and sometimes from five, seems to me to be very difficult. I suspect that it's the right answer however. Deciding when to lead from the five card suit is no doubt tricky, but I doubt you should wait for KQJxx at either imps or matchpoints.

If we take the fundamental point to be that you should not just look at your five card suit, find the fourth best, and lead it, this seems like very good advice.

In the free-jb hand I am virtually certain I would have led a heart but the logic for selecting a spade is substantial.

[pigpenz]

When I was still living in Omaha Mike Albert was talking about one of Soloways Rules and it was that he lead the four card suit not the five card suit....but that being 25 yr ago i dont remember if this applied to matchpoints or imps or both.

But i would tend to say the best thing to do is just crunch numbers with bridgebrowser or some deal making program.


obviuosly we have to look at the quality of the suits

[hotShot]

This is a very complex problem.

1) What deals to look at?
Those that where opened 1NTor any deal that ended in 3[NT].
2) How often do you hit partners double?
Obviously 8 cards given to 3 players averages to 2.66... so partner should have 3 cards about 2 of 3 times. So going for the 4 card suit is an antipercentage play for the 5 card suit.
It is logical that the quality of the 5 card suit is very important.
A weak suit hitting partners double will cost a lot, while with a strong suit partners double does not matter.
3) Of cause the combined quality of the 4 card suit is important. Playing a "weaker" 4 card suit, is in fact betting on partners unknown hand.
4) When developing a suit, the main question is: Do i have enough entries to win my tricks once the suit is developed. Obviously if the answer is no, developing the suit is a waste of tempo.

So if you don't have a good 5 card suit and have problems with entries, hitting partners double in your 5 card suit will be expensive. Leading your 4 card suit might be better.
With a strong suit and enough side entries, i don't think playing the 4 card suit will pay off.

[zasanya]

Jlall, on Oct 11 2006, 12:29 PM, said:

When I did lead my 4 card suit, I would have an excellent reason (like having QJT9 and xxxxx). Basically, when I lead a 4 card suit I am doing it for a very strong reason and expect it to be right a majority of the time. When I lead my 5 card suit it is often because I have nothing better to do, and think it's right to lead my long suit.

Could you give us any examples of either type from your/some other expert's tournament play?