% PBN 2.1
% EXPORT
%Content-type: text/x-pbn; charset=ISO-8859-1
%Creator: BridgeComposer Version 5.64.2
%Created: Sat Oct 21 17:32:50 2017 -0400
%BCOptions Center STBorder STShade
%BidAndCardSpacing Thin
%BoardsPerPage 1
%CardTableColors #e1e1e1,#ffffff,#aaaaaa
%EventSpacing 0
%Font:CardTable "Arial",11,400,0
%Font:Commentary "Times New Roman",12,400,0
%Font:Diagram "Times New Roman",12,400,0
%Font:Event "Times New Roman",12,400,0
%Font:FixedPitch "Lucida Console",10,400,0
%Font:HandRecord "Arial",11,400,0
%GutterSize 500,500
%HRTitleDate 2017.10.18
%HRTitleEvent "2017-10-18E"
%HRTitleSetID ""
%HRTitleSetIDPrefix ""
%HRTitleSite ""
%HtmlClubs entity,"http://bridgecomposer.com/suitimg/c.gif"
%HtmlDiamonds entity,"http://bridgecomposer.com/suitimg/d.gif"
%HtmlHearts entity,"http://bridgecomposer.com/suitimg/h.gif"
%HtmlNavBar 0.75,#cfe2f3
%HtmlSpades entity,"http://bridgecomposer.com/suitimg/s.gif"
%Margins 500,500,500,500
%PaperSize 1,2159,2794,2
%ParaIndent 0
%PipColors #000000,#ff0000,#ff0000,#000000
%PipFont "Symbol","Symbol",2,0xAA,0xA9,0xA8,0xA7
%ScoreTableColors #e6e6e6,#000000
%SelectedBoard 1
%ShowBoardLabels 2
%ShowCardTable 2
%TSTCustomSortOrder Default
%TSTReport List
%TSTReportOrder ByNumber
%TSTReportShade Yes
[Event "Which Major -- Problem"]
[Site ""]
[Date "2017.10.18"]
[Board "1"]
[West ""]
[North ""]
[East ""]
[South ""]
[Dealer "S"]
[Vulnerable "None"]
[Deal "S:AQ76.72.QT98.KJ8 JT3.JT854.65.942 K98.A93.K43.AQT7 542.KQ6.AJ72.653"]
[Scoring ""]
[Declarer "S"]
[Contract "3NT"]
[Result ""]
{

What would you lead, playing (a) matchpoints, or (b) IMP scoring?

Using methods similar to those employed by David Bird and Taf Anthias in their
ground-breaking book, <i>Winning Notrump Leads</i>, I analyzed this opening lead
situation. The authors make it clear that a major suit is a strong favorite
against this auction, so I considered only the standard leads of the \SJ and
\HJ.

<i>Please turn the page to read my conclusions.</i>

<b>Method of Analysis, for those who care:</b>

To start, I fired up Dealmaster Pro, entered the West hand, went into the
Simulator, and specified these parameters for the North and South hands:

-------- North Definitions --------\n
N1 Acc HP10-14       \S 2-3\H 2-3\D 2-5\C 2-5\n
N2 Acc HP 9-12        \S 1-3\H 1-3\D 6-13 \C 1-4\n
N3 Acc HP 9-12        \S 1-3\H 1-3\D 1-4\C 6-13\n
N4 Acc HP10-15 Bal \S 3-4 \H 3-4\D 3-3\C 3-3\n
-------- South Definitions --------\n
S1 Acc HP15-17 Bal \S2-4 \H 2-4\D 2-6\C 2-6\n
S2 Acc HP15-16 Bal \S 2-5 \H 2-3\D 2-3\C 2-3\n
S3 Acc HP15-16 Bal \S 2-3 \H 2-5\D 2-3\C 2-3

Bird & Anthias did not consider openings with 5-card majors, as I do.  

Using these specs, I dealt 500 hands, setting South as dealer on all hands. 
With text editors and tools on Cygwin, I prepared two copies of the deal file
for analysis, one where the SJ was always led, and one with the HJ led.  I
opened each file with a separate instance of Bridge Composer, which I then told
to "GIB Bid and Play All Boards."  (GiB is a separate product.  Bridge
Composer's double dummy analysis unfortunately ignores the contract and opening
lead.)  This is takes hours for 500 deals, so I let the two jobs run in
parallel, overnight.

Back in the morning, I used this pipeline on Cygwin to create a file with the
number of tricks taken for each deal in the Lead=SJ file, one per line:

grep Result *=SJ | tr -d "[a-zA-Z\\ \\"]" > SJ.txt

This takes each input line of the form "[Result "#"]" and produces a line in
SJ.txt containing "#".  (I used this method, so that the Windows CR of the
CR-LF pair would be retained at the end of each line.)

It was then easy to paste the results for each file into Excel, and calculate
the results given on the next page.}
[BCFlags "1f"]
[Hidden "NES"]
[Score ""]
[Auction "S"]
1NT {notrump opener -- 2-5 C; 2-5 D; 2-4 H; 2-4 S; 15-17 HCP}
Pass {No suitable call -- 20- total points}
3NT {2-5 C; 2-5 D; 2-4 H; 2-4 S; 10-13 total points}
AP

[Event "Which Major -- Solution"]
[Site ""]
[Date "2017.10.18"]
[Board "1"]
[West ""]
[North ""]
[East ""]
[South ""]
[Dealer "S"]
[Vulnerable "None"]
[Deal "S:AQ76.72.QT98.KJ8 JT3.JT854.65.942 K98.A93.K43.AQT7 542.KQ6.AJ72.653"]
[Scoring ""]
[Declarer "S"]
[Contract "3NT"]
[Result "12"]
{

<i>Sorry, the actual opponents were playing the 1NT opening as 12 - 14 HCP, which</i>
<i>I did not remember until I started entering the full deal above.  Bird and</i>
<i>Anthias say the strength of the opening bid makes little difference, to the</i>
<i>analysis.  If this bothers you, swap the \CA with the \CJ, for example.</i>

My lead of the \SJ was not a success; nor was partners return of a spade, after
getting in with the \DA.  But what was the "right" lead?  That is, what lead
should produce the best result over time?  Here are the results of my analysis:

          When the \HJ was led, declarer averaged 10.158 tricks, and was set 49
times.

          When the \SJ was led, declarer averaged 10.040 tricks, and was set 60
times.

At matchpoints, the \SJ saves 0.118 tricks per deal, on average, or a trick
every 8.47 deals.  How about an extra three tricks a session, if you had these
odds on each deal!

At IMPs, the \SJ sets the contract 60/500 times or 12% of the time.  The \HJ
achieves only 49/500, or 9.8%.  This 2.2% bulge on a full session is not as
impressive, setting slightly more than half an extra contract, on average, over
24 boards.  

As we all know, good luck trumps good play.  I still like the odds for the \SJ
on my side.

<i>- Pete Matthews, Jr.</i>}
[BCFlags "1f"]
[DoubleDummyTricks "ab8bbab8bb1151111511"]
[OptimumScore "NS 450"]
[Score ""]
[Auction "S"]
1NT {notrump opener -- 2-5 C; 2-5 D; 2-4 H; 2-4 S; 15-17 HCP}
Pass {No suitable call -- 20- total points}
3NT {2-5 C; 2-5 D; 2-4 H; 2-4 S; 10-13 total points}
AP
[Play "W"]
SJ
[OptimumResultTable "Declarer;Denomination\2R;Result\2R"]
N NT 10
N  S 11
N  H  8
N  D 11
N  C 11
S NT 10
S  S 11
S  H  8
S  D 11
S  C 11
E NT  1
E  S  1
E  H  5
E  D  1
E  C  1
W NT  1
W  S  1
W  H  5
W  D  1
W  C  1