[Gerben47]

In the Reisinger Semi-Final at this moment, south in board 5 held:

---

And in board 6 south had a deja-vu as he held:

---

8 of the cards are the same as in the previous hand, and the singleton club 9 is duplicated. Now what are the odds to that?

[helene_t]

The point probability of 8 in the (13;52;13)-hyper-geometrical distribution is 0.0002. The recurence of the stiff club is not remarkable since two hands with 8 common cards usually have a similar pattern. So if you play 1000 boards a year (some 42 club evenings a 24 boards) this will happen once in five years to you.

[Antoine Fourrière]

It seems to me that there are as many chances to draw eight of your previous thirteen cards as there are to draw eight of any thirteen cards, such as eight of the thirteen spades, eight of the thirteen hearts, eight of the thirteen diamonds or eight of the thirteen clubs...

[Flame]

I think when cards are drawn by a computer the chances of getting the same cards are much bigger since the comp dealer isnt complete random.

[scoob]

Flame, on Nov 28 2004, 04:04 AM, said:

I think when cards are drawn by a computer the chances of getting the same cards are much bigger since the comp dealer isnt complete random.

they're not?

[PriorKnowledge]

Flame, on Nov 28 2004, 06:04 AM, said:

I think when cards are drawn by a computer the chances of getting the same cards are much bigger since the comp dealer isnt complete random.

Actually, it is exactly the opposite. Computer dealt hands are much more random than shuffled hands due to poor shuffling. The simplist computer algorithm I have ever seen does not repeat for 3,000,000,000 hands and most software uses a much better algorithm.

But since these hands were in separate decks, the point is mute. I'll take helene_t's word. Once every 5 years sounds about right.

[han]

It is true that computers are never completely random, but don't be mislead by this: they are so random that you will never know the difference while playing bridge.

Indeed, drawing cards by hand is far les random.

[joker_gib]

In general people think that hands are too wild when dealt by computer but in fact, they are too flat when dealt by hand !

[epeeist]

Computer deals (e.g. in tournaments) are, as a previous poster noted, generally more truly random because physical cards tend to be poorly shuffled. That assumes a proper method of computer dealing, of course.

The most recent issue of the Bridge Bulletin, in discussing bridge software, includes a note as to whether the program is capable of (i.e. might) deal any possible hand. I think that only "Jack" (?) was capable of dealing any possible hand (by which I mean, distribution of cards to all four hands), but I don't have the issue handy. I'm not sure what BBO's system does.

[hotShot]

The largest integer number a random generator on a 32 bit system can deliver is (2^31)-1. This is much smaller than the total number of possible deals at bridge.
So a lot of algorithms use more than one random number to shuffle the cards. But the random numbers a computer generates are not random, so there is a chance to get the same series of random numbers again, if the random generators seed is the same.
Since the seed is often an integer number, there are only (2^31)-1 possible seeds. Leading to only as many different deals.

So more has to be done to make sure any possible deal can be shuffled (easy to do) and with the same propabillity (very hard to do).

[PriorKnowledge]

hotShot, on Dec 3 2004, 08:23 AM, said:

The largest integer number a random generator on a 32 bit system can deliver is (2^31)-1. This is much smaller than the total number of possible deals at bridge.
So a lot of algorithms use more than one random number to shuffle the cards. But the random numbers a computer generates are not random, so there is a chance to get the same series of random numbers again, if the random generators seed is the same.
Since the seed is often an integer number, there are only (2^31)-1 possible seeds. Leading to only as many different deals.

So more has to be done to make sure any possible deal can be shuffled (easy to do) and with the same propabillity (very hard to do).

Actually, this is not correct. Just because a machine's architecture is 32-bit, does not mean that simulating longer bit integers is not possible. Many early generators used 64-bit seeds. The last I heard, 128-bit seeds were in common use.

[hotShot]

PriorKnowledge, on Dec 3 2004, 04:06 PM, said:

Actually, this is not correct. Just because a machine's architecture is 32-bit, does not mean that simulating longer bit integers is not possible. Many early generators used 64-bit seeds. The last I heard, 128-bit seeds were in common use.

Of cause it is possible to implement one, but the build in random generator is limited by the system and the seed is given as an argument. This argument must be in a native format, in this case a 4 byte integer.

This is why serious deal programs have to have their own random generator and a "handmade" library to handle very long integers.