Free's 3NT lead hypothesis

[inquiry]

To follow up on the post in response to Josh about 5NT contracts, a little more data. First, one argument sent to me was that the average BBO players really isn't that good. Well, I disagree in principle (and just because the average player is self-rated as expert or higher). But I addressed this issue in two different ways. First, I want to assure you that no matter what database you look at, 5NT on average is the worse contract (if we leave out specifically some doubled and redouble contracts).

 

Here is the result of a contract search in an OKBridge Tournament database, The assumption most people seem to believe is that since people pay to be a member of OKBridge, only serious players play there (it is worth noting that thousands of these serious bridge players have moved to BBO, disproving the general claim).

 

From the chart below, looking just at the row for 5NT when the contract is played from the opener's side, we see almost identical numbers to that found on the BBO except there is a much higher frequency of hands played at matchpoints as opposed to imps. In the bbo database, 5NT earned an average of -4.08, in okbridge it was slightly worse at -4.42. The average matchpoint result was essentially identical (36.36 versus 37.23 ).

 

http://1.forumer.com/uploads/homebaseclub/post-25-1160756639.jpg

 

I think this is beginning to demonstrate that, for other than josh, 5NT on the average is a bad spot to play. The graph above also gives you a feel for the relatively frequency of contracts from 4H to 5NT. Note that 4NT and 5NT are very rare (compare to 4H and 4S for example). The averages here, like in the earlier post, hides some of the good info you can get at. For example, you can click on 4S and see how 4S faired when opening side had at least a 9 card fit, or a 10, or you can look to see how playing with 6-5 distribution compared to playing with 5332 distribution or all other distributions. But what we are interested in, again, is why is 5NT so bad. We can click on 5NT and see how 5NT faired when it made on the noise (no overtricks, no under tricks), and compare that to the data for the bbo database both in terms of frequency and average imps and matchpoints (see chart below)

 

http://1.forumer.com/uploads/homebaseclub/post-25-1160756590.jpg

 

This chart looks sort of familiar. Look two post up to the same plot of tricks taken on BBO when playing in 5NT. On BBO, 5NT making on the noise earned 2.26 imps, on Okbridge in their tournament? 2.05 imps And on the two sites, the average MP for making were very close, 64.30 versus 64.25. How about percentage. As we can see of the 721 5NT contract on okbridge, 214 made on the noise, almost 30%.

 

Such data is reproducible across different databases from both sites.

 

So the next argument is, no doubt, that the problem is the people playing the hands were on average too poor a player. So I searched a larger okbridge database for all 5NT contacts and I FORCED declarer to have a lehman's rating of at least 55 lehman rating and a partner with at least a 52 lehman. First I looked at the result for all 5NT plays, then I looked at the results for players with the minimum requirements. Note how close the full dataset data is to that from the other two databases studied.

 

FULL 5NT DATA

http://1.forumer.com/uploads/homebaseclub/post-25-1160760543.jpg

 

Now the data with the high lehman requirement. While this data should be biased considerably, in that their average result will surely get better. To do this search, I created a board file with all the hands in which anyone played 5NT. There where 7779 such boards. Then I combined a number of searches (the blued tabs show which ones I combined... Player, where I made at least one partner have a 55 rating, the other at least 55, the contract tab, which I made 4H to 5NT so you could see how different games faired when.. NOTE these contacts are restricted to 7779 boards that someone played 5NT on, I used the view tab to be sure the averages to speed up the search, and the bid analysis tab to generate the table shown in the image....

 

http://1.forumer.com/uploads/homebaseclub/post-25-1160767252.jpg

 

So what you see that on these hands, 5NT was still the worse contract, compared to the other contracts at game but below slam. For all these contracts, the declaring side had a minimum of a 52 lehman (I used OKBRIDGE's precise ratings for these players) and the second player a min of 55. Since all the "weak" players had been weeded out, this is true of all contracts. The average is still horrible, -2.75 imps and below average, but both are better than the general population. If you examine why, you will discover that 5NT made on the noise a total of 270 times, and that is out of 728 contracts, or 37%. When it made on the noise, the average result was +2.88 imps and 66.07 MP. If you use excel to lower these to the percentage of making on the noise contract (throw out 20% of the good results), the data will overlap that when 5NT made for the general population. Thus, the reason the good players did better, on average, in 5NT is that more of them made it (up from 30 to 37%). It is worht noting that on these "good hands" on average, if the partnership stopped in anything game less than 5NT they would have done better. Thus, 5NT being the worse contract still holds.

 

What if we restrict both players to a lehman of 60? That search is doable, but of course you tremeddously limit the number of hand you find. As you can imagine, a rating of 60 is VERY VERY good, and catching two players together with a 60 rating is fairly rare. I think it is safe to say that players with a OKBridge Rating of 60 or more are very good players (note, using okbridge rating, those with 0 will not be shown, nor those with hiddent ratings, as I have set minimum and max ratings). There were only 41 contracts total in 5NT when 60+ played opposite 60+ out of 16,6 million contracts. 38 of which after they opened the bidding, with 13 at imps, for -0.43 and 25 at MP for average 41.77. This just shows that better players get better results in the same contract, a point that can be driven home by seeing how people with rating s from 25 to 40 faired in 5NT, but I didn't bother. THe point is, you can use soooo many differnt criteria, in choosing the population of hands. I am not sure what Josh's rating is to average 8 imps per board, but it must be something like a 300 lehman's (yes, I know it only goes to 75).

[Free]

His only bidding example is (his side silent):

1♠ - 2♣

2NT - 3NT

This is not my only example. I also mentioned a 6-4 and some other hands which I didn't define. We now have the hand I played with Jillybean, but as I said in my post, I don't keep a database on these hands. Tonight I had one in my local club and it didn't matter if I lead my 4- or 5-card suit: same result. But I still tested it (got a top, but that was because declarer blew his line of play, not because of the lead)...

 

I agree that it doesn't have to be 1NT-3NT auctions, that partner was usually silent (otherwise you get other lead directing info), both of leader's suits are unbid is probably right as well, and that the strength in both suits should be equal or better in the short suit. But not that both players had to be silent, and that both suits need to be unbid. Ofcourse, if opponents bid your short suit it's probably not such a good idea to still lead it, but if they bid your long suit it might still pay off to lead it anyway.

 

Don't really want to react to Ben's post, it's just ridiculous to attack me while I say the same things Justin said (and you agree with this, but not with my words), but a bit more direct. You just keep going with your BRBR, perhaps we'll get some useful info, perhaps not. Whatever you say, I still don't think this is a job for BRBR, and we both know we won't be able to convince each other to change ones mind. You don't need to explain yourself, I just don't care after such reply of yours...

[sfbp]

I average playing in 5N about 3-4 times a year, and win about 8 imps/bd on those hands (I am estimating since I haven't written down all my results anywhere). I think I have a statistical sample of 20 boards in real live bridge which isn't a huge sample. but its enough to know this isn't a fluke (I have played that contract much more since I started playing relay. It probably was only a once a year contract before that)...

How often have you** played 5NT in online bridge, Josh? How did it work out?

 

Stephen

 

**or of course, partner