Programming SAS to Recognize Straights in Poker Sim

[kbry1180]

I am doing a poker simulation in SAS.  So far I have created a deck, shuffled and dealt 5 card hands (no draw), and kept track of pairs, trips, 4 of a kind, and flushes using proc freq.  Now I am trying to get SAS to recognize straights.  I have assigned numerical values to the card ranks (2=2, J=11, A=14, etc.).  My professor suggested using an Array to do this process and showed me a sample code.  The first part is the code I am using where I assigned numerical values.  The other data steps (data 1, 2, and 3) are the array.  Could someone offer suggestions as to how I would use the array with the code I am using?  

%let decksize=52;
%let handsize=5;
%let nsims=10000;
%let nhands=5;

data _null_;
ncards=&handsize*&nhands;
call symput("ncards", ncards);
run;

data deck;
call streaminit(&seed);
do suit="C","D","H","S";
do value="2","3","4","5","6","7","8","9","T","J","Q","K","A";
if value="A" then numval=14;
else if value="2" then numval=2;
else if value="3" then numval=3;
else if value="4" then numval=4;
else if value="5" then numval=5;
else if value="6" then numval=6;
else if value="7" then numval=7;
else if value="8" then numval=8;
else if value="9" then numval=9;
else if value="T" then numval=10;
else if value="J" then numval=11;
else if value="Q" then numval=12;
else numval=13;
 if suit="C" then suitval=1;
 else if suit="D" then suitval=2;
 else if suit="H" then suitval=3;
 else suitval=4;
order=rand("uniform");
output;
end;
end;
run;

*Array commands shown by professor;
data one;
  do i=2 to 6;
    if i^=6 then cards=i;
      else cards=6;
    output;
  end;
run;

data two;
  set one;
  array hand{5} c1-c5;
  retain hand;
  hand{_n_}=cards;
  if _n_=5;
run;  

data three;
  set two;
  straight=1;
  array hand{5} c1-c5;
  do i= 1 to 4;
    if hand{i+1}-hand{i}^=1 then do;
      card=i;
      straight=0;
      leave;
    end;
  end;  
  if card=4 then if hand{5}-hand{4}=9 then straight=1;
run;  

commands my professor showed me.  

[ballardw]

I don't know what that last line of "if card=4" is for.

This identifies straights using that approach.

data example;
   do c1= 2 to 14;
   do c2= 2 to 14;
   do c3= 2 to 14;
   do c4= 2 to 14;
   do c5= 2 to 14;
      output;
   end;
   end;
   end;
   end;
   end; 
run;
data three;
  set example;
  straight=1;
  array hand{5} c1-c5;
  call sortn(of hand(*));

  do i= 1 to 4;
    if hand{i+1}-hand{i}^=1 then do;
      card=i;
      straight=0;
      leave;
    end;
  end;  
run;  

but for further hands you'll need to do something to keep the suit with the rank.

[kbry1180]

Hello Ballard, 

The last line is accounting for Aces being valued 1 and 14.  So the way he wrote it was that whenever there was a sequence:  2-3-4-5-A, it would calculate that last difference as 9 and count it as a straight as well.  

Before seeing your reply, I ran a program that did not result in any errors or warnings, yet I am not quite sure it is doing what I intended i.e. identify straights.  Here is what I have so far:

%let seed=1210; 
%let decksize=52;
%let handsize=5;
%let nsims=10000;
%let nhands=10;

data _null_;
ncards=&handsize*&nhands;
call symput("ncards", ncards);
run;

data deck;
call streaminit(&seed);
do suit="C","D","H","S";
do value="2","3","4","5","6","7","8","9","T","J","Q","K","A";
if value="A" then numval=14;
else if value="2" then numval=2;
else if value="3" then numval=3;
else if value="4" then numval=4;
else if value="5" then numval=5;
else if value="6" then numval=6;
else if value="7" then numval=7;
else if value="8" then numval=8;
else if value="9" then numval=9;
else if value="T" then numval=10;
else if value="J" then numval=11;
else if value="Q" then numval=12;
else numval=13;
 if suit="C" then suitval=1;
 else if suit="D" then suitval=2;
 else if suit="H" then suitval=3;
 else suitval=4;
order=rand("uniform");
output;
end;
end;
run;

proc sort data=deck;
by order;

/*proc print data=deck;
run;*/

proc surveyselect data=deck noprint
method=srs
seed=&seed
sampsize=&ncards
rep=&nsims
out=cards;
run;

/*proc print data=cards;
run; */

data hands;
set cards;
hand=ceil(_n_/&handsize);
run;

data three;
  set hands;
  straight=1;
  array draw{&handsize} c1-c&handsize;
  do i= 1 to 4;
    if draw{i+1}-draw{i}^=1 then do;
      card=i;
      straight=0;
      leave;
    end;
  end;  
  if card=4 then if draw{5}-draw{4}=9 then straight=1;
run;  
      
  
/*proc print data=hands;
run;*/  

/*proc print data=three;
run;*/


proc freq data=hands noprint ;
tables value/out=freqval;
by hand;
run;

proc freq data=hands noprint;
tables suit/out=freqsui;
by hand;
run;

proc freq data=three noprint;
tables straight/out=freqstr;
by hand;
run; 

data repthree;
set freqstr;
simno=ceil(hand/&nhands);
run;

data rephand;
set freqval;
simno=ceil(hand/&nhands);

/*proc print data=repthree;
run;*/

proc means data=rephand noprint;
var count;
by simno;
output out=maxnum max=kind;
run;

/*proc freq data=maxnum;
tables kind;
run;*/

proc means data=rephand noprint;
var count;
by simno;
output out=win max=winkind;
run;

proc means data=repthree noprint;
var straight;
by simno;
output out=maxstraight max=winstraight;
run;

proc freq data=win;
tables winkind;
run;

data rephandsuit;
set freqsui;
simno=ceil(hands/&nhands);
run;

proc means data=rephandsuit noprint;
var count;
by simno;
output out=win max=winflush;
run;

/*proc print data=rephandsuit;
run;*/

proc freq data=win;
tables winflush;
run;

proc freq data=maxstraight;
tables winstraight;
run;

[kbry1180]

Quick follow up:

I did not get any error messages but the ouput for straights (and flushes) do not seem right.  For instance, it shows that a hand with 5 matching suits (flush) had a frequency of 100% and that hands that were not straights had a frequency of 100%.    So I either used the wrong code to track these hands or the output I created using proc freq is incorrect.  

[art297]

I would go about it slighly differently. I only took the following as far as checking for Straights. You can find more ideas for checking the other scenarios in the thread at: https://communities.sas.com/t5/SAS-Procedures/Allocating-Values-to-Poker-Hands/td-p/146459

Here is my suggested approach (Note: You'd have to create four subrankings based on suit and values. Thus a straight flush would actually have a range of 40 subranks to account for the possible range of values and suits. You could set them to have values of 9.01 to 9.40):

%let decksize=52;
%let handsize=5;
%let nsims=10000;
%let nhands=5;

/*missing the following*/
%let seed=0;

data _null_;
  ncards=&handsize*&nhands;
  call symput("ncards", ncards);
run;

data deck;
  call streaminit(&seed);
  length suit $1;
  do suit='C','D','H','S';
    do value=2 to 14;
      order=rand("uniform");
      output;
    end;
  end;
run;

proc sort data=deck;
  by order;
run;

data game (drop=value suit order);
  array handv(&handsize.);
  array hands(&handsize.) $;
  do hands=1 to &nhands.;
    do cards=1 to &handsize.;
      set deck;
      handv(cards)=value;
      hands(cards)=suit;
    end;
    call sortc(of hands(*));
    call sortn(of handv(*));

    /*create ranks based on*/
      /*9=Straight Flush*/
      /*8=Four of a Kind*/
      /*7=Full House*/
      /*6=Flush*/
      /*5=Straight*/
      /*4=Three of a kind*/
      /*3=Two pair*/
      /*2=Pair*/
      /*1=High card*/

    /* check for flush */
    if hands(1) eq hands(5) then rank=6;

    /* check for straight or straight flush */
    if handv(5)-handv(4) eq 1 and
       handv(4)-handv(3) eq 1 and
       handv(3)-handv(2) eq 1 and
       handv(2)-handv(1) eq 1 then do;
         if rank eq 6 then rank=9;
         else rank=5;
    end;
    else if handv(5) eq 14 then if handv(4)-handv(3) eq 1 and
            handv(3)-handv(2) eq 1 and
            handv(2)-1 eq 1 then do;
             if rank eq 6 then rank=9;
             else rank=5;
    end;
    /* check for four of a kind */
    else if (sum(of handv(*))-handv(1) eq (4*handv(5))) or
            (sum(of handv(*))-handv(5) eq (4*handv(1))) then rank=8;
  end;
run;

Art, CEO, AnalystFinder.com

[kbry1180]

Hello Art,

Thank you for your help.  I tried running your code and got some error messages.  Please excuse me if I forgot to do something that would account for these errors as I am new to SAS.

Here are the errors:

 90         data game (drop=value suit order);
 91           array handv(&handsize.);
 92           array hands(&handsize.) $;
 93           do hands=1 to &nhands.;
 ERROR: The variable type of hands is invalid in this context.
 93           do hands=1 to &nhands.;
                 _
                 133
 ERROR 133-185: A loop variable cannot be an array name or a character variable;  It must be a scalar numeric.
 
 ERROR: Illegal reference to the array hands.
 94             do cards=1 to &handsize.;
 95               set deck;
 96               handv(cards)=value;
 97               hands(cards)=suit;
 98             end;
 99             call sortc(of hands(*));
 100            call sortn(of handv(*));

[art297]

Sorry, that was an error on my part. I posted without testing .. something one should never do! Try the following instead:

%let decksize=52;
%let handsize=5;
%let nsims=10000;
%let nhands=5;

/*missing the following*/
%let seed=0;

data _null_;
  ncards=&handsize*&nhands;
  call symput("ncards", ncards);
run;

data deck;
  call streaminit(&seed);
  length suit $1;
  do suit='C','D','H','S';
    do value=2 to 14;
      order=rand("uniform");
      output;
    end;
  end;
run;

proc sort data=deck;
  by order;
run;

data game (drop=value suit order);
  array handv(&handsize.);
  array hands(&handsize.) $;
  /* deal for each player and score their hand */
  do hand=1 to &nhands.;
    do cards=1 to &handsize.;
      set deck;
      handv(cards)=value;
      hands(cards)=suit;
    end;
    call sortc(of hands(*));
    call sortn(of handv(*));

    /*create ranks based on*/
      /*9=Straight Flush*/
      /*8=Four of a Kind*/
      /*7=Full House*/
      /*6=Flush*/
      /*5=Straight*/
      /*4=Three of a kind*/
      /*3=Two pair*/
      /*2=Pair*/
      /*1=High card*/

    /* check for flush */
    if hands(1) eq hands(5) then rank=6;

    /* check for straight or straight flush */
    if handv(5)-handv(4) eq 1 and
       handv(4)-handv(3) eq 1 and
       handv(3)-handv(2) eq 1 and
       handv(2)-handv(1) eq 1 then do;
         if rank eq 6 then rank=9;
         else rank=5;
    end;
    else if handv(5) eq 14 then if handv(4)-handv(3) eq 1 and
            handv(3)-handv(2) eq 1 and
            handv(2)-1 eq 1 then do;
             if rank eq 6 then rank=9;
             else rank=5;
    end;
    /* check for four of a kind */
    else if (sum(of handv(*))-handv(1) eq (4*handv(5))) or
            (sum(of handv(*))-handv(5) eq (4*handv(1))) then rank=8;
    output;
  end;
  stop;
run;

Art, CEO, AnalystFinder.com

[kbry1180]

Thank you again Art.  

[art297]

Interestingly, this is the first time I've tried to write a program that ranks poker hands. My code isn't complete .. for one, it doesn't account for 2nd level ties (e.g.,  7 7 6 6 5 compared with 7 7 6 6 4).

However, in taking the code as far as I have, I discovered some things that you really ought to consider. Most importantly, ranking must be done using whole numbers .. otherwise you run into numeric precision issues. That is always a risk, with computers, attempting to do comparisons of calculations as if they were done using base 10 math.

Anyhow, here is what I've done so far:

%let decksize=52;
%let handsize=5;
%let nsims=10000;
%let nhands=5;

/*missing the following*/
%let seed=0;

data _null_;
  ncards=&handsize*&nhands;
  call symput("ncards", ncards);
run;

data deck;
  call streaminit(&seed);
  /*  set suit a card values*/
  /*  suits 1=clubs 2=diamonds 3=hearts 4=spades*/
  do suit=1 to 4;
    do value=2 to 14;
      order=rand("uniform");
      output;
    end;
  end;
run;

proc sort data=deck;
  by order;
run;

data hands (drop=value suit order hand cards);
  array handv(&handsize.);
  array hands(&handsize.);
  /* deal for each player */
  do hand=1 to &nhands.;
    do cards=1 to &handsize.;
      set deck;
      handv(cards)=value;
      hands(cards)=suit;
    end;
    call sortn(of hands(*));
    call sortn(of handv(*));
    output;
  end;
  stop;
run;

/*For testing purposes only: create test hands dataset
data hands;
  input handv1-handv5 hands1-hands5;
  cards;
2 3 4 5 14 1 1 1 1 1
10 11 12 13 14 1 1 1 1 1
9 10 11 12 13 4 4 4 4 4
8 10 11 12 13 4 4 4 4 4
7 9 9 9 9 1 1 2 3 4
4 5 6 7 8 1 1 2 3 4
2 3 4 4 4 1 1 2 3 4
5 6 6 6 7 1 2 2 3 4
8 9 9 9 10 10 2 2 3 3 4
5 6 7 7 7 1 2 3 3 4
2 5 5 6 6 2 2 3 3 4
2 3 4 4 5 1 1 1 2 3
9 10 11 12 14 1 2 2 3 4
4 5 6 8 9 1 1 1 2 3
;
*/

data game;
  set hands;
  array handv(&handsize.) handv1-handv5;
  array hands(&handsize.) hands1-hands5;

    /*create ranks based on*/
      /*900=Straight Flush*/
      /*800=Four of a Kind*/
      /*700=Full House*/
      /*600=Flush*/
      /*500=Straight*/
      /*400=Three of a kind*/
      /*300=Two pair*/
      /*200=Pair*/
      /*100=High card*/

  rank=0;
    /* check for flush */
  if hands(1) eq hands(5) then rank=600+handv(5);

    /* check for straight or straight flush */
  if (handv(5)-handv(4)) eq 1 and
     (handv(4)-handv(3)) eq 1 and
     (handv(3)-handv(2)) eq 1 and
     (handv(2)-handv(1)) eq 1 then do;
         if rank gt 600 then rank=900+handv(5);
         else rank=500+handv(5);
  end;
    /* check for an Ace thru 5 straight or straight flush*/
  else if handv(5) eq 14 and
          handv(4) eq 5 and
          handv(3) eq 4 and
          handv(2) eq 3 and
          handv(1) eq 2 then do;
           if rank gt 600 then rank=900+handv(4);
           else rank=500+handv(4);
  end;
    
  if rank lt 500 then do;
      /* check for four of a kind */
      if (sum(of handv(*))-handv(1)) eq (4*handv(5)) or
              (sum(of handv(*))-handv(5)) eq (4*handv(1)) then
              rank=800+handv(2);

      /* check for full house */
      else if (handv(1) eq handv(2) and handv(3)=handv(5)) or
              (handv(1)=handv(3) and handv(4)=handv(5)) then
              rank=700+handv(3);

      /* check for three of a kind */
      else if (sum(of handv(*))-handv(5)-handv(4)) eq (3*handv(1)) or
              (sum(of handv(*))-handv(1)-handv(5)) eq (3*handv(2)) or
              (sum(of handv(*))-handv(1)-handv(2)) eq (3*handv(3)) then
              rank=400+handv(3);

      /* check for two pairs */
      else if (handv(1) eq handv(2) and handv(3) eq handv(4)) or
           (handv(1) eq handv(2) and handv(4) eq handv(5)) or
           (handv(2) eq handv(3) and handv(4) eq handv(5)) then
              rank=300+handv(4);

      /* check for one pair */
      else if (handv(1) eq handv(2)) or
           (handv(2) eq handv(3)) or
           (handv(3) eq handv(4)) or
           (handv(4) eq handv(5)) then do;
        if handv(1) eq handv(2) then rank=200+handv(2);
        else if handv(2) eq handv(3) then rank=200+handv(3);
        else if handv(3) eq handv(4) then rank=200+handv(4);
        else if handv(4) eq handv(5) then rank=200+handv(5);
      end;

      /* check for high card */
      else rank=100+handv(4);
  end;
run;

Of course, the above isn't tested for accuracy .. that part is up to you.

Art, CEO, AnalystFinder.com