Trying to get more precise number (Poker Hands Generator)

[kinghippo423]

Hello Everyone,

I did a poker program in Java that essencially finds the strenght of a poker hand created Randomly. My program is doing OK...but I'm pretty sure it can be optimised.

This is my results with 1 million hands:



Number of hands generated : 1000000
Straight Flush : 0.0012 % In theory : 0.0012%
Four Of A Kind : 0.017 % In theory : 0.0240%
Full House : 0.1535 % In theory : 0.1441%
Flush : 0.2285 % In theory : 0.1967%
Straight : 0.3104 % In theory : 0.3532%
Three Of A Kind : 2.2372 % In theory : 2.1128%
Deux Pairs : 4.6104 % In theory : 4.759%
Pair : 41.777 % In theory : 42.2569%
Nothing : 50.6648 % In theory : 50.157%

There is big differences between my results and the theory. I thought at first that it was because I didn't used enough hands.

This is my results with 50 million hands:



Number of hands generated : 50000000
Straight Flush : 0.0012 % In theory : 0.0012%
Four Of A Kind : 0.0183 % In theory : 0.0240%
Full House : 0.1489 % In theory : 0.1441%
Flush : 0.2286 % In theory : 0.1967%
Straight : 0.3086 % In theory : 0.3532%
Three Of A Kind : 2.2266 % In theory : 2.1128%
Deux Pairs : 4.6107 % In theory : 4.759%
Pair : 41.7652 % In theory : 42.2569%
Nothing : 50.6918 % In theory : 50.1570%

And again, the numbers are pretty much the same so it means that thereare flaws in my logic. The problem is I can't find those flaws. My question will be about my isFourOfAKind() method because I get approximalety 20% less Four Of A Kind hands then the theory. A pretty good difference in my opinion.

In general, my program is built like this

Card --> suit, strenght.
Hand --> Card1, Card2, Card3, Card4, Card5.

This is a representation of my method (Card1 = the strenght of Card1. In reality Card1 would be replace by Card1.getCardSt renght())

if(Card1 == Card2 && Card1 == Card3 && Card1 == Card4 && Card1 != Card5)
return true
else if(Card1 == Card2 && Card1 == Card3 && Card1 != Card4 && Card1 == Card5)
return true
if(Card1 == Card2 && Card1 != Card3 && Card1 == Card4 && Card1 == Card5)
return true
if(Card1 != Card2 && Card1 == Card3 && Card1 == Card4 && Card1 == Card5)
return true
else
return false

I'm pretty sure this logic is good. The problem is my results tells me that my logic is NOT good. Where is the flaw in that method?

Thanks in advance and pardon me for English mistakes...it isn't my first language

kinghippo423

P.S. I saw lots of poker hands generator like mice that works WAY faster than mine. Is there any website that explain some technics to have a "faster" program? It sucks to wait 1 minute ot generate 1 million hands...I hope I explain that well.

[Ganon11]

If you have translated your results correctly, then I see a 0.6% difference between your results and theory. It looks like your program is doing an adequate job of randomly generating hands - remember, nothing matches theory in practice.

[kinghippo423]

Theory almost never matched in practice.

By doing a test with 50 millions hands, I should have a really accurate result...specia ly with isFourOfAKind() because I have only 5 cases (XOOOO, OXOOO, OOXOO, OOOXO, OOOOX).

[JosAH]

Card --> suit, strenght.
Hand --> Card1, Card2, Card3, Card4, Card5.

This is a representation of my method (Card1 = the strenght of Card1. In reality Card1 would be replace by Card1.getCardSt renght())

if(Card1 == Card2 && Card1 == Card3 && Card1 == Card4 && Card1 != Card5)
return true
else if(Card1 == Card2 && Card1 == Card3 && Card1 != Card4 && Card1 == Card5)
return true
if(Card1 == Card2 && Card1 != Card3 && Card1 == Card4 && Card1 == Card5)
return true
if(Card1 != Card2 && Card1 == Card3 && Card1 == Card4 && Card1 == Card5)
return true
else
return false
 

What if a four of a kind are to be found in card#2, card#3, card#4 and card#5?
I'm about sure that the above logic is incomplete.

kind regards,

Jos

[Nepomuk]

What if a four of a kind are to be found in card#2, card#3, card#4 and card#5?
I'm about sure that the above logic is incomplete.

True, you said yourself that there are 5 possibilities, but you only check 4! Try to figure out, which one is missing. ^^

Greetings,
Nepomuk

[JosAH]

True, you said yourself that there are 5 possibilities, but you only check 4! Try to figure out, which one is missing. ^^

Greetings,
Nepomuk

Yep, and it statistically answers the question that the four of a kind score is
about 20% too low. Problem solved ;-)

kind regards,

Jos

[kinghippo423]

I was probably smooking that day. I correct that really dumb answer and this is my results after 50 millions hands. The results are as bad but reversely:

Nombre de mains généré : 50000000
Straight Flush : 0.0011 % In theory : 0.0012%
Four Of A Kind : 0.0354 % In theory : 0.0240%
Full House : 0.1419 % In theory : 0.1441%
Flush : 0.2264 % In theory : 0.1967%
Straight : 0.3069 % In theory : 0.3532%
Three Of A Kind : 2.1676 % In theory : 2.1128%
Deux Pairs : 4.5608 % In theory : 4.759%
Pair : 41.5571 % In theory : 42.2569%
C'est rien : 51.003 % In theory : 50.1570%

As you can see, now after 50 millions hands the difference between results and theory are at 0.0114%. That's almost 50% more F-O-A-K than theory. Now I know that my logic is correct thanks to all of you. But still, I'm still with that problem.

This is my new question. Throughtout my program, I check the strenght of the hand by checking all the possibilities. For example, for a pair there is 10 possibilities : XXOOO, XOXOO, XOOXO, XOOOX, OXXOO, OXOXO, OXOOX, OOXXO, OOXOX, OOOXX. I double checked my isPair() method and the logic is perfect. But still, I'm almost a 1% off the theory and I'm using 50 millions hands and that'a a lot.

Is it the optimal way to check the strenght of a poker hand. Is there anyone here that used a different approch? I'm not trying to get a easy answer from all of you (because I will probably re-do my progran with piles and other concepts and I learned in the last year) but you open my mind because in my mind, that's the only way to check the strenght of a poker hand...is by checking every possibilities. My method for my isTwoPairs() method is huge because there is 30 way to manipulate XXOOJ (J in that case is a non-full-house card) without duplication.

My method is excellent? Great? Good 'nouf? Please come again?

BTW, the random Java Class is trustable? I'm starting to suspect that something isn't right with that class.

Oh and By the Way again, what is exactly the seed in the Random Class? I randomize the seed for better result but is it a good way to do it? Should I keep the seed fixed?

Thanks you all agai for the advises.

[Ganon11]

Random number generators are, in fact, rarely Random. What they do is reference a very long list of integers that are 'random' - a portion of this list may be:

...324845, 348723, 876334, 3984, 4358,...

The seed tells the generator where along the list to begin. If you keep the seed constant, you will be getting the same 'random' number every time, so randomizing the seed is essential.

[kinghippo423]

The seed tells the generator where along the list to begin. If you keep the seed constant, you will be getting the same 'random' number every time, so randomizing the seed is essential.

But Randomizing the seed doesn't makes the Random numbers really Random though because Math.Random() will never be "Pure" Random. Am I correct?

P.S. Maybe too much Random in that sentence :)

[JosAH]

My method is excellent? Great? Good 'nouf? Please come again?

Thanks you all agai for the advises.

Mwah, I'd call it so-so. I once did it like this: create two char arrays, one for the
ranks (strengths) and one for the suits:

char[4] suits= new char[4];
char[15] ranks= new char[15];
Arrays.fill(suits, '0');
Arrays.fill(ranks, '0');
 

Note the fifteen elements in the ranks array; the first two elements are unused
because there are no 0 and 1 cards; aces are high (14) and all cards have a
rank equal to their, ahem, rank.

For five cards in a hand update the two arrays:

for (Card card: hand) {
   suits[card.getSuit()]++;
   ranks[card.getRank()]++;
}
 

... and convert the arrays to Strings:

String s= new String(suits);
String r= new String(rank);
 

And here comes the fun: use regular expressions for finding the score:
Four of a kind is found as follows:

boolean fook= r.matches("0*40*");
 

A full house is found like this:

boolean fh= r.matches("0*20*30*") || r.matches("0*30*20*");
 

Checking whether or not all cards have the same suit is easy too:

boolean ss= s.matches("0*40*");
 

I hope you get the picture.

kind regards,

Jos