cards
bridgetalk.com
Issues for the bridge community


Pages: (7) [1] 2 3 ... Last »  ( Go to first unread post ) addREPLY  newTOPIC  newPOLL  

> Computer generated hands (UK club bridge)
BruceR
Posted since your last visit  Posted: Jan 21 2004, 12:16 PM

bridgetalk member
*

Group: Members
Posts: 1,615
Joined: 17 October 03
Location: UK




When surfing bridge club web sites I came across the following statement (regarding mostly, I think, the club's bi-weekly open pairs level 3 evenings)

"To ensure randomness the hands are dealt by computer"

Returning to duplicate Bridge after a 20+ years absence, this statement was new to me. After 25+ years in commercial software houses, I cannot say the above quote filled me with confidence. On the other hand, I can see the great advantage of already having a record of what the hands are.

Are there any EBU guidelines on this ?
I quickly scanned the ebu site and the Orange and White books, but found nothing.
whereagles
Posted since your last visit  Posted: Jan 21 2004, 12:41 PM

bridgetalk member
*

Group: Members
Posts: 482
Joined: 3 December 03
Location: Portugal



It is known that hand shuffling may not be 100% random. The patterns you get on a new deal may be reflective of what happened in the deal before (thus the old saying "the queen follows the jack").

There are ways to manually shuffle hands that generate good enough random deals. For instance: shuffle 1-by-1 (like an automatic dealing machine does), repeat 7 times, cut, deal 1-by-1. In practice, players tend to be lazy and they don't shuffle the cards enough times for the outcome to be random to a good enough approximation. This is why some people prefer computer-generated hands.

Now, I know of some problems regarding computer randomization. That could cause problems as well. But the programs I've seen so far do a good job of presenting a random outcome, so I trust them. I compared the results with the theoretical odds and got an almost perfect fit.

Some people have prejudices against computer-generated hands. They believe that computer hands are more freakish and harder to bid. Actually, I have seen such a thing happening in a major championship: roughly half the hands were crazy fits, the other half total misfits. At the time I began to wonder if this was normal, and concluded that those hands were indeed somewhat far from the theoretical odds. Rather than blaming the computer, I found a much much simpler reason: the organizers sorted out the hands so that they could have a nice show in the VU-graph!
bluejak
Posted since your last visit  Posted: Jan 21 2004, 03:05 PM

Forum Host
*

Group: IBLF
Posts: 5,984
Joined: 28 August 03
Location: Liverpool UK



It is great to have an excuse when your play is not good enough to win an event, thus there are always some complaints. rolleyes.gif

But in practice computer-generated hands have been proved again and again and again to be fair. Hand dealt ones tend to be flatter because of insufficient shuffling.

There have been many major championships with quite boring and flat hands. However, when there is one that is freakish, people suggest the organisers have picked the hands. They do not.

I have been involved at many levels in organising bridge. The only events where hands are ever picked in any way are Novice Events, and Simultaneous Pairs, where sometimes the commmentators are given some choice. Not all, by any means: the Bill Hughes Sim Pairs, run by me in late June every year [and available nowadays to clubs anywhere in the world] is done by a straight computer deal, no choosing.

Overall, you can trust computer hands to be fair: hand dealt are less fair, though consistently so, thus you know what to expect.

Talking of organisers and picking hands, I reached the Gold Cup Final some years back, Great Britain's top event [won this year by Frances Hinden, who often posts here]. The last 24 boards were to be shown on Vu-Graph. The first 8, dealt in advance by one of the helpers, were quite freakish. Since he was winning, one of our opponents, Tony Forrester, wanted to be sure there was no picking of hands, so told the organisers this was not acceptable. As a result, we dealt the boards ourselves at the table, and duly dealt the only ten card suit I ever remember. biggrin.gif


--
David Stevenson ... ... ... <webjak666[at]googlemail.com>
Liverpool, England, UK ... ... ... Currently at home
Lawspage: http://blakjak.org/lws_menu.htm
BruceR
Posted since your last visit  Posted: Jan 21 2004, 05:28 PM

bridgetalk member
*

Group: Members
Posts: 1,615
Joined: 17 October 03
Location: UK



So I take it that as there are no guidelines, but experience has shown that none are necessary (yet !) cool.gif

My reservations were based on the fact that with software, like anything else, "you gets what you pays for", and "if it's not on the tin, then it probably doesn't do it". frown.gif

If it says on the box that it can be used for duplicate events at clubs, I am happy to believe it. But the club may be using software that only costs 29.95 and does not say that on the box. PC generated deals are not random, they are only "pseudo-random". And how pseudo-random numbers are "seeded", and then the results turned into a bridge hand is another issue. What's good enough for "home use", and statistical or analytical purposes, may not necessarily be good enough other uses. After all, 10,000 pre-dealt "random" hands sitting on a flat file somewhere would be enough to meet those requirements ! However, this is a futile mathematical digression if experience has shown there is no problem with using such software. sheep.gif

On the other hand, anything would be better than what passes for shuffling in some parts of New Zealand these days laugh.gif

BruceR
Posted since your last visit  Posted: Jan 21 2004, 05:30 PM

bridgetalk member
*

Group: Members
Posts: 1,615
Joined: 17 October 03
Location: UK




Sorry, "as" in first line should not be there biggrin.gif
bluejak
Posted since your last visit  Posted: Jan 21 2004, 07:34 PM

Forum Host
*

Group: IBLF
Posts: 5,984
Joined: 28 August 03
Location: Liverpool UK



I see your worry. I am totally convinced that the software used by national organisations has passed the tests set for it. That does not mean that it did when introduuced, but computer dealing has been around for over 20 years now.

However, I grant you that it may not be so true of computer dealing in the clubs. But I believe that someone's home-grown hand dealer with pseudo-random deals, not very well done, is still probably better than dealing by hand.


--
David Stevenson ... ... ... <webjak666[at]googlemail.com>
Liverpool, England, UK ... ... ... Currently at home
Lawspage: http://blakjak.org/lws_menu.htm
Gerardo
Posted since your last visit  Posted: Jan 21 2004, 09:36 PM

bridgetalk member
*

Group: Members
Posts: 7
Joined: 1 September 03
Location: Quilmes - Buenos Aires - Argentina



See the Documentation for Hans Van Staveren's Big Deal, see if that convinces you.
See also Kaj G Backas' BigDeal for Windows

It is good as for documentation, and affordable (at $0) biggrin.gif
Frances Hinden, UK
Posted since your last visit  Posted: Jan 22 2004, 10:09 AM

Unregistered









By having 20+ years of absence from duplicate, you have been lucky enough to miss out on the evolution of computer dealing programs. When they first came in, there were multiple problems caused either by the randomisation process (as you mention) but also from incorrect dealing algorithms.

(Digression: a common way to write dealing programs at one point was to take the ace of spades & give it to one of 4 hands with 25% probability, then the King of spades & use 25% probability for each hand.... there was a phase of very long club suits!)

There have been a couple of incidents of the same set of hands appearing more than once. Originally these may have been due to problems with the random number seed, but I believe most turned out to be human error.

However, as David says, things have progressed now to the point that computer dealing is generally considered far superior to hand dealing.

(Another digression: I saw an account of a novice pairs which used computer dealt hands, but then 'picked' them to ensure that each player averaged 10 HCP over the session. They thought this was a way of making things fair; I thought it was a great opportunity to psyche at the right time!)
BruceR
Posted since your last visit  Posted: Jan 22 2004, 10:49 AM

bridgetalk member
*

Group: Members
Posts: 1,615
Joined: 17 October 03
Location: UK



Driving home last night I realised I had forgotten something .... oops.gif

The computer cannot deal hands at all. Some poor soul(s) has to go through all the effort of sorting each pack out into suits, gathering together the cards for each hand, checking he has got it right .... rolleyes.gif

I am with Tony Forrester. Just SHUFFLE the b*****y things !! biggrin.gif

You can mathematically show that 3 drop shuffles are enough for a game like bridge, and 5 are ample. To overcome stickiness ("Queen over Jack") you need to riffle shuffle. So for Rubber Bridge: two riffle shuffles, five drop shuffles, and a cut. And that's as "random" as random can be.


But driving to work this morning, I realised that for Duplicate Bridge you hardly need to shuffle at all ! biggrin.gif

Why ? Because the cards are NOT gathered together into tricks, but each hand is kept separate until the next deal. laugh.gif

At Rubber Bridge, because the cards are gathered together into tricks in the order played, if you did not shuffle, you would get one card each from each of the previous tricks, and thus more or less the same hand shape, and cards played to the same trick last time would be in different hands again. So you have to shuffle thoroughly.

But at Duplicate Bridge the hands are kept discrete, then put together, then dealt one at a time. So if you did not shuffle, then each person would get one quarter of each of the previous hands. If players make one quick drop shuffle before putting the hands back, as they are supposed to, it would be a random quarter too. And that would be 3 or 4 days later, after playing about 20 other packs in between. So, what the hell.

As long as the hands are kept separate until the next deal, collated, and then dealt one at a time, only minimal shuffling is required. Which is exactly what the White books says. cool.gif


The leads me to believe:

A) 3 drop shuffles are enough for duplicate, which takes about 4 seconds, and the one-at-a-time deal is more important than the shuffle

B) "Queen over Jack" is not applicable to duplicate (there is no "stickiness")

C) New Zealand's problems must have come from incorrect *dealing*, not insufficient shuffling. They must have been dealing the first 13 cards to one hand, the next 13 to the next, instead of one-at-a-time

D) Using computers to generate "random" hands for one-off Duplicate Bridge (as opposed to simultaneous events) is a totally mistaken idea

Bruce

bluejak
Posted since your last visit  Posted: Jan 22 2004, 01:10 PM

Forum Host
*

Group: IBLF
Posts: 5,984
Joined: 28 August 03
Location: Liverpool UK



I am not sure I would like to be on the road when you are travelling to work! biggrin.gif

You are making a number of assumptions which I believe are incorrect.

First, just because the pattern of cards is different at duplicate from rubber does not mean there is no discernable pattern. The effects of card play are such that there are patterns even when the cards are kept in hands and good shuffling is needed to get rid of their effects.

Second, you say something about drop shuffles. If by drop shuffles you mean the old-fashioned shuffle called an "overhand" shuffle then anything lesss than thirty is inadequate to get any sensible effect, and there is some doubt as to whether you ever get randomisation.

To shuffle adequately requires seven riffle shuffles. Not one pack in twenty gets adequately shuffled when playing duplicate or rubber bridge in clubs or elsewhere.

Third, you say players should shuffle the cards before replacing them in their hand. While this might help, in practice many players are brought up to sort them as a matter of good manners: many players do not shuffle: most players who shuffle use a single overhand shuffle which is as much use as taking the largest card out and putting it at one end of the hand.

Fourth, you refer to 3 or 4 days later. This seems to me to suggest that we are talking of remembering a deal. But that is not the point: it is whether deals are fair and random. Ordinary dealt hands are not: computer dealt hands are. Bridge technique depends in part on the odds and they are not accurately present in hands that are inadequately shuffled.

Fifth, as computer dealt hands are becoming more common, so are dealing machines. So many of these computer dealt hands are not put in the boards by a person. In the cases where they are it is normal for players to make the boards up themselves on the first round so that instead of shuffling and dealing each player helps make up a couple of boards. This takes much the same time to get much fairer hands.

Sixth, do not miss the point of my story about Tony Forrester: he has faith in computer-dealt thands. What he did not have faith in was hand dealt boards which he had not seen dealt, and could therefore have been picked.

Seventh, when playing with computer dealt hands players usually get a copy of the hands at the end, and they love that.

Computer dealt hands are fairer and more popular.



--
David Stevenson ... ... ... <webjak666[at]googlemail.com>
Liverpool, England, UK ... ... ... Currently at home
Lawspage: http://blakjak.org/lws_menu.htm
BruceR
Posted since your last visit  Posted: Jan 22 2004, 01:10 PM

bridgetalk member
*

Group: Members
Posts: 1,615
Joined: 17 October 03
Location: UK



Sorry, when I said "White Book", what I really meant was Law 6B. I knew I had read it somewhere.

Further thought leads me to another conclusion ...

E) At Duplicate Bridge, the cards should NOT be riffle shuffled.

If you take out the hands, collate them, then do two 'perfect' riffle shuffles, and then deal one-at-a-time, you will reproduce the previous hand almost exactly. In other words, two 'perfect' riffle shuffles at Duplicate is equivalent to not shuffling at Rubber.

biggrin.gif
bluejak
Posted since your last visit  Posted: Jan 22 2004, 01:16 PM

Forum Host
*

Group: IBLF
Posts: 5,984
Joined: 28 August 03
Location: Liverpool UK



QUOTE

As long as the hands are kept separate until the next deal, collated, and then dealt one at a time, only minimal shuffling is required. Which is exactly what the White books says.


Where?


--
David Stevenson ... ... ... <webjak666[at]googlemail.com>
Liverpool, England, UK ... ... ... Currently at home
Lawspage: http://blakjak.org/lws_menu.htm
BruceR
Posted since your last visit  Posted: Jan 22 2004, 01:51 PM

bridgetalk member
*

Group: Members
Posts: 1,615
Joined: 17 October 03
Location: UK



Thank you David, for your long reply. I do appreciate it.

I meant Law 6B, not the White book. And I was really referring to dealing one-at-a-time.

I think this thread has got slightly out of scope. oops.gif

I was talking in the context of your average club's bi-weekly level 3 open pairs. That's what the original web quote was referring to (I think). I had a mental image of some poor TD spending an hour and half arranging hands from a sheaf of laser A4.

More prestigious events, where there is more money, man-hours, and machinery available, are a different kettle of fish.

Can I ask one question ?

You talked of dealing machines. Does this mean that there are machines that will deal the hands generated by software ? If that's the case, count me as a supporter of the idea.

biggrin.gif

Bruce


whereagles
Posted since your last visit  Posted: Jan 22 2004, 04:29 PM

bridgetalk member
*

Group: Members
Posts: 482
Joined: 3 December 03
Location: Portugal



Bluejak: don't get me wrong. I'm just mentioning *one* particular computer-dealt event where most of the hands were quite crazy, thoughout 300+ deals. There was absolutely nothing strange about all the other computer-dealt events I played, and at the time I had already played a few. Therefore my surprise about that one particular event. I know it could happen... but it just felt strange.
bluejak
Posted since your last visit  Posted: Jan 22 2004, 07:42 PM

Forum Host
*

Group: IBLF
Posts: 5,984
Joined: 28 August 03
Location: Liverpool UK



QUOTE
At Duplicate Bridge, the cards should NOT be riffle shuffled.

If you take out the hands, collate them, then do two 'perfect' riffle shuffles, and then deal one-at-a-time, you will reproduce the previous hand almost exactly. In other words, two 'perfect' riffle shuffles at Duplicate is equivalent to not shuffling at Rubber.


All I can really say is that this is wrong. At any game of cards that requires shuffling you should always riffle shuffle since thirty seconds of riffle shuffling is equivalent to about an hour of overhand shuffling.

QUOTE
As long as the hands are kept separate until the next deal, collated, and then dealt one at a time, only minimal shuffling is required. Which is exactly what the White books says.

Sorry, when I said "White Book", what I really meant was Law 6B. I knew I had read it somewhere.


Law 6A requires "thorough" shuffling. Minimal shuffling is illegal. Thorough shuffling means seven riffle shuffles [which happens one time in thirty or so], about twenty minutes or so of overhand shuffling [which happens once every five years or so] or computer deals.

QUOTE
I think this thread has got slightly out of scope.


It contains matters of general interest so it matters not.

QUOTE
I had a mental image of some poor TD spending an hour and half arranging hands from a sheaf of laser A4.


As explained that is unlikely to happen. It may happen in other events. Certainly my wife and I have made up various sets of boards over the years, sometimes from computer deals.

QUOTE
You talked of dealing machines. Does this mean that there are machines that will deal the hands generated by software ? If that's the case, count me as a supporter of the idea.


Certainly there are such machines. I wonder which club started you thinking about this - some clubs have their own.


--
David Stevenson ... ... ... <webjak666[at]googlemail.com>
Liverpool, England, UK ... ... ... Currently at home
Lawspage: http://blakjak.org/lws_menu.htm

addREPLY  newTOPIC  newPOLL  

Quick Jump