[sfbp]

It's all coming back to me.

If we believe in Pearson points, something magic should happen at 15 PP. Anyway everyone seems to agree that the 15PP hands should be opened.

So here are the interesting cases which are only 14PP

14 points and no spades
13 points and 1 spade
12 points and 2 spades
11 points and 3 spades

Clearly the last two are the only critical ones... most people can see the sense of opening a 13 count no matter what.

Just for fun I looked at 11 points and 2 spades, and 11 points and 1 spade, as well as 12 points and 1 spade.

Anyway here is the chart for 11 points and exactly 3 spades.



These numbers (for opening 1C, 1D, 1H) are hugely in favour of opening. Of course you don't open 1S. Half an imp or 5 MP% is a big margin in favour of opening.

Just in case you are wondering (I was) the 20.71 over on the right is the average pointcount held by one side. Not sure which as it depends on where the mouse was, and I wasnt looking at the time.

I don't show the variances but as expected with these large frequencies, they are negligible.

I'll try 12 points and 2 spades next.

Stephen

[sfbp]

Here's the chart for 12 points and 2 spades



It's even worse passing here... -0.79 imps is a huge loss. A full 1 imp or 7 MP% worse than opening.

To put this in perspective, you should look at all standard, agreed actions by the majority should always end up 0.0 imps or 50%.

Stephen

[Miron]

I've read this thread. And if I got it right:
On the fourth seath open, even weak hands without spades.
Is it this way?
Thanx

[sfbp]

Quote

I've read this thread. And if I got it right:
On the fourth seath open, even weak hands without spades.
Is it this way?
Thanx

Essentially what the data appear to say is that you should use whatever other criteria you use to decide whether to open normally.

I may be able to follow up with some more data (if everyone isn't bored and overwhelmed already) but my sense of it is this: if you have more points than average (10), BID, with the exception I noted higher up the thread for balanced hands. One thing that set me thinking about this was a single incident where I played against one of our better players locally who opened last seat on a nondescript NINE count, and she got an excellent matchpoint score. I would personally never do this since I open all 12 counts in 1st, 2nd and 3rd seats, whereas that player might have been allowing for her partner passing a 12-count.

I think there might be a "Pearson Effect" but it certainly doesn't indicate you need FIFTEEN Pearson points. Opening 4441 11 counts (with a spade singleton) is probably not a good move, but I wouldnt put it any stronger than that, without further data.

Perhaps it's time for you or some others to get wet feet

Stephen

[inquiry]

Miron, on Apr 28 2006, 02:50 AM, said:

I've read this thread. And if I got it right:
On the fourth seath open, even weak hands without spades.
Is it this way?
Thanx

The short answer is, yes, this data suggest opening light hands in 4th seat rather or not you hold some number of spades.

The long answer has to do with being able to read the data on that table shown on the Bid Analysis tool (the graphic in the post above yours) can be difficult to grasp at first. If you look at it you will see a table, the first column of which is a labelled "bid" just above the white area... from Pass then dbl, rdbl, 1C, 1D, 1H all the way to 7NT. The next six columns are for when someone "bids". The second column is "labelled" open (for opener). Here SFBP forced the opener to open with pass, so it should probably be labelled dealer as that is the case in this example. You will see that pass by dealer (with the other requirements of his search) totalled 28,858. And you will see no dealer opened 1♣ (the 0 in the row labeled 1♣ under the "open" column, and similar 0's for all other bids (other than pass). This is because the search criteria FORCED delearer to pass.

The third and 4th columns are for hand after dealer and then dealer's partner. Again the search criteria only found hands where these hands passed. So it is the same 28,858 hands and they all pass (in addition to the first three hands passing, the fourh hand had to hold EXACTLY 2 spades, and EXACTLY 12 hcp... so out of 23 million hands, the auction went pass=pass=pass 28,858 times WHEN the 4th hand had both 12 hcp and 2 spades).

What the first three bids where was controlled by the search parameters, they had to be pass of the hands were not "found". However, what the fourth hand bids with his 12 hcp and 2♠'s was not controlled. So you see some people "passed" (in fact, 5947 people passed the hand out). But now you see some people opened 1♣ (in fact, 7,333 times), some 1♦ (9,799), etc. Even one person opened 1♠ on his doubleton ♠.

Now what isn't obvious in this table at first gkance is how to tell what different bids (left hand row) in different seats (top of the columns) earned. Above the columns is a "slider". Basically this is an arrow that points to one column of the other (it is in the narrow white area above the headers to each column), Currently the slider is pointing at "Adv", Adv stands for "advancer". In bidding, the terms opener, overcaller, responder, advancer refer to the four seats in this example, opener means "dealer" so advancer (adv) means 4th seat. With the slider pointing to advancer, it means the numbers in "aveMP" (average mp score) and "aveIMP" (meaning average imp score) is what the the bid displayed for "advancer" earned. In the rows next to average imps and average MP is rows labeled "# MP" and "# IMP", this referres to the number of hands figuring into the average imps and average mp hands.

So for example, only one person opened 1♠ on in 4th seat on the doubleton spade. IF we look down the row entitled "ave IMP", we see the opening of 1♠ averaged -1.70 imps, and we see 1 in the number of imp hands that were opened 1♠ (that was this one psyche or mositio opening), For average mp, we see 50.00 but we also see there where no MP hands where 1♠ was opened, explainig the 50.00 (you get 0 average for imps if no hands, and 50.00 for mp if no hands).

Having seen how this works, you can look at hands where the 4th seat passed... you see out of nearly 6000 hands (5947 to be exact) the results for passing out was not very good. This hand pattern (12 hcp, 2 spades) was passed out at imps a total of 3,179 times and passing earned (on average), -0.79 imps. At matchpoints, pass out occured 2768 times, for an average of 44.41 matchpoints.

By comparison, you can look to see how bidding did, by opening bid, opening 1C, 1D, 1H, 1NT all did better than passing. The full table also show the averages for opening 2C throuigh 7NT.

A few other things can be done, the slider can be moved over the columns that say "CtrOS" and "CtrNO". These stand for "contract opening side", and "contract non-opening side". If you move the slider, you the averages in the average imp and average MP columns will change to correspond with where the slider is.

In this example, opening side was the "dealer" so we see a large number of contracts for the "non-opening side" compared to the opening side, for instance, 1NT was played 2460 times by 4th seats side compared to only 155 times by the dealer side (after starting with 3 passes). We can not see from the "static" table (the screen capture) how well or poorly 1NT did. The number shown in the column are for 1NT opening bids by 4th seat with 12 hcp. However, using bridgebrowser you could move the slider over the CtrNO and then read off how 1NT did. If you had been able to move the slider over, you would have seen that 1NT averaged 0.65 imps and 58.98 matchpoints for the 4th seat side. If the opener's side played 1NT, they did much worse, averaging 0.09 imps and 49.55 matchpoints.

You can go further is you want. You can figure out, using other bridge browser tools such as the "hand stats" function for example that 4th seat playing in 1NT averaged 7.17 tricks at imps, and 7.36 tricks at MP and that their average hcp total was 21.15 and 21.09 when they played in exactly 1NT. Or you can figure out how many hcp they averaged as a team when they took 6 tricks, 7 trick, 8 tricks, 9 tricks. etc

6 tricks = (494 times) 20.60 and 20.79
7 tricks = (768 times) 21.01 and 21.03
8 tricks = (655 times) 21.69 and 21.27
9 tricks = (399 times) 21.92 and 21.47

The actual range of tricks playing in 1NT after opening in 4th seat with 12 hcp and 2 spades was from 1 tricks to 11 tricks. Obviously taking as few as 1 to 4 or as many as 11 were very rare events. Playing in 1NT 4th seat took only 1 trick once (at matchpoints), took 2 tricks only once (at imps), took 3 tricks only 7 times, and took 11 tricks only 7 times. So these extremes were only 0.6% (less than 1%) off all the 1NT contracts.

Or you can examine individual hands, for example if you "clck" on that "one" for the hand that opened 1♠ in fourth chair it will call up not only that hand where the 1♠ was bid, but also all the duplicates of that hands where some other (more reasonable) bid was made and inspect each one if you like.

Ben

[inquiry]

sfbp, on Apr 28 2006, 08:48 AM, said:

I think there might be a "Pearson Effect" but it certainly doesn't indicate you need FIFTEEN Pearson points. Opening 4441 11 counts (with a spade singleton) is probably not a good move, but I wouldnt put it any stronger than that, without further data.

Actually my study examined "distibution" in addition to just number of spades and HCP, and I looked specifically at 1444 hands with singleton spades with 13, 12, and 11 hcp.

With a singleton spade and 12 hcp (13 pearson points), it was "better" to bid with 5431 patterns than 1444 patterns. Passing with 5431 earned -0.94 imps ((+/- 0.13) and 38.88% (+/- 0.33); while passing with 4441 pattern cost much less --- only -0.18 (+/- 0.19) at imps, but 43.62 (+/- 1.92) at matchpoints.

[sfbp]

Ben:

I suppose I meant that I wanted to see the same results within the framework of the methodology I myself had set up.

Since I open all hands with 11 points and a 5-card suit, that hand type never even occurred to me as a candidate for passing. Duh! It should have.....

I'm looking at the overall picture for 11 points and 1 spade right now, and overwhelmingly it suggests bidding.

11 points and 2 spades wasnt so clear, interestingly.

If we actually believe this (and the variations don't have quite enough data for me to be sure about this just yet) then my conclusion is the whole Pearson thing is a chimera.

Stephen

[mike777]

What hands are you assuming Partner in second seat is passing on?
1) Is partner in second seat opening all 11 hcp and many unbalanced 10 hcp hands?
2) Is partner only opening Roth/Stone style in second seat?

[inquiry]

mike777, on Apr 28 2006, 10:25 AM, said:

What hands are you assuming Partner in second seat is passing on?
1) Is partner in second seat opening all 11 hcp and many unbalanced 10 hcp hands?
2) Is partner only opening Roth/Stone style in second seat?

For the study I did I listed my assumptions... I restricted 1st, 2nd, and 3rd chair to no more than 12 hcp, and then assigned a specific number of HCP and distirbution (with specific number of spades specified...so some were 5431 with 1 spade and some where 5431 with 3 spades for example, I treated these differently).

I am not certain what constraints stephen used other than three passes up to the fourth chair. You could, if you want, place extra strong or lax standards on the 2nd seat. For example if you open any 11, you could exclude all hands with 11+ hcp in second seat. If you open any 10 with five card major, you could exclude that in 2nd seat as well.

[hotShot]

Didn't you restrict to bidding to: pass-pass-pass ?

So a weak 2 made in any seat, would have eliminated the hand?

[inquiry]

There are other ways to appoach this problem. One is to simply hold the HCP and the distribution constant, and then to change the the legnth of the spade suit. I tried this out using 4th seat 5431 and exactly 11 hcp. The average results for passing these hands out were...

     IMPS     MP's
1♠   -0.77   43.97
3♠   -0.79   43.50
4♠   -1.71   49.11
5♠   -0.97   34.55

For this study, I change to a dataset with "only" 16.9 million hand plays, just to change the souce data. A few things to note.

1) There was no difference between passing with 1 or 3 ♠ and 11 hcp and 5431 distribution, and passing was bad in both cases.

2) Passing with 4 or 5 ♠ and 11 hcp gave some odd data. First, at MP, passing with 4 was "safer" than passing with 1 or 3 spades. Second bidding with 4♠ was better at imps than bidding with 5♠'s. It should be noted that the few people who choose to pass with 11 hcp and five spades so the SD is very high making the absolute numbers somewhat questionable.

Using the same dataset, I examine 5422 distributions and 11 hcp

     IMPS     MP's
2♠   -0.25    49.63
4♠   -0.57    49.98
5♠   -0.05    41.35

More people passed with 2♠ than with 5♠ by a lot.

[awm]

One thing that might be interesting is to restrict the search to balanced hands. I think you excluded 5-5 hands the first time, but even a 5431 11-count is a "normal opening" in most people's methods. It seems pretty clear that "blindly following Pearson points is a bad idea" but most of us knew that anyway. Perhaps a more interesting claim is "if you have a hand that you would not open in 1st/2nd seat, then you should use Pearson points to decide."

An easy way to address this might be to look at balanced 11s and see if having more spades makes opening more favorable. In fact, if you restrict my pattern to some (4432) and 11 hcp, it would be somewhat surprising to me if opening the hands with four spades wasn't better than opening the hands without.

[inquiry]

You appear to be at least half right (for imps at least). Again using the constraints earlier and the 16.9 milliion hand database, I searched for hands with 4432 and 11 hcp, only changing by number of ♠'s.

Pass earned

-0.01 and 48.30 with two spades
-0.29 and 48.35 with three spades
-0.47 and 48.08 with four spades

The data suggest that there is no real difference at matchpoints between passing at bidding with 4432 and 11 chp (all low 48% with standard deviatiions close enough to suggest these are not statistically different). However, at imps, the difference beween pass and bidding based upon the number of spades appears to be significant. Of course, I only used 16 million hands out of a pool of 100 million, so the data could be expanded to confirm this, and 10 hcp with four and five spades versus short spades could be tested as well.

[han]

I have been critical when people (imo) jumped to conclusions, but I must admit that the data in this thread is very interesting and impressive.

Ben, can you suggest an explanation for why the behavior we see at IMPs is so different from the behavior at MPs?

[Al_U_Card]

1 matchpoint can be 10 score points (620 vs 630) whereas 1 imp is 20 score points etc. based on the imp scale. Small differences in matchpoints carry larger importance, whereas larger differences in imps carry larger importance. Is that not at least part of the reason?

[sfbp]

Quote

For the study I did I listed my assumptions... I restricted 1st, 2nd, and 3rd chair to no more than 12 hcp, and then assigned a specific number of HCP and distirbution (with specific number of spades specified...so some were 5431 with 1 spade and some where 5431 with 3 spades for example, I treated these differently).

The main difference I can see between my data process and ben's is that I guaranteed that at least SOMEONE passed out every one of the deals I looked at.

So I specifically exclude ALL hands where the opening fourth seat is obvious to everyone. It's possible there are rare glitches, of course, and Ben is correct, I should constrain the first three bids to exclude Moscito, Strong Pass and that famous system MisClick. But there are lots of people passing 12 and even 13 counts in 1st and second seats, and I don't want to exclude those hands in any way.

If the software had more sophisticated criteria I could look for hands where at least TWO tables passed it out

Stephen

[inquiry]

Hannie, on Apr 28 2006, 12:09 PM, said:

I have been critical when people (imo) jumped to conclusions, but I must admit that the data in this thread is very interesting and impressive.

Ben, can you suggest an explanation for why the behavior we see at IMPs is so different from the behavior at MPs?

Short answer, I don't know. But in general what I see is what one would expect after opening light in the 4th chair: if the side that opens gets to play the contract below 2NT they pretty much average a plus score, and any plus score is good at matchpoints and imps. There biggest problem is if they overbid. If the dealer side plays the contract, the 4th seat side generally does well too.

Let's take, for example opeing 4th seat with 11 hcp and 4 ♠. 4th seat passed these out 41.7% of the time and got worse than average (). They bid the remaining 57.3% of the time and averaged good results. In fact, let's futher subdivide the hands where 4th seat bid, into those where 4th seat played the hands (4980 times, or 48.1%), and where opener's side played the hands (1051 or 10.2%). In both cases, the 4th seat side got the better of both the matchpoint and imp score, despite the number of tricks being won by both sides is roughly the same when the bought the contract. When 4th seat declared, they average winning 7.91 tricks at imps and 7.76 tricks at mp and average (in the aggregate) +0.18 imps and 52.16 MP. On the other hand, if 4th seat didn't pass, and dealer's side played the hand they averaged just about the same number of tricks (7.79 at imps. and 7.83 at MP) but they scored much less, -0.38 imps and 48.46 MP.

However, if you multiply the number of tables, by the scores, they don't come to zero (close, but not zero). For instance for imps, in this dataset there where 6136 imp hands. Multiply the average imps won by the number of board then total that and divide by the number of hands, the average imps per board (which should be zero) is -0.06. This means as expected, results at other talbles that did not start with 3 passes have not been factored in. And indeed those hand are not returned with this search, but they affect the calculated scores obviously.

[hotShot]

Can it be that 4th seat openings are more often played in a major, while dealers side defend in a minor?
This would explain why the same amount of tricks is worth less.

[Al_U_Card]

hotShot, on Apr 28 2006, 01:14 PM, said:

Can it be that 4th seat openings are more often played in a major, while dealers side defend in a minor?
This would explain why the same amount of tricks is worth less.

Shouldn't affect the matchpoint results as much as the Imp result, I would think. Even tho the "rank" factor would have to be considered. (to play in diamonds against 2H you must be in 3....etc.)

[mike777]

So if I understand some conclusions drawn from this data are:
1) 15 Pearson pts=open
2) 11+ hcp with a stiff or void=open
3) pass the rest?