Reruns

[gnasher]

The program should normally generate its own seed, but it should also allow the operator to enter a seed himself. That protects you from operator errors without creating obstacles for someone who does need to recreate an old set.

[billw55]

Perhaps seeds should be obtained from random number suppliers such as hotbits. And of course, records should be kept of seeds that have been previously used. Each instance of the software should do so, and ACBL should require all used seeds to be submitted and kept in a central database.

[pran]

Perhaps seeds should be obtained from random number suppliers such as hotbits. And of course, records should be kept of seeds that have been previously used. Each instance of the software should do so, and ACBL should require all used seeds to be submitted and kept in a central database.

I really do not see any reason for keeping used seeds in any form of archive provided the seeds are created in some satisfactory manner.

 

In my own program the seed is a combination of the system clock (resolution 1 mS, range several thousand years) together with the time in milliseconds within 0,25 seconds (8 bits) for each key and/or mouse activity from the program is launched until the seed is actually needed. The result is a full 360 bits seed with no chance of duplication and the majority of which are completely random.

[barmar]

From the ACBL website:

Many computers use their internal clocks as a "random" starting value to assign each of the cards an "address." In practice, however, this method isn't sufficiently random enough. The software that ACBL uses to generate deals, therefore, requires an additional step. The staff person whose job it is to create the deals must physically deal out a set of 52 cards, just as you would do at the bridge table. The four hands are entered into the computer. This information, along with the time at which it was entered, create the starting value for the process and ensures that it is completely random.

[pran]

From the ACBL website:

I am extremely sorry to state so, but this reference reveals a major lack of understanding the concepts of randomness.

 

It is quite true that the system clock is in no way "random", but then neither is it not used that way. However, it is fully acceptable to use it as one of the sources for creating a starting seed to a (pseudo-)random number generator.

 

The important feature for a starting seed is that it shall never be possible to produce the same seed in separate runs.

 

Provided that the PRNG itself is satisfactory the difference between seeds is unimportant, they may differ in just one single bit or in as many bits as are convenient.

 

What is important to understand is that the suggested procedure of having someone manually shuffle a deck of cards and then enter the resulting card sequence as a starting seed is (like all human processes other than dice throwing) inferior to any decent computer seed generator.

[aguahombre]

What is important to understand is that the suggested procedure of having someone manually shuffle a deck of cards and then enter the resulting card sequence as a starting seed is (like all human processes other than dice throwing) inferior to any decent computer seed generator.

Yeh, it isn't random. But the combo of the shuffle and a clock or other computer seed generator should be enough to ensure no duplicate set, while just using a computer generation might be subject to a glitch.

[pran]

[...]But the combo of the shuffle and a clock or other computer seed generator should be enough to ensure no duplicate set, while just using a computer generation might be subject to a glitch.

If you look at my description (post #28) you might note that this is precisely my own procedure.

 

(And registering the 8 bits fraction within the quarter of a second for each keyboard or mouse activity should ensure that each such registered group of 8 bits contributed to the seed is truly random.)

[aguahombre]

Yes, yours gets it done. My intent was just that although not random by definition, the ACBL method doesn't suck.

[billw55]

Of course the ACBL method is OK, but it still makes me chuckle. Why go through such tedium when you could just roll a handful of d20. There must be many methods for humans to generate seeds which are less time consuming. I think they want to offer something familiar and comforting to the members, something the members will believe in, who are often suspicious of "computer hands".

[gnasher]

Does the ACBL procedure have to be followed for each set of boards, or is it just once a day?

[PhilKing]

Does the ACBL procedure have to be followed for each set of boards, or is it just once a day?

 

The latter - they use the time as a seed at exactly 6PM.

[pran]

The latter - they use the time as a seed at exactly 6PM.

Oh dear!

 

Is this so they can be perfectly sure that everybody use the same basic seed on the same day?

[RMB1]

Is this so they can be perfectly sure that everybody use the same basic seed on the same day?

Oh dear!

[bluejak]

I think people seem to miss the point. If the Correction Period is over, then the results stand. Law 6D2, like any other Law, is subject to the Correction Period. Of course the organisers should refund the entry fees to the OP, and anyone else who complains.