[mikeh]

I found this thread interesting, but I agree that there are probably some self-filtering aspects to the sample group.

I was interested in another aspect.

My suspicion is that 1N openings gain their edge at mps on the partscore hands and at imps on the game hands, and was wondering if it is possible to check this?

My reasoning is as follows (more or less, I left out some permutations):

Opposite weak responding hands, 1N will be the final contract more than 1M: responders pass 1N with many hands that have 5-8 hcp with which they would respond to 1M. And the opps compete more over 1M than over 1N, especially in passout seat.

So on the weak responding hands, either opener will get too high (falling in love with his 16 count) or end up in an inferior spot at the 2-level or end up defending a contract on which opener's score is a minus or a smaller plus than was available in 1N.

Alternatively, on weak responding hands, opener in 1N scores 120 as against 110 in 3m or 2M, or scores +90 as against -50 or -100 or +80 etc.

At imps, some of this remains as a factor, but the 110 v 120 or 90 v 80 etc factor becomes essentially irrelevant, thus offsetting a significant element of the putative edge that the 1N gains at mps.

On game hands, the difference between 420 in a major 5-3 and 400 in notrump is minimal at imps. But, the difference between scoring 9 tricks in both denominations is huge. Furthermore, the concealment of a 5 card major (1N 3N) will frequently result in a lead into the major... often on the very hands that would defeat 4M and almost as often on hands which would also defeat 3N if the major had been disclosed (1M 2M 2N 3N etc).

Now, obviously, at both mps on the partscores and imps on the games, the 1M opening will sometimes triumph because the major fit, if it exists, offers declarer more tricks. And the ability to ruff will provide a stopper that is unavailable at notrump and so on.

There is also, at a high level of skill, a tiny edge to the 1M opener in that suit contracts generally offer more opportunity for technique: who ever heard of a trump squeeze in notrump? While many trump contracts can be reduced to essentially notrump endpositions, the converse is not true. And sometimes a side suit in dummy can be established via a ruff and so on. However, very few players would even recognize a trump squeeze, let alone pull one off so the technical edge is surely invisible in the sample under discussion.... but should it be ignored by those aspiring to play at the highest levels?

These are only the most apparent (to me) factors which seem likely to underlie the apparent success of the 5332 1N openers.... and I am beginning, as I am wont to do, to ramble, so I will stop there.

[1eyedjack]

awm, on Aug 9 2006, 06:45 PM, said:

Why do more people open 1NT on okbridge? Because okbridge generally charges for their services, there are fewer casual players online there. This makes the field (on average) stronger than it is on BBO.

I may have misread something, but I understood that while the numbers differed a bit, the broad trend within the OKB data was similar to that on BBO, ie a slight edge in favour of opening 1N. If that is the case, then the fact that OKB "filters out" the casual player MAY be a factor in favour of the proportion of players opening 1N but I do not see it as being a factor affecting the trend of results.

[kenberg]

The following came to mind as I was waking up. Suppose you ran the following study: You find, as you did, all the hands with a 5-3-3-2 distribution (5 in M) and a NT opening range. You find the folks who opened 1NT and those who opened 1M. You record the imp or mp swing. All this you did. Now you take each person that you have examined, and record the score he received on the next board he played. So: Joe opened 1NT on one of the samples and received a swing of x imps. You look at Joe's next board, whatever it is, and see that he received a swing of y imps. Compare the average x with the average y. This could give some evidence of the extent to which the positive average x is due to general superiority. It would not be conclusive proof (conclusive proof of anything tends to be elusive) since even if the average y was larger than the average x one could say, yes the players who open 1NT are indeed better players, and one aspect of that betterness is that they open 1NT. Still, like the original result, it would be interesting.


I can think of a theoretically better way to address the "it's just that they are better players" issue, but I doubt it's practical. On each hand you find, you look for every pair, Joe and Moe, such that Joe opens 1NT and Moe opens 1M. You record the swing x. You then scan their hand records to find, if you can, another hand played by both Joe and Moe but this time where the auction was precisely the same at each of their tables. Now you record the swing on that hand, call it z, and average the z scores. This should be a somewhat credible indicator of play advantage. Unless the average x exceeds the average z, it would be reasonable to expect the perceived advantage to the 1NT opening is simply the result of generally superior play.

[inquiry]

awm, on Aug 9 2006, 12:45 PM, said:

So the 1NT-opening crowd automatically filters out the weakest players.

Of course it's going to be hard to test this theory, but here's one simple check:

Stipulate that the 1♥ and 1NT openers both end up playing 4♥ by opener. .... How do they score? If the 1NT openers are (on average) the better players, you'll see them still scoring better despite reaching the same contract from the same side.

My feeling is that better players, on average, open 1NT with 5332 distribution and a five card major. That doesn’t mean you are a poor player if you open 1M, but rather, beginners will always open 1M (as a rule) so the population who open 1M include the weaker players. This is the essences of awn comment quoted above. and in fact, this was a leading possible explanation I proposed for the reproducible differences between opening 1M and 1NT with this pattern.

Awn suggest the following study. Find 4H contracts and determine if those who opened them 1NT or 1H did better on them. I am not certain this is the ideal way, but we will give it a go. The reason is, the way the data is collected. Let’s see how it works and try to decide if the collection method is “flawed”.

First, we run BRidgeBRowser to find all openers which match our requirements. To save time, I will do this for first seat only (thus an indexed search). Then, choose only the hands where opener side ended up in 4H, and run “bid analysis” on this. This will give you all opening bids and the result of those. Thus, those who opened 1NT or 1H will end up in 4H (as will those who opened, for instance 1C). Since all final contracts will be 4H, you can look and see how 1NT opening (ending in 4H) compared to 1H (and 1C opening for that matter).

For this initial study, I used a new database (not used in earlier studies). This was OKB database, but one of the several small ones (2.6 million hands). Hands that matched the initial search requirement (dealer holding 5332, 15-16 pts, 5 Hearts) that ended up in 4H totaled 1827 hands.

AT MP, (all final contracts = 4H)
Opening bid 1H = 46.89% (709), 1NT = 47.94% (137)

So for matchpoints, it seems there is little difference, but the “n” is small

At IMPs ((all final contracts = 4H)
1H = -0.04 (785), while 1NT = +0.71 (192), suggesting perhaps awn is correct. But how this doesn’t explain why the results are similar at MP but different at imps

For what it is worth, opening 1C that matched the requirements, occurred only 4 times, so no need to show those numbers, it can not be statistically valid. So we will have to bite the bullet and do this study on the larger databases. In the meanwhile, I have discovered from reading the help file and talking to Stephen Pickett there is a way to look up all 4H hands where opener opened 1H, and then, using only those hands, look up results for matching hands that opened 1NT. I will have to try this new search method as well.

Thus I tried this same study on the 23 million hand database (again looking only at the first seat opening, and to maximize the hands, I removed the "quality of the heart suit" from the equation. Any five hearts would do). The result of this larger study will be published in the next post, along with a an analysis of MikeH's propositio that partscores, on average, are better if you open 1NT and games are better if you open 1H. Another issue is how often you declarer the hand if you open 1H versus 1NT and how you do if you opponents buy the contract.

[mikeh]

Thanks, Ben. Actually my proposition was that 1N gains on both partscores and imps: but gains at partscores at mps, while probably being no better than break-even on game hands at mp; and gains on games at imps and does no better than break-even on partscore hands at imps.

The reason for the effect I suspect to exist lies in the scoring methods. I expect that 1N, on partscore hands, will result in a lot of tiny gains, often completely insignificant at imps but very important at mps, and a lesser number of fairly substantial losses.... 1N down 200 when 2M or 2m after a forcing 1N would be plus is a typical example.

6 hands on which 1N averaged .5 of an imp gain would be overwhelmed by 1 hand with a 5 imp loss at imps, but the 6 hands with a 1/4 board edge would be more than enough to offset the zero that -200 got once every 7 boards.

Conversely, I expect 1N to result in a significant number of large gains at imps: it is rare for a two balanced hands to make 4M and be unable to make 3N, but the converse is not true. Game hands are less frequent than partscore hands, even when we assume one hand has a 5332 15-16 count... we expect a mean of 8 points or so in responder, and most pass 1N with a balanced 8 count. So at mps, the partscore hands dominate by sheer number, while at imps, the game hands dominate by size of swing when swings occur.

[inquiry]

awm, on Aug 9 2006, 12:45 PM, said:

I suspect that most beginners, looking at a five-card major, would not consider opening 1NT. Stipulate that the 1♥ and 1NT openers both end up playing 4♥ by opener. Of course there is a slight difference in information content because the auctions are different, but I suspect this will matter a lot less when you reach 4♥ than other contracts. How do they score? If the 1NT openers are (on average) the better players, you'll see them still scoring better despite reaching the same contract from the same side.

This hypothesis, along with MikeH's were testable. I think the data disproves Mikeh's but leaves awn still quite viable.

The following table shows the results when openers (who opened 1H) played the following contracts. The first column shows the final contract, the second column the average Imps earned, the third column the number of hands played at imps, the fourth column the average MP result, and the last column, the number of mp contracts. The opening side played 37,236 contracts out of 48,216 times they opened 1H, meaning that they played 77.2% of the contract. They averaged minus 0.309 imps on the imp hands they played (16761 hands), and 48.65% on the MP hands they played (20,475 hands).

Using this same table, we can answer the first question Mikeh posed. How did game versus part score contracts compare. When opening 1H, the part-score average was -0.606 imps (5652 hands) and 48.87% (7242 hands). The game contract (3NT, 4H, 4S, 5C, 5D) averaged -0.097 imps (9448 hands) and 49.08% (11,235 hands).

Contract	AvIMP	[space][space][space]# Imp [space] [space] [space]AvMP	[space] [space] [space]#MP
1H[space] [space] [space] [space] [space]0.92[space] [space] [space] [space] [space]372[space] [space] [space] [space] [space]59.76[space] [space] [space] [space] [space]517
1S[space] [space] [space] [space] [space]-0.34[space] [space] [space] [space] [space]5[space] [space] [space] [space] [space]53.78[space] [space] [space] [space] [space]8
1NT[space] [space] [space] [space] [space]-0.44[space] [space] [space] [space] [space]1048[space] [space] [space] [space] [space]48.43[space] [space] [space] [space] [space]912
2C[space] [space] [space] [space] [space]-1.46[space] [space] [space] [space] [space]141[space] [space] [space] [space] [space]43.13[space] [space] [space] [space] [space]113
2D[space] [space] [space] [space] [space]-0.72[space] [space] [space] [space] [space]122[space] [space] [space] [space] [space]49.65[space] [space] [space] [space] [space]134
2H[space] [space] [space] [space] [space]-0.16[space] [space] [space] [space] [space]1150[space] [space] [space] [space] [space]52.57[space] [space] [space] [space] [space]1678
2S[space] [space] [space] [space] [space]-0.14[space] [space] [space] [space] [space]401[space] [space] [space] [space] [space]47.81[space] [space] [space] [space] [space]396
2NT[space] [space] [space] [space] [space]-1.58[space] [space] [space] [space] [space]490[space] [space] [space] [space] [space]42.2[space] [space] [space] [space] [space]730
3C[space] [space] [space] [space] [space]-1.41[space] [space] [space] [space] [space]157[space] [space] [space] [space] [space]41.93[space] [space] [space] [space] [space]157
3D[space] [space] [space] [space] [space]-2.38[space] [space] [space] [space] [space]88[space] [space] [space] [space] [space]46.22[space] [space] [space] [space] [space]165
3H[space] [space] [space] [space] [space]-0.84[space] [space] [space] [space] [space]1422[space] [space] [space] [space] [space]48.12[space] [space] [space] [space] [space]2128
3S[space] [space] [space] [space] [space]-1.28[space] [space] [space] [space] [space]185[space] [space] [space] [space] [space]41.47[space] [space] [space] [space] [space]219
3NT[space] [space] [space] [space] [space]0.37[space] [space] [space] [space] [space]3257[space] [space] [space] [space] [space]53.15[space] [space] [space] [space] [space]3970
4C[space] [space] [space] [space] [space]-2.09[space] [space] [space] [space] [space]44[space] [space] [space] [space] [space]28.42[space] [space] [space] [space] [space]30
4D[space] [space] [space] [space] [space]-1.53[space] [space] [space] [space] [space]27[space] [space] [space] [space] [space]43.06[space] [space] [space] [space] [space]55
4H[space] [space] [space] [space] [space]-0.31[space] [space] [space] [space] [space]4760[space] [space] [space] [space] [space]47.34[space] [space] [space] [space] [space]6074
4S[space] [space] [space] [space] [space]-0.33[space] [space] [space] [space] [space]1174[space] [space] [space] [space] [space]45.89[space] [space] [space] [space] [space]1022
4NT[space] [space] [space] [space] [space]-0.96[space] [space] [space] [space] [space]32[space] [space] [space] [space] [space]44.78[space] [space] [space] [space] [space]30
5C[space] [space] [space] [space] [space]-0.64[space] [space] [space] [space] [space]123[space] [space] [space] [space] [space]19.98[space] [space] [space] [space] [space]46
5D[space] [space] [space] [space] [space]-1.44[space] [space] [space] [space] [space]102[space] [space] [space] [space] [space]40.08[space] [space] [space] [space] [space]93
5H[space] [space] [space] [space] [space]-1.85[space] [space] [space] [space] [space]493[space] [space] [space] [space] [space]36.66[space] [space] [space] [space] [space]636
5S[space] [space] [space] [space] [space]-3.25[space] [space] [space] [space] [space]46[space] [space] [space] [space] [space]44.72[space] [space] [space] [space] [space]60
5NT[space] [space] [space] [space] [space]-7.92[space] [space] [space] [space] [space]10[space] [space] [space] [space] [space]23.26[space] [space] [space] [space] [space]18
6C[space] [space] [space] [space] [space]-1.52[space] [space] [space] [space] [space]40[space] [space] [space] [space] [space]63.66[space] [space] [space] [space] [space]53
6D[space] [space] [space] [space] [space]-1.47[space] [space] [space] [space] [space]71[space] [space] [space] [space] [space]36.94[space] [space] [space] [space] [space]28
6H[space] [space] [space] [space] [space]1.23[space] [space] [space] [space] [space]633[space] [space] [space] [space] [space]44.7[space] [space] [space] [space] [space]594
6S[space] [space] [space] [space] [space]-1.23[space] [space] [space] [space] [space]122[space] [space] [space] [space] [space]53.4[space] [space] [space] [space] [space]182
6NT[space] [space] [space] [space] [space]-0.87[space] [space] [space] [space] [space]180[space] [space] [space] [space] [space]54.2[space] [space] [space] [space] [space]288
7C[space] [space] [space] [space] [space]1.74[space] [space] [space] [space] [space]3[space] [space] [space] [space] [space]65.35[space] [space] [space] [space] [space]4
7D[space] [space] [space] [space] [space]-6.98[space] [space] [space] [space] [space]4[space] [space] [space] [space] [space]55.28[space] [space] [space] [space] [space]5
7H[space] [space] [space] [space] [space]0.69[space] [space] [space] [space] [space]32[space] [space] [space] [space] [space]57.93[space] [space] [space] [space] [space]36
7S[space] [space] [space] [space] [space]-6.75[space] [space] [space] [space] [space]7[space] [space] [space] [space] [space]61.25[space] [space] [space] [space] [space]32
7NT[space] [space] [space] [space] [space]2.19[space] [space] [space] [space] [space]20[space] [space] [space] [space] [space]56.41[space] [space] [space] [space] [space]62

For the record, the results when the Non-opening side (dealer opens 1H) played the contract was plus +0.187 imps, and a at 49.44% matchpoint score.

How did the contracts do if you opened 1NT? This is shown in the following table. Here we see that 4♥ contract (and 3NT contract) faired much better than the same contracts if you opened 1♥ (discussed more below the table). We can also see that the both the part-score and game contracts faired better if you opened 1NT than 1♥. For example, part-scores averaged 0.286 imps (3381 hands) after opening 1NT compared to a dismal -0.606 imps (5652 hands) if you open 1H, likewise, MP partscores average 53.03% (2577 hands) if you open 1NT compared to 48.87% (7242 hands) if you opened 1♥. So mikeh was dead on correct with this part of his prediction. However, games after opening 1NT averaged 0.406 (5463 hands) and 54.56% (4308 hands) compared to -0.097 imps (9448 hands) and 49.08% (11,235 hands) for all games after opening 1♥. I will not show the table for contracts defended, but when defending after opening 1NT, the opening side averaged +0.50 imps (1880 hands) and +52.13% (1350 hands), compare that to +0.187 imps, and a at 49.44% matchpoint score after opening 1♥.

Contract	AvIMP	[space][space][space]# Imp [space] [space] [space]AvMP	[space] [space] [space]#MP
1H[space] [space] [space] [space] [space]0[space] [space] [space] [space] [space]0[space] [space] [space] [space] [space]50[space] [space] [space] [space] [space]0
1S[space] [space] [space] [space] [space]0[space] [space] [space] [space] [space]0[space] [space] [space] [space] [space]50[space] [space] [space] [space] [space]0
1NT[space] [space] [space] [space] [space]0.65[space] [space] [space] [space] [space]1796[space] [space] [space] [space] [space]55.07[space] [space] [space] [space] [space]1132
2C[space] [space] [space] [space] [space]-1.17[space] [space] [space] [space] [space]6[space] [space] [space] [space] [space]28.29[space] [space] [space] [space] [space]2
2D[space] [space] [space] [space] [space]0.64[space] [space] [space] [space] [space]11[space] [space] [space] [space] [space]44.2[space] [space] [space] [space] [space]12
2H[space] [space] [space] [space] [space]0.04[space] [space] [space] [space] [space]262[space] [space] [space] [space] [space]55.44[space] [space] [space] [space] [space]268
2S[space] [space] [space] [space] [space]0.68[space] [space] [space] [space] [space]446[space] [space] [space] [space] [space]55.85[space] [space] [space] [space] [space]298
2NT[space] [space] [space] [space] [space]-0.93[space] [space] [space] [space] [space]218[space] [space] [space] [space] [space]48.28[space] [space] [space] [space] [space]235
3C[space] [space] [space] [space] [space]0.83[space] [space] [space] [space] [space]119[space] [space] [space] [space] [space]43.48[space] [space] [space] [space] [space]90
3D[space] [space] [space] [space] [space]-0.83[space] [space] [space] [space] [space]98[space] [space] [space] [space] [space]49.12[space] [space] [space] [space] [space]93
3H[space] [space] [space] [space] [space]-0.5[space] [space] [space] [space] [space]292[space] [space] [space] [space] [space]52.04[space] [space] [space] [space] [space]293
3S[space] [space] [space] [space] [space]-1.31[space] [space] [space] [space] [space]93[space] [space] [space] [space] [space]45.98[space] [space] [space] [space] [space]106
3NT[space] [space] [space] [space] [space]0.45[space] [space] [space] [space] [space]3072[space] [space] [space] [space] [space]57.84[space] [space] [space] [space] [space]2379
4C[space] [space] [space] [space] [space]-0.2[space] [space] [space] [space] [space]26[space] [space] [space] [space] [space]49.23[space] [space] [space] [space] [space]20
4D[space] [space] [space] [space] [space]-3.98[space] [space] [space] [space] [space]14[space] [space] [space] [space] [space]46.43[space] [space] [space] [space] [space]28
4H[space] [space] [space] [space] [space]0.26[space] [space] [space] [space] [space]1201[space] [space] [space] [space] [space]50.6[space] [space] [space] [space] [space]1132
4S[space] [space] [space] [space] [space]0.68[space] [space] [space] [space] [space]985[space] [space] [space] [space] [space]53[space] [space] [space] [space] [space]662
4NT[space] [space] [space] [space] [space]-1.19[space] [space] [space] [space] [space]38[space] [space] [space] [space] [space]51.61[space] [space] [space] [space] [space]53
5C[space] [space] [space] [space] [space]-0.45[space] [space] [space] [space] [space]79[space] [space] [space] [space] [space]24.65[space] [space] [space] [space] [space]29
5D[space] [space] [space] [space] [space]-0.83[space] [space] [space] [space] [space]88[space] [space] [space] [space] [space]30.57[space] [space] [space] [space] [space]53
5H[space] [space] [space] [space] [space]-1.97[space] [space] [space] [space] [space]100[space] [space] [space] [space] [space]43.96[space] [space] [space] [space] [space]97
5S[space] [space] [space] [space] [space]-1.2[space] [space] [space] [space] [space]38[space] [space] [space] [space] [space]42.67[space] [space] [space] [space] [space]39
5NT[space] [space] [space] [space] [space]-4.54[space] [space] [space] [space] [space]2[space] [space] [space] [space] [space]38.22[space] [space] [space] [space] [space]10
6C[space] [space] [space] [space] [space]1.73[space] [space] [space] [space] [space]66[space] [space] [space] [space] [space]57.82[space] [space] [space] [space] [space]37
6D[space] [space] [space] [space] [space]0.51[space] [space] [space] [space] [space]52[space] [space] [space] [space] [space]23.73[space] [space] [space] [space] [space]19
6H[space] [space] [space] [space] [space]1.63[space] [space] [space] [space] [space]243[space] [space] [space] [space] [space]50.9[space] [space] [space] [space] [space]128
6S[space] [space] [space] [space] [space]-2.07[space] [space] [space] [space] [space]117[space] [space] [space] [space] [space]61.16[space] [space] [space] [space] [space]137
6NT[space] [space] [space] [space] [space]1.33[space] [space] [space] [space] [space]136[space] [space] [space] [space] [space]54.35[space] [space] [space] [space] [space]228
7C[space] [space] [space] [space] [space]-15.51[space] [space] [space] [space] [space]1[space] [space] [space] [space] [space]21.09[space] [space] [space] [space] [space]1
7D[space] [space] [space] [space] [space]-3.7[space] [space] [space] [space] [space]4[space] [space] [space] [space] [space]50[space] [space] [space] [space] [space]0
7H[space] [space] [space] [space] [space]-10.79[space] [space] [space] [space] [space]7[space] [space] [space] [space] [space]50.57[space] [space] [space] [space] [space]9
7S[space] [space] [space] [space] [space]-2.01[space] [space] [space] [space] [space]12[space] [space] [space] [space] [space]62.92[space] [space] [space] [space] [space]24
7NT[space] [space] [space] [space] [space]-2.55[space] [space] [space] [space] [space]8[space] [space] [space] [space] [space]65.05[space] [space] [space] [space] [space]34

So MikeH's theory is not proven, opening 1NT seems to do better across the board at part-scores, games,

[inquiry]

(couldnt get this in the last post.. no idea why.. continued here)

games, and when defending. But the data lends support to awm;s theory that it is because there are better players. But is this really the cause? Let's take a closer look at the 4H contracts,

When you open 1♥, these are the statistics for the 4♥ contract

IMP HANDS
4760 [space] [space] [space]Tricks [space] [space]Score [space] [space] [space]IMP [space] [space] [space] [space]HCP [space] [space] [space] Trumps [space] [space]HP+Dist
Avge	10.01 [space] [space] 306 [space] [space] [space] -0.31 [space] [space] 24.89 [space] [space] 8.62 [space] [space] [space] 27.27
Std Err	0.01 [space] [space] [space] 5 [space] [space] [space] [space] [space] 0.06 [space] [space] [space]0.03 [space] [space] [space]0.01 [space] [space] [space] [space]0.03

MP hands
6074 [space] [space] [space]Tricks [space] [space] [space]Score [space] [space] MP % [space] [space] HCP [space] [space] [space] Trumps [space] [space]HP+DistHP+Dist[space]Avge [space] [space] [space]9.91 [space] [space] [space] [space]286 [space] [space] [space] 47.34 [space] [space] [space]24.86 [space] [space] 8.66 [space] [space] [space]27.34
Std Err	0.01 [space] [space] [space] [space] [space]4 [space] [space] [space] [space] [space] [space]0.34 [space] [space] [space] [space]0.03 [space] [space] [space] 0.01 [space] [space] [space] 0.03

When opening 1NT, these are the statistics for the 4♥ contract.

IMPS
1201 [space] [space] [space]Tricks [space] [space]Score [space] [space] [space]IMP [space] [space] [space] [space]HCP [space] [space] [space] Trumps [space] [space]HP+Dist
Avge [space] [space] [space] 10.37 [space] [space] 414 [space] [space] [space] [space]0.26 [space] [space] [space] 24.8 [space] [space] [space] 9.18 [space] [space] [space] 27.48
Std Err [space] [space]0.03 [space] [space] [space] 9 [space] [space] [space] [space] [space]0.11 [space] [space] [space]0.07 [space] [space] [space]0.02 [space] [space] [space]0.07
 [space] [space] [space]
Matchpoints [space] [space]
1132 [space] [space] [space]Tricks [space] [space] [space]Score [space] [space] MP % [space] [space] HCP [space] [space] [space] Trumps [space] [space]HP+Dist
Avge [space] [space] 10.22 [space] [space] 368 [space] [space] [space] [space] 50.6 [space] [space] [space]24.93 [space] [space] [space]9.11 [space] [space] [space]27.67
Std Err [space] [space]0.03 [space] [space] [space] [space]9 [space] [space] [space] [space] [space]0.74 [space] [space] [space]0.07 [space] [space] [space] 0.02 [space] [space] [space] 0.07

Now, there is something to note here: The hands that end up in 4♥ when you open 1NT are different from the hands that end up in 4♥ when you open 1♥. What is the difference? It is not hcp (averagignn aroudn 24.85 in all cases. It is different in the number of trumps held. Note, the ones that end up in 4♥ after opening 1NT has more trumps on average. And it is the extra trump that accounts for the additional trick. Let's look more closely.

The 1H--> end in 4H occurred 10834 times and average 8.64 trumps and 9.95 tricks. The 1NT-->end in 4H occurred 2333 times and averaged 9.15 trumps and 10.30 tricks. The 1NT-->4H players had 0.51 trumps more to play with and as a result, took 0.35 more tricks. In an earlier study, an extra trump was worth 1/3 of a trick on average. This clearly suggest the "extra" trick in hearts was due not to brilliance of the players but the presenses of more 5-4 fits (on average). This suggested the following study... same opener requirement, but require responder to have four hearts, (so trumps = 9 exactly) and see how the 1NT opener versus the 1H opener faired in 4H. Here are the stats from these two studies (not trumps now = 9.0 for all conditions).

Open 1H
1623 [space] [space] [space]Tricks [space] [space]Score [space] [space] [space]IMP [space] [space] [space] [space]HCP [space] [space] [space] Trumps [space] [space]HP+Dist
Avge [space] [space] [space]10.06 [space] [space] 347 [space] [space] [space] [space]0.07 [space] [space] [space]24.83 [space] [space] 9 [space] [space] [space] [space] [space] 27.2
Std Err [space] 0.03 [space] [space] [space] [space]8 [space] [space] [space] [space] [space] [space]0.1 [space] [space] [space] 0.07 [space] [space] [space] 0 [space] [space] [space] [space]0.07
 [space] [space] [space]
Unconstrained [space]Matchpoints [space] [space]
2340 [space] [space] [space]Tricks [space] [space] [space]Score [space] [space] MP % [space] [space] HCP [space] [space] [space] Trumps [space] [space]HP+Dist
Avge [space] [space] [space]10.21 [space] [space] [space] 355 [space] [space] [space]50.39 [space] [space] 24.93 [space] [space] 8.99 [space] [space] [space] [space]27.46
Std Err [space] [space]0.02 [space] [space] [space] [space] 6 [space] [space] [space] [space]0.52 [space] [space] [space] [space]0.05 [space] [space] [space]0 [space] [space] [space] [space] [space] [space]0.05

Open 1NT
626 [space] [space] [space]Tricks [space] [space]Score [space] [space] [space]IMP [space] [space] [space] [space]HCP [space] [space] [space] Trumps [space] [space]HP+Dist
Avge [space] [space] [space] [space]10.4 [space] [space] 430 [space] [space] [space] [space]0.39 [space] [space] [space]25.35 [space] [space] [space]9 [space] [space] [space] [space] [space] 27.82
Std Err [space] [space] [space] 0.04 [space] [space] [space] 11 [space] [space] [space] [space]0.16 [space] [space] [space]0.09 [space] [space] [space] 0 [space] [space] [space] [space] [space]0.09
 [space] [space] [space]
Matchpoints [space] [space]
684 [space] [space] Tricks [space] [space] [space]Score [space] [space] MP % [space] [space] HCP [space] [space] [space] Trumps [space] [space]HP+Dist
Avge [space] [space] 10.46 [space] [space] [space]421 [space] [space] [space] [space]52.24 [space] [space] [space]25.41 [space] [space] 9 [space] [space] [space] [space] [space]28.09
Std Err [space] [space]0.04 [space] [space] [space] 9 [space] [space] [space] [space] [space] 0.92 [space] [space] [space] [space]0.06 [space] [space] [space]0 [space] [space] [space] [space] [space] 0.06

Oddly enough, the results still favored the 1NT opening, but when we forced the heart legnth to be "correct", the 1H bidders bid to game on less values by about a half of a pont, which reduced their trick taking potential. So we force one more condition. We required the team total HCP to be exactly 25 hcp. Doing so, essentially gave the two sides the same "hands". The results show the 1H openers scoring better at both imps (.59 versus 0.24) and mp (53.29 versus 51.84).

Open 1H
205 [space] [space] [space]Tricks [space] [space]Score [space] [space] [space]IMP [space] [space] [space] [space]HCP [space] [space] [space] Trumps [space] [space]HP+Dist
Avge [space] [space] [space] [space]10.02 [space] [space] [space] [space]348 [space] [space] [space] [space]0.59 [space] [space] [space] [space]25 [space] [space] [space] [space]9 [space] [space] [space] [space]27.24
Std Err [space] [space] [space] [space]0.06 [space] [space] [space] [space]19 [space] [space] [space] [space]0.25 [space] [space] [space] [space]0 [space] [space] [space] [space]0 [space] [space] [space] [space]0.02
 [space] [space] [space]
Matchpoints [space] [space]
437 [space] [space] Tricks [space] [space] [space]Score [space] [space] MP % [space] [space] HCP [space] [space] [space] Trumps [space] [space]HP+Dist
Avge [space] [space] [space] [space]9.86 [space] [space] [space] [space]289 [space] [space] [space] [space]53.29 [space] [space] [space] [space]25 [space] [space] [space] [space]9 [space] [space] [space] [space]27.18
Std Err [space] [space] [space] [space]0.04 [space] [space] [space] [space]13 [space] [space] [space] [space]1.16 [space] [space] [space] [space]0 [space] [space] [space] [space]0 [space] [space] [space] [space]0.03
 [space] [space] [space]
Open 1NT
125 [space] [space] [space]Tricks [space] [space]Score [space] [space] [space]IMP [space] [space] [space] [space]HCP [space] [space] [space] Trumps [space] [space]HP+Dist
Avge [space] [space] [space] [space]10.28 [space] [space] [space] [space]408 [space] [space] [space] [space]0.24 [space] [space] [space] [space]25 [space] [space] [space] [space]9 [space] [space] [space] [space]27.24
Std Err [space] [space] [space] [space]0.08 [space] [space] [space] [space]25 [space] [space] [space] [space]0.35 [space] [space] [space] [space]0 [space] [space] [space] [space]0 [space] [space] [space] [space] [space] [space]0.03
 [space] [space] [space]
Unconstrained [space]Matchpoints [space] [space]
156 [space] [space] Tricks [space] [space] [space]Score [space] [space] MP % [space] [space] HCP [space] [space] [space] Trumps [space] [space]HP+Dist
Avge [space] [space] [space]10 [space] [space] [space] [space] [space]325 [space] [space] [space] [space]51.83 [space] [space] 25 [space] [space] [space] [space] 9 [space] [space] [space] [space] [space] [space] 27.3
Std Err [space] [space]0.07 [space] [space] [space] 23 [space] [space] [space] [space] 1.92 [space] [space] [space] [space]0 [space] [space] [space] [space] 0 [space] [space] [space] [space] [space] [space] 0.05

From this, I think we can propose a new theory that might begin to explain why the 1NT opening bids with these hands work better, even on the game hands. The tables above suggest that the difference in 4H is based upon both the better fit (9 more often) rather than difference in playing skill (see the last table). For this to be true, then a prediction would exist. With a 5332 in opener and a generally balanced hand opposite including 3 card support, 3NT tends wto be a better contract. With BRidgeBRowser, that proposition is testable too.

Ben

[sfbp]

A co-interesting point is raised at

http://www.microtopi...ridge/day2.html

Do you want to play in 3NT or 4H after a 2NT opener with a 5-3 fit? This study gives RESPONDER five hearts, but it raises some of the same points.

I guess the same sort of study might be done with a 1NT opener.

A couple of advantages I would propose for 1NT opening five card majors that weren't mentioned:

a. the better range of tools for the responder (eg puppet stayman, transfers)
b. the ability to play in RESPONDER's suit at the 2 level despite opener's 5-card major.

Stephen

[mikeh]

Interesting While I think that maybe Ben misunderstood my theory, his results prove that my theory (on either view of it) was wrong! Oh well.

I was very interested to see that opening 1N was even more effective at imps on partscore hands than it was on game hands: on partscores, 1N gained about .89 of an imp per board while on game hands it gained on average only .5 imps per board. My expectation was that the per board imp gain would be a lot more on game hands than on partscores, and the evidence is to the contrary.

And I had expected that the 1N edge at matchpoints would be higher for partscores than for games, and, again, I was dead wrong At mps, on games, the 1N had a 5.5% edge while on partscores it was only 4.16%.

This all seems counter-intuitive to me... and it may be that the sample results are distorted by a number of factors such as awn has proposed, but so far the results look compelling. Thanks again, Ben.

[inquiry]

sfbp, on Aug 10 2006, 11:15 AM, said:

A co-interesting point is raised at

http://www.microtopi...ridge/day2.html

Do you want to play in 3NT or 4H after a 2NT opener with a 5-3 fit? This study gives RESPONDER five hearts, but it raises some of the same points.

I guess the same sort of study might be done with a 1NT opener.

A couple of advantages I would propose for 1NT opening five card majors that weren't mentioned:

a. the better range of tools for the responder (eg puppet stayman, transfers)
b. the ability to play in RESPONDER's suit at the 2 level despite opener's 5-card major.

Stephen

Well, oddly enough, I doubt that puppet stayman helps much when compared to simply opening 1NT. Having said that, one place where opening 1H was a clear loser was when responder's initial bid was 3♥. I presume this is due to this frequenlty being a preemptive bid. Preempting when your partner is big and balanced is probably not a good idea.

Note, just as when you can compare the average result for a first bid (like 1H or 1NT) regardless of the auction, you can look at the result of intermediate bids, like responders bids. Both 2H and 3H raises of a 1H opening bid was surprising a loser. Also auctions that went 1H-4H accounts almost entirely for the lower point count for hands that ended up in 4H on nine card fits after opening 1H when compared to hands that opened 1NT. Those 4 card preemptive raises thus may have accounted for the most of the worsing results until we forced the same HCP total.

[inquiry]

An interesting paradox. In trying to decide how to pursue the question of 5332 one hand opposite a three card fit for the question of 3NT or 4M better, I tested some search strategies in a small database.....

Opening 1NT was again the clear winner (final results restricted to 3NT or 4H). And 3NT with the 5-3 major fit was a run-away winner over 4H (I forced responder to have a three card fit and no singleton or void). Opening 1NT, was a huge winner (without respect to rather 3NT or 4H was reached) compared to opening 1H. But when I looked at the tricks for opening 1NT versus 1H (without regard to the final contract of 4H or 1NT), those who opened 1H took more tricks on average!!!

Here is the results...

OPEN 1NT... 
222  Tricks  Score   IMP  HCP	Trumps	HP+Dist
Avge	9.22	420	1.16	25.48	8	26.76
Std Err	0.07	19	0.32	0.08	0	0.41
      
Unconstrained  Matchpoints    
146  Tricks  Score   MP %  HCP	Trumps	HP+Dist
Avge	9.47	395	55.35	25.96	8	28
Std Err	0.1	24	2.27	0.12	0	0


OPEN 1H
383  Tricks  Score   IMP  HCP	Trumps	HP+Dist
Avge	9.64	329	-0.2	25.17	8	26.81
Std Err	0.06	17	0.26	0.09	0	0.1
      
Unconstrained  Matchpoints    
347  Tricks  Score   MP %  HCP	Trumps	HP+Dist
Avge	9.72	261	47.07	25.63	8	27.37
Std Err	0.07	18	1.62	0.12	0	0.13 

The situation is, however, that if you open 1NT, you played 3NT more than 90% of the time, if you opened 1H you played 3NT only 18% of the time. The 0.3 to 0.4 extra trick was not as much value in most of the time when you opened 1H (Average score 261 and 329) compared to when you opened 1NT (average between 395 and 429). This clearly has to be related to the taking the 9 tricks on average in 3NT as compared to 4M.

But there was another paradox, one that requires futher thought before a design is set up. Playing 3NT after you open 1NT averaged 1.25 imps and 55,71%. Ok, normal enough. Playing 3NT after opening 1H, averaged 1.42 imps, but ony 48.14%. The question becomes why the worse average at MP? Does hiding openers 5 card major provide some advantage?

Do we think the advantage of 3NT over 4H (if it remains reporducible) with a 5332 pattern opposite a hand with no singleton or void will remain even if the 5332 hand is weaker and the responder hand is stronger? How should such a study to approached to avoid artificial bias as much as possible?

[inquiry]

Another approach... tried... I used the current bbo main room database, and search for the first 300,000 hands (all imps) with no selection criteria at all (any bid is fine). You may be interested in two findings on these hands. The first:

Quote

Opening
Bid      AveIMP    total hands
1C         0.03         33283
1D         -0.01         31319
1H         0.01         23071
1S         0.02         23728
1NT         0.1         17390
2C         0.1         3078
2D         0.25         2174
2H         0.07         3226
2S         0.22         3224
2NT         0.27         2566
3C         0.49         1138
3D         0.02         947
3H         0         851
3S         0.31         658
3NT         -1.12         171
4C         0.23         87
4D         -1.67         72
4H         0.69         596
4S         0.31         456
4NT         -1.16         25
5C         0.66         62
5D         -0.48         59
5H         -9.67         1
5S         -17.13         1
5NT         -12         1
6C         -3.63         2
6D         1.71         6
6H         6.62         3
6S         -6         4
6NT         0         0
7C         -16.53         1
7D         14.67         1
7H         5.8         2
7S         0         0
7NT         -18.86         2

A couple of issues, most "normal" opening bids quickly average to 0.00 plus or minus a few, see 1C, 1D, 1H, 1S above. Hands where people may, or may not bid (ike preempts) vary more, and of course, if we apply some selection factors, the values can be slanted as well. For these same 300,000 hands, I show what the different contracts averaged for the opening side (less than 300,000 obviously)...this data tells you what you already know, bidding game generally is worht more than "average" on line (see 4H, 4S, 3NT. And as bad as people think 2NT is (averages -1,25 wih lots of plays), 5NT is always a worse contract :-)

The same study can be done on millions of hands

Quote

Contract  AvIMP      # Imp
PO      -0.02      2069
1C      0.16      348
1D      -0.01      431
1H      0.47      613
1S      0.74      927
1NT      0.27      6475
2C      -0.41      1226
2D      -0.37      1980
2H      0.23      4013
2S      0.18      5445
2NT      -1.25      3787
3C      -0.8      2310
3D      -0.81      3081
3H      -1.05      3595
3S      -1      4229
3NT      0.75      24067
4C      -1.18      1139
4D      -1.45      1370
4H      0.32      14164
4S      0.27      15583
4NT      -1.36      358
5C      -1.45      2288
5D      -1.48      3009
5H      -1.79      1897
5S      -2.15      1545
5NT      -3.79      156
6C      -0.25      856
6D      0.24      1079
6H      -0.52      2007
6S      -0.72      2224
6NT      -1.28      1840
7C      -2.49      86
7D      -1.68      117
7H      -2.83      152
7S      -0.47      161
7NT      -1.49      260

Any way, having done this, I check to see what opening 1H or 1S on 5332 distribution averaged (without regards to hcp). The 1H opening with 5332 (a total of 5069 or 22% of the time. The average result of opening 1H with this pattern was -0.03, (sd of 0.03)and given the remarkably tight SD, this was "significantly" worse than opening 1He in general (0.04 with SD of 0).

what does the contract situation look like?

Quote

1H      0.2      101
1S      0.39      17
1NT      0.28      289
2C      -1.67      20
2D      -1.51      29
2H      0.41      363
2S      0.35      93
2NT      -1.57      155
3C      -1.78      24
3D      -1      43
3H      -1.1      332
3S      -1.75      54
3NT      0.51      602
4C      -2.12      11
4D      -1.89      12
4H      -0.08      1080
4S      0.15      171
4NT      -0.41      6
5C      -3.74      12
5D      -1.14      25
5H      -2.37      120
5S      -3.41      10
5NT      -7.58      5
6C      -0.69      7
6D      1.19      13
6H      0.87      124
6S      0.17      37
6NT      -1.78      38
7C      5.6      1
7D      5.4      2
7H      -0.32      6
7S      8.47      4
7NT      4.13      7

Notice 3H and espically 4H did not fair well (remenber in general games win about 0.35 imps on line). But 3NT did very well indeed. What factors play a role in these result. BRidgeBRowser has some plotting features, so lets look at the 4H contracts,,,, using it....

First of the 1080 4H contract, responder bid 4H at his first opportunity a total of 384 times (with or without interference). 81 of those times with 3 card support. Those averaged -0.19 imps (sd 0.65). Bid 4H with four hearts a total of 135 times, averaged a nice 0.43 imps for that bid, and bid with 5 or more hearts 147 times but averaged minus 0.27 imps. But here I have run into a wall. I wold like to look at all the 4H contracts on only 8 trumps and compare to all the 3NT contracts with 5-3 heart fit. But here, in this method, I have hit a wall (avoiding any selection other than 5-3-3-2 in opener). It would be nice if there was a plotting feature that showed the total number of cards in each suit split between the partnership when you reach a final contract. We will have to ask Stephen for that!!!

[inquiry]

Ok.. I have examined the question of 3NT or 4M another way. I set dealer up with the following requirements.

1) 5332
2) Five spades
3) 9 or more hcp
4) A team total of 21 to 29 hcp

In addition, I required his partner to hold 3 Spades and no singleton or void (so responser is 4333, 4432, 5332, 6322 in each case with 3 Spades).

That is it, then I ran the analysis. If the 5-3-3-2 opposite no "ruffing value" is right to play in NT (at least on average), this test should show it. For this study, there was no correction for opening bid what so ever. It could be 2S, 1S, pass, 1C, 1NT, 2C, what ever. To make it handable, of course, I ran the test with the hand with five spades being the dealer. The fact that the one player only has 3S should keep wild preemptive raise of 1S mostly out of the picture.

The not surprising result, given the earlier study on hearts, was:

3NT = average 0.93 +/- 0.07 imps (3565 hands), 55.64% +/- 0.59% MP (2966 hands)
4S = average -0.22 +/-0.04 imps (11,525 hands), 45.41 % +/- 0.26% MP (10,869 hands)

Of course, this study removed the major advantage of playing in spades, as it is difficult to ruff a card in the three card suit hand with this hand pattern as responder. So we will need to repeat the same search except with responder having 3 spades and either a singleton or a void (shapes like 5431, 6331, 7330, etc). The more shape responder has with his three card fit, obviously, the more useful the 8 card fit is (can anyway say I fought the law)? I will try to do this later...

Ben

[jtfanclub]

inquiry, on Aug 10 2006, 12:23 PM, said:

Opening 1NT was again the clear winner (final results restricted to 3NT or 4H). And 3NT with the 5-3 major fit was a run-away winner over 4H (I forced responder to have a three card fit and no singleton or void). Opening 1NT, was a huge winner (without respect to rather 3NT or 4H was reached) compared to opening 1H. But when I looked at the tricks for opening 1NT versus 1H (without regard to the final contract of 4H or 1NT), those who opened 1H took more tricks on average!!!

I don't claim to understand most of this, but for that last part, there seems to be a couple of possiblities...

1. People reach game more often after a 1NT opener.

2. People end up with the contract more often after a 1NT opener (ie., 1NT openers were more likely to win the contract).

I'd be willing to bet it's the 2nd one. Would you be willing to run a test to see how many times 1♥ and 1NT openers each ended with them winning the contract, and what the result was when they didn't?

Thanks.

[inquiry]

Follow up study... hand opposite the 5332 hand has three card support for spades, but also has a singleton or void somewhere.

Now, the results are as one might predict... playing 4S is much, much better than playing 3NT...

3NT = -0.69 imps (488 hands) and 45.55% MP (542 hands)
4S  = +0.37 imps (4401 hands) and 50.35% MP (3612)

The conclusion you might draw from this data is that if one hand is 5332 with a five card major, it is very important that the hand with 5332 be able to not only show the five card suit, but also express the "balanced" nature of the hand. The data also suggest that missing the 5-3 fit is not a bad thing when responder is relatively balanced with three card support, and if you are not worried about moysein fits, with an unbalanced responder hand yhou could still try for 4M (question this raises, how would the moysein fit do opposite an unbalanced responder?).

To take this study one step further, I made no restrictions on third seat, and had dealer hold 5S, 9 to 29 hcp, 5332 pattern, and dealer's partnership holds 21 to 29 combined hcp. There is no other restictions what so ever. Now you may have only 5 spades combined, or you may have 13. How did 3NT compare to 4♠?

3NT = 0.80 (5151 hands), 54.77% (4869 hands)
4S = -0.07 (18,185 hands), 46.93 (16,827)

Couple of issues.... remember from a few threads back, any random opening bid that ends in game wihtout some kind of restirction usually averages about 0.37 imps or so. Here 4S was both minus at imps and MP (less than 50%) when one hand held 5332 and five spades and the other hand any thing (but combined hcp 21 to 29). While 3NT was markedly better (0.80 and 54.77%).

This suggest two things.

1) don't worry so much about missing 5-3 major fit when yuou open 1NT (or 2NT) with 5332.
2) Maybe puppet stayman with balanced type hands (no singleton) is not so important opposite 2NT (or even 1NT).

I hadn't expected either of these findings when this started...

[glen]

inquiry, on Aug 11 2006, 01:21 AM, said:

This suggest two things.

1) don't worry so much about missing 5-3 major fit when yuou open 1NT (or 2NT) with 5332.
2) Maybe puppet stayman with balanced type hands (no singleton) is not so important opposite 2NT (or even 1NT).

I hadn't expected either of these findings when this started...

I'm not surprised by all this. Moving 5-3-3-2s with a 5cM into the bid handling that balanced range seems to have a small net gain, though many will remain unconvinced.

The counterpoint is that responder (i.e. the hand opposite the balanced hand type) must be able to show all singleton/voids on game going hands in order that the right game is reached. This is also criticial on hands without a five card major, as the short suit is the suit most often attacked by the defense.

[the hog]

From purely anecdotal evidence: I have played that all 5332 shapes in the appropriate range are opened with 1NT. We have hd very few problems with this, and the gains in the auction have been well worth it.
Ron

[1eyedjack]

inquiry, on Aug 11 2006, 02:21 AM, said:

The conclusion you might draw from this data is that if one hand is 5332 with a five card major, it is very important that the hand with 5332 be able to not only show the five card suit ...

Sorry, can you clarify please why the data shows it to be very important that the hand with 5332 be able to show the 5 card suit? I expect it is in there somewhere, but I was thinking that it might be sufficient that responder be able to show the singleton or void.

[inquiry]

1eyedjack, on Aug 11 2006, 12:48 AM, said:

Sorry, can you clarify please why the data shows it to be very important that the hand with 5332 be able to show the 5 card suit? I expect it is in there somewhere, but I was thinking that it might be sufficient that responder be able to show the singleton or void.

This was an extrapolation from the data that showed if partner had a short suit and 3 card support it was better to play in 4S than 3NT (the numbers were

3NT = -0.69 imps (488 hands) and 45.55% MP (542 hands)
4S  = +0.37 imps (4401 hands) and 50.35% MP (3612) )

Now I neglected to show the improvement for 2S contracts. It is better to play in 2S with 5332 opposite three with ruffing value than opposite 3 without ruffing values as well. But you are correct, in general, the study showed the factor in the decision process is responders hand not openers (although the effect of a small doubleton has not been looked at yet).

[pbleighton]

"I don't claim to understand most of this, but for that last part, there seems to be a couple of possiblities...

1. People reach game more often after a 1NT opener.

2. People end up with the contract more often after a 1NT opener (ie., 1NT openers were more likely to win the contract).

I'd be willing to bet it's the 2nd one."

Generally agree, but I'd rather say that opening 1NT gives you an advantage in competition. You can't make a meaningful takeout double after a NT opening, and overcalls are riskier. DONT, etc., are better than passing, but don't make up for the inherent slipperiness of bidding over a 1NT opening.

And even if the opps get the contract, it may be a disaster.

The NT opener's partner also knows what to do in a competitive auction better (usually) than the 1M opener's partner.

Peter