BBO random deal generator

[Erkson]

I have been kibitzing a good number of boards. I am under the impression that the number of unusual distributions is much greater than normal.

 

IMHO the random generator of this program should be examined.

 

Erkson

[inquiry]

Actually, it is clear that the BBO doesn't use a random hand generator. My recollection of a long ago discussion, maybe in this forum somewhere or maybe in online chat, is that the algorithm is slanted to decrease the number of average hands. One thing for sure, I love the more than occassional 9 card suits, the 7-6 hands, etc. It makes the game more exciting IMHO.

 

Ben

[Erkson]

Hi, Ben, and many thanks for the explanation.

 

I would never prefer anything slanted :)

 

I just saw a perfectly called 7H going down 1 because of a 5-0 trump distribution.

http://bbo.bridgebase.com:81/perl/history....060432636-28560

 

It is not fair.

 

So goes the world.

 

The random generator is far too much slanted.

 

Best regards.

Erkson

[hrothgar]

Actually, it is clear that the BBO doesn't use a random hand generator. My recollection of a long ago discussion, maybe in this forum somewhere or maybe in online chat, is that the algorithm is slanted to decrease the number of average hands. One thing for sure, I love the more than occassional 9 card suits, the 7-6 hands, etc. It makes the game more exciting IMHO.

 

Ben

 

Where are you getting this information?

 

I talked to Fred about a related issue a couple years ago.

As I recall, he explicitly said that the the hand generator is designed for "fair" dealing and that this had been statitically validated.

 

Potentially you are confusing a related issue:

 

Hands that are dealt manually are often biased towards flat distributions.

This occurs because of flaws in manual shuffling techniques.

As a result, dealing programs often appear to be biased because their deals represent the "true" probability density function.

 

You are asserting that the dealer program has been deliberate modified to distort the hand generation PDF away from what is expected., which is a VERY different thing.

[keylime]

It takes 7 shuffles of a normal deck of cards to guarantee suffcient randomness within the 52 cards. Also, when the deck is not shuffled sufficiently the chances of freakness increases markedly. Something else to consider too is when you look at the hands at high level competitions, it seems that there are always those groups of hands that are deemed "wild".

 

Hand generators use normally some form of a random number generator. 3rd year computer science students write pretty good generators, so to inferred that BBO's is substandard in any way doesn't make a lot of sense. The programmers for our service do excellent work and the client has improved in many areas. I utterly fail to see the assertion that the hands are "rigged" in any way.

 

One last thought: in one of my mid-chart tourneys I had a hand that went like this in the bidding:

 

1S (3C) 4S (6H)

 

At nearly every table this sequence was duplicated, but at some tables there was a 6S sacrifice offered. This is an example of how proper hand evaluation and strategy can counter a 9 card suit.

 

Folks I'm en route to St. Louis as we speak -- will see all later today.

 

D.

[fred]

Actually, it is clear that the BBO doesn't use a random hand generator. My recollection of a long ago discussion, maybe in this forum somewhere or maybe in online chat, is that the algorithm is slanted to decrease the number of average hands. One thing for sure, I love the more than occassional 9 card suits, the 7-6 hands, etc. It makes the game more exciting IMHO.

 

Ben

 

The hand generator is certainly intended to be

completely random and the extensive statistical

tests that I have done indicate that it really is!

 

Fred Gitelman

Bridge Base Inc.

www.bridgebase.com

[hrothgar]

It takes 7 shuffles of a normal deck of cards to guarantee suffcient randomness within the 52 cards. Also, when the deck is not shuffled sufficiently the chances of freakness increases markedly.

 

I'm pretty sure that this is backwards

Insufficent shuffles lead to abnormally flat hands, not abnormally distributional ones.

[JRG]

It takes 7 shuffles of a normal deck of cards to guarantee suffcient randomness within the 52 cards. Also, when the deck is not shuffled sufficiently the chances of freakness increases markedly.

 

I'm pretty sure that this is backwards

Insufficent shuffles lead to abnormally flat hands, not abnormally distributional ones.

I remember, perhaps somewhat vaguely, the article about "seven shuffles". Perhaps because my first run at university was in the maths and sciences and I have always retained an interest in the mathematical.

 

Anyway, as I remember it, the study was that based on the kind of physical "shuffle" of the deck that most people do, it would take a MINIMUM of seven such shuffles to create (more or less) random deals.

 

I took this sufficiently to heart, that I often count how many times I'm shuffling when playing face-to-face (riffle ("one"), riffle ("two")...). I have got into the habit of (not sure of the name - "riffle"? -, but split the deck into roughly equal parts and then run thumbs along a corner of two halves, mixing the two parts together and then squaring up the deck) a couple of times and then shuffling a couple of times (i.e. take about 7/8 of deck off bottom in one hand, slide about 1/8th or so onto top of pile in other hand, repeat until whole deck back together). I do this until I've riffled the deck 7 times (which is more than what the formal mathematical study came up with).

 

Why the explanation? Well, this takes a bit of time (not much more than most players take, because I have fair dexterity and shuffle and deal quite fast), so I have time to watch the other players at the table shuffling the other boards (shuffling doesn't really require looking at my hands). I've noticed that "most" (i.e. a subjective "most") people only shuffle the deck about 3 or 4 times.

 

So I suspect (as I seem to remember the original article stated), that most bridge deals are not really "randomly" dealt. ...and most hands appear to be fairly "flat" (more so than seems (again subjectively) statistically correct).

 

For what it is worth!