[inquiry]

Ok, I tried to ask the following question of bridge browser. What is the AVERAGE imp and Matchpoint result when a partnership with two players within a certain lehman range play together (defend or declearer). Unfortunately, I discoverd that the random search of hands is not totally random, it finds all the declarered ones first appartnetly, so the data I show is for just declared hands. The data shows:A nearly linear finding with regards to matchpoints and a non-linear fit for imps. The actual data was... To do this search, I used 5 point lehman ranges from 30 to 75, and plotted the data versus the middle point of each range. Also note, BRIDGEBROWSER treats anyone with a 0.00 lehman as hanving a 50.00 lehman. So I searched on 45 to 49.99 and from 50.01 to 55 to exclude these players of variable ability or hidden lehmans.

In the chart below, I allowed the size of the symbol at each value represent the uncertainty in the measurements. Since there are fewer partnerships on the ends of the spectrum, their average imps had larger standard errors (I could have run much longer searches to get smaller SE, but the values were changing very little and I grew tired).



This chart shows, first of all, that average lehman is not really 50.00, if one wants to believe the curve fit. The average imps for the 50.01 to 55 (plotted as 52.5) was 0.04, suggesting perhaps 50 should be very close to average. No doubt the curve fit is not accurate. A similar trend was observed in the matchpoint data (I did not bother to force teh symbols to represnet the standard error for mp, the trend is the aame.



Of course, you have to realize something. If two pairs of 65 down to play and they are evenly matched, they will earn an average of 0 imps and 50% matchpoints over a large number of boards. So drawing too many conclusions based upon average imps/MP earned is sort of silly. I suspect most truly great players only play against other very good players, which compresses the averages at the top. I suspect that bad players will have a hard time finding like minded soles, so they are playing "UP", and average players play against pairs both above them and below them in ability. Anyway, just thought this might interest somebody. I will test the defensive theory next.

[inquiry]

The normal meaning of 1H-Pass-3H

As you are probably aware, 1H-3H use to show a strong, GF raise. That has long been replaced by a "limit raise" showing, according to most authorities, something like 10-12 points (counting distribution) and generally 4 card support.

More recently, pairs have begun migrating to 1H-p-3H being a preemptive raise based upon four card support with a theoretical 0 to 7 hcp, perhaps coupled with Bergen Raises when holding limit raises. It is perhaps not surprising that the bidding on BBO shows that the limit raise is still by far the most popular, but the definition of "limit raise" has been some what altered from the 10-12 points, lets see how it goes.

What does bidding on BBO show? When players on the BBO with ratings that would equate to a lehmans of 58 or higher responded 3S, 66% of the time they held 10-12 points (counting distribution), and only 3.8% of the time held a hand with 15 points (Dist + hcp). These remainder of the time they held 9 or fewere distributional ponts (about 16% total) and 13 to 14 DP 24% of the time. Here is a chart of their point count for a 1H=P=3H bid.. Note the seldom had 13 hcp and with only three exceptions held l3 or less.



From this data, believing the "lehman type rating" and that 58 or higher must be a fairly good player, the standard raise 1M-3M still seems to fall in the 10=12 "point" range (66% in that range, and another 24% within two points one on top, one on bottom.... .. so if you extend this to 9 to 13 "points" you will account for 82% of all 1H=3H bids by players with a lehman of 58 or higher on the BBO

How about the data for all players (not separated by estimated lehmans? The data is shown below. First a table showing all the 1H-P-3H auctions...


Now a figure showing the average hcp (and DP in green). The lower panel exanded the range so you could see the bids with 0 and 1 hcp and 23 DP...



A couple of things right off the bat.. while the general trend is similar to that for the higher ranking players. Most 3H bids had 10 to 12 "points" (the green bars, the yellow ones are hcp). The order of the points was 11>12>10>13>9 for both the "experts" and total players. Suggesting that the general population is following closely the ranges played by the "experts." For what it is worth, I bid 3H with a hand considerably weaker than 11 DP, so my bids will be down there on the lower left hand side. The fact that there were bids with 24 hcp and similar high point count, however, that was missing from the expert auctions suggest that the strong jump raise is still alive and well. There is also a cute thing in the expanded chart. Two times 1H wss raised to 3H on 0 HCP and 0 DP. That means the raiser was 4333. Of those, one hand was xxx xxx xxx Txxx. the other time the raise held four hearts to ten. The numbers at the bottom of the blown up image show the average result for 1H-3H on 0 or 1 hcp. Curiously the average was positive, and even more curious the average for the the two highest point counts turned out to be negative (hugely so).

So it is safe to assume that the "average" 1H-p-3H bid on BBO will be on hands roughly equivlanent to limit raise. I was surprized that the preemptive raise was not more popular.

I looked at 100 of the weakest 1H-P-3H bids, none had more than 6 hcp. Of these bids, only 10 were alerted as either weak, preemptive, bergen or just alerted with no explaination (yes, BridgeBrowser stores alert information too). So it seems we have a way to go in getting people to alert their bids. I think all preemptive raises to the 3 level should be alerted, especially given the fact that 82% of all such bids fall into the normal limit raise ranges.

[hrothgar]

Hi Ben

Here's an example of the type of search that i would find especially interesting.

Choose one of BBO's name players like Fred or Papi or Jimmy Cayne or some such.
Ideally the player should be highly recognizable and play a single system. (Alternatively, you might be able to able to chose a star who only plays a single system with a given partner)

Chose one opening bid. (Perhaps a first seat 1NT opening. Better yet, a third seat, white versus red 1NT opening)

Provide a PDF graphing the strength of all the 1NT opening bids using a couple different metrics.

[inquiry]

hrothgar, on Oct 18 2006, 01:40 PM, said:

Hi Ben

Here's an example of the type of search that i would find especially interesting.

Choose one of BBO's name players like Fred or Papi or Jimmy Cayne or some such.
Ideally the player should be highly recognizable and play a single system. (Alternatively, you might be able to able to chose a star who only plays a single system with a given partner). He opened with 13 hcp 3 times, 18 hcp 4 times, and 19 hcp twice. You can click on the corresponding bars to call up those hands...

Chose one opening bid. (Perhaps a first seat 1NT opening. Better yet, a third seat, white versus red 1NT opening)

Provide a PDF graphing the strength of all the 1NT opening bids using a couple different metrics.

Sadly, Fred rarely plays so any one BBO database has too few of his hands. I did a follow up to the 1H-P-3H auction above, where the auction went 1H-P-4H...

I was surprised to find that 42% of the times the auction went 1H-4H, the 4H bidder had only 3 hearts!. I was also surprised to find that just under 50% of the time the responder had at least 11 hcp, and often much more. That was not divided by player skill level. So clamping the declarer down to 58+ lehman (estimated lehman), I found a clear splitting of the ways. The players with 58+ bid 4H on a 3 card suit, less than the general population (22% versus 42%). But still an alarming (to me) percent of the time. I wonder how many of the pairs bidding 4H on three cards were palying polish club or precision so that 1H was very limited.

One of the online database is hands played by Okbridge player papi (B Garozzo)... there are 8300 hands he played in this datafile (it includes the other tables when he is playing of course). Here is a list of all the hand he either opened (first column), overcalled (second column), responded (third column) or advanced (4th column).. that is, all the bids shown in this talbe are his, because I choose playere only (not player plus partner). If you look at the red box, it shows his average result at imps for each of his opening bids. With the exception of 3D, 4D and 5D, they are all positive imps (well 3C too, but that is essentially zero).. Some one needs to tell him to stop preempting in diamonds :-)



The next chart shows who he partnered with (and you could get his average result against any, or all opponents if you wanted. This table also generally shows the player lehman ratings at imps and at matchpoint. I have hidden those ratings.



If you click on any give partner, you will retrieve all the hands with that partner, you can also get several partners at once as shown in the next table, except here this is a table of all opponents that papi played against. I have highlighted four opponents and could pull up just the hands papi played against them...



Of course you asked about 1NT openings.. papi opened (not his partners, just him) 1NT a total of 587 times is this database. The follwoing chart shows his high card point range...


One 19 point hand was both vul, he was first seat and held AK98, AJ98, KQ5, Q2

The second 19 point hand he was vul first seat, and held AK54, A9, A432, A54 both times with dupont.

It is difficult to tell from the following distributional chart, but papi opened with a singleton three times. In each case he had a singleton King. Once with 5431 and twice with 4441. He had 17, 17 and 15 hcp when he did... the time with 15 hcp he was 5431 with five diamonds AKQxx.



The fact that you can click on any of these bars and easily review the ahnds, or save all the results in a board file for further probing (what did he do vul versus novul, or novul versus vul) makes this even more powerful... and wondering about those diamond preempts? you can pull up just his 3D, 4D, 5D opening bids very easily if you felt like it and examine what was going on, even to the level of suit quality....

[inquiry]

an added note.. when these hands were taken for the database, papi had a lehman of 63 and averaged 0.75 imps per board.. .just about dead on the estimate you would get for his lehman from his imp average in the first chart in the first post of this thread....