BBO Discussion Forums: One sided points distribution - BBO Discussion Forums

Jump to content

  • 2 Pages +
  • 1
  • 2
  • You cannot start a new topic
  • You cannot reply to this topic

One sided points distribution One side gets the best hands

#1 User is offline   Askehall 

  • Pip
  • Group: Members
  • Posts: 4
  • Joined: 2020-April-23

Posted 2020-November-09, 15:19

Hi,

I play with 3 friends at our own table in the Casual bridge section. The table is set up as "Total Points". We play for about 1 hour, 3 times a week. I've noticed that on each session the hands seem to favour one side. Fine if it's your side with the points but demoralising for the opposition. Towards the end of the evening it is getting pretty obvious that the opposition has lost the will to play. Also I notice that we are presented with many more hands with 6 and 7 card suits. Far more than I would have got in an hours at our bridge club session.
Is this just my perception or are the hands pre-prepared and presented to favour one side or the other.
What has been a fantastic facility during this lockdown period,is becoming a bit tedious for one of the couples each night.

Thanks Mike
0

#2 User is offline   hrothgar 

  • PipPipPipPipPipPipPipPipPipPipPip
  • Group: Advanced Members
  • Posts: 15,372
  • Joined: 2003-February-13
  • Gender:Male
  • Location:Natick, MA
  • Interests:Travel
    Cooking
    Brewing
    Hiking

Posted 2020-November-09, 15:27

Quote

Is this just my perception


Yes

Quote

are the hands pre-prepared and presented to favour one side or the other.


No

Quote

I notice that we are presented with many more hands with 6 and 7 card suits.
Far more than I would have got in an hours at our bridge club session.


If your normal bridge club deals things by hands, then there is a very real chance that the hands are artificially flat. (People are bad at dealing unbiased hnads. Computers are good at it. So you might have an inaccurate sense of what's what.

Please note:

I've been on these forums for umpteen years.
I have seen complaints to yours raised over and over again
I've seen a whole bunch of people waste a lot of time and effort investigating these claims
And it never pans out...

If you take a statistically valid sample, the hand generators are fair
Alderaan delenda est
1

#3 User is online   mycroft 

  • Secretary Bird
  • PipPipPipPipPipPipPipPip
  • Group: Advanced Members
  • Posts: 7,059
  • Joined: 2003-July-12
  • Gender:Male
  • Location:Calgary, D18; Chapala, D16

Posted 2020-November-09, 15:44

It's your perception. Let's assume that you can get through 10 hands in your "about 1 hour".

The chance that 8 or more out of 10 hands go one direction is 10% (5% each way). 7 or more: 20%. This is often enough for human pattern matchers (and that's what bridge players are) to notice. Along with "all the finesses are wrong today" or "the queen is always above the jack", or any of the other conspiracy theories that people notice when it's there and don't notice when it isn't.

I can't tell you the number of times I heard in real life "the cards ran N-S (or E-W) this session". I can't tell you the number of times I've *said* that. And that's 27 deals. And it probably wasn't that much running, it's just that "all the games were N-S, and when E-W had the cards, N-S had long suits and could pre-empt".

In order to notice actual trends, you need to be looking at thousands (if not millions) of deals.

From a theoretical point of view, what's in it for BBO to cook the hands at all? It's a lot of work, and the only thing they'll get from it is griping.

As far as the less-balanced hands go, if you were dealing cards, it is very likely that you were insufficiently shuffling, and that tends to cause flatter hands than the math would expect.

I understand the wish to play total point, but one of the big draws of duplicate is that even when you don't have the cards, you can still do "less badly" than the other pairs you're comparing against, and score well. It's the only way I (as a known bad card holder) survive in this game. [Yes, that last line is a joke, but it's *my* conspiracy theory]
When I go to sea, don't fear for me, Fear For The Storm -- Birdie and the Swansong (tSCoSI)
1

#4 User is offline   steve2005 

  • PipPipPipPipPipPipPip
  • Group: Advanced Members
  • Posts: 3,148
  • Joined: 2010-April-22
  • Gender:Male
  • Location:Hamilton, Canada
  • Interests:Bridge duh!

Posted 2020-November-09, 18:48

wish i had $1 every time someone makes this type of unfounded complaint.
Sarcasm is a state of mind
0

#5 User is offline   dsLawsd 

  • PipPipPipPip
  • Group: Full Members
  • Posts: 299
  • Joined: 2017-September-15

Posted 2020-November-10, 00:25

Can you specify the deal type? If so, you could ask for Vu-graph deals either random or specific event (as can be done with you plus 4 robot tables).
Else can you pick IMPS or matchpoints or even close the table periodically and re-open?
0

#6 User is online   thepossum 

  • PipPipPipPipPipPip
  • Group: Advanced Members
  • Posts: 2,362
  • Joined: 2018-July-04
  • Gender:Male
  • Location:Australia

Posted 2020-November-10, 03:25

Its a strange thing. When I first started playing Bridge, computers were still at a very basic level. I doubt there was very much close to a Bridge computer/dealer, especially not one widely available to thousands of people. Occasionally during a whole session, maybe a few hours, of rubber bridge I could go through the whole session without even making a bid; sometimes you felt the cards were totally stacked against you

Having said that, I am convinced though, that sometimes the hands we get on this site are not random
0

#7 User is offline   nige1 

  • 5-level belongs to me
  • PipPipPipPipPipPipPipPipPip
  • Group: Advanced Members
  • Posts: 9,128
  • Joined: 2004-August-30
  • Gender:Male
  • Location:Glasgow Scotland
  • Interests:Poems Computers

Posted 2020-November-10, 19:59

View Postpilowsky, on 2020-November-09, 22:36, said:

Just for reference here is a link to an excellent chart of HCP and shape distributions. Some of them are a little surprising.
From the sublime to the ridiculous :) Arithmetic is not my long suit but those random deal statistics seem too good to be true :( What do Helene_T and Cherdano think?
0

#8 User is offline   nige1 

  • 5-level belongs to me
  • PipPipPipPipPipPipPipPipPip
  • Group: Advanced Members
  • Posts: 9,128
  • Joined: 2004-August-30
  • Gender:Male
  • Location:Glasgow Scotland
  • Interests:Poems Computers

Posted 2020-November-11, 09:15

View Postpilowsky, on 2020-November-10, 20:15, said:

I hope not. I rely on them and they work just fine!
The agreement between generator practice and theoretical calculation seems too perfect to me.
0

#9 User is offline   hrothgar 

  • PipPipPipPipPipPipPipPipPipPipPip
  • Group: Advanced Members
  • Posts: 15,372
  • Joined: 2003-February-13
  • Gender:Male
  • Location:Natick, MA
  • Interests:Travel
    Cooking
    Brewing
    Hiking

Posted 2020-November-11, 10:35

View Postnige1, on 2020-November-11, 09:15, said:

The agreement between generator practice and theoretical calculation seems too perfect to me.


Have you looked at the sample size?

5.68971224e10 is an incredibly large number

For kicks and giggles, I ran the following a few times
(Note, I am using a sample size that is 1/100th of the true value because I only have 16GB of memory on this machine)

sample_size2 <- 568971224
bar <- rbinom(sample_size2,1,.215512)

mean(bar)

The results are very tight

(Sorry, that I am too lazy to calculate confidence bounds more formally)
Alderaan delenda est
0

#10 User is offline   nige1 

  • 5-level belongs to me
  • PipPipPipPipPipPipPipPipPip
  • Group: Advanced Members
  • Posts: 9,128
  • Joined: 2004-August-30
  • Gender:Male
  • Location:Glasgow Scotland
  • Interests:Poems Computers

Posted 2020-November-11, 11:58

View Posthrothgar, on 2020-November-11, 10:35, said:

Have you looked at the sample size? 5.68971224e10 is an incredibly large number
Hrothgar might well be right :) No big deal :) At the head of the table the agreement seemed plausible but the exact agreement at the foot of the table surprised me :)
9-2-2-0	4,677,922  0.0082 			0.0082
 9-4-0-0	549,810 	0.0010 			0.0010
10-2-1-0   625,482    0.0011 			0.0011
10-1-1-1   225,405    0.0004 			0.0004
10-3-0-0   87,554      0.00015   		0.00015
11-1-1-0   14,180      0.00002   		0.00002
11-2-0-0	6,655   	0.00001   		0.00001
12-1-0-0   164   		0.0000003   	0.0000003
13-0-0-0	0              0.0000000000 0.0000000006

0

#11 User is online   smerriman 

  • PipPipPipPipPipPipPip
  • Group: Advanced Members
  • Posts: 3,724
  • Joined: 2014-March-15
  • Gender:Male

Posted 2020-November-11, 13:21

View Postnige1, on 2020-November-11, 11:58, said:

At the head of the table the agreement seemed plausible but the exact agreement at the foot of the table surprised me :)

The reason it's not as surprising as it first seems is due to the choice of the number of significant figures to display.

For example, to 2sf, the theoretical 12-1-0-0 split is 0.00000032%, while the table would give 0.00000029%.

Restricting it to 1sf means it would match the theoretical figure if there were anywhere between 143 and 199 occurrences of that distribution, which is a gigantic margin of error.

Likewise, if the 10-3-0-0 shape was measured to 3sf, there would be a slight deviation in the table (0.000154 vs 0.000155) - 2sf would work for anywhere between 82501 and 88190 occurrences, which is inevitable.
0

#12 User is offline   hrothgar 

  • PipPipPipPipPipPipPipPipPipPipPip
  • Group: Advanced Members
  • Posts: 15,372
  • Joined: 2003-February-13
  • Gender:Male
  • Location:Natick, MA
  • Interests:Travel
    Cooking
    Brewing
    Hiking

Posted 2020-November-11, 13:23

The variance for an individual bucket of a binomial distribution is

N(p)(1-p)

What happens to (p)*(1-p) as p gets further and further away from .5?
Alderaan delenda est
0

#13 User is online   smerriman 

  • PipPipPipPipPipPipPip
  • Group: Advanced Members
  • Posts: 3,724
  • Joined: 2014-March-15
  • Gender:Male

Posted 2020-November-11, 14:07

View Posthrothgar, on 2020-November-11, 13:23, said:

The variance for an individual bucket of a binomial distribution is

N(p)(1-p)

What happens to (p)*(1-p) as p gets further and further away from .5?

Actually, the opposite of what you're implying :)

While sqrt(p(1-p)) gets smaller as p gets smaller, it gets larger in proportion to p - for example, when p is 0.1, you get 0.3 (three times p), while when p is 0.00001, you get 0.003: 300 times p.

So you would expect greater deviation in the results for smaller p - it's the fact this is completely overwhelmed by the rounding in the table that the numbers match so well.
0

#14 User is offline   hrothgar 

  • PipPipPipPipPipPipPipPipPipPipPip
  • Group: Advanced Members
  • Posts: 15,372
  • Joined: 2003-February-13
  • Gender:Male
  • Location:Natick, MA
  • Interests:Travel
    Cooking
    Brewing
    Hiking

Posted 2020-November-11, 15:11

View Postsmerriman, on 2020-November-11, 14:07, said:

Actually, the opposite of what you're implying :)

While sqrt(p(1-p)) gets smaller as p gets smaller, it gets larger in proportion to p - for example, when p is 0.1, you get 0.3 (three times p), while when p is 0.00001, you get 0.003: 300 times p.

So you would expect greater deviation in the results for smaller p - it's the fact this is completely overwhelmed by the rounding in the table that the numbers match so well.


Aren't we just concerned about the absolute variance here and not how large it is relative to p itself?
Alderaan delenda est
0

#15 User is online   smerriman 

  • PipPipPipPipPipPipPip
  • Group: Advanced Members
  • Posts: 3,724
  • Joined: 2014-March-15
  • Gender:Male

Posted 2020-November-11, 15:48

View Posthrothgar, on 2020-November-11, 15:11, said:

Aren't we just concerned about the absolute variance here and not how large it is relative to p itself?

Yes and no. If you want to look at absolute errors, consider the following:

The 5-5-2-1 figure of 3.1739% gives sqrt(p(1-p)) = 0.1697.

The 10-3-0-0 figure of 0.00015% gives sqrt(p(1-p)) = 0.0012.

So the SD for the second figure is about 1/140th the size of the first.

But the first figure would need an absolute error of 5e-5 to give a different rounded result, while the second figure would need an error of 5e-6 to give a different rounded result.

So you need an absolute error of a tenth the size - but with 1/140th the standard deviation.

So while you are interested in the absolute error, the fact it is smaller for smaller numbers isn't important - it's only about how that compares to how much you're scaling down the required accuracy.

If you measured both to the same number of significant figures, the smaller values would be more volatile.

If you measured both to the same number of decimal places (so it's solely about absolute error), the smaller values would be more accurate - but they're clearly not doing that in the table.
1

#16 User is offline   tadyan 

  • Pip
  • Group: Members
  • Posts: 1
  • Joined: 2020-June-21

Posted 2023-January-06, 18:27

I absolutely agree with the point raised (one sided distribution). I play 25 casual boards twice a week with 3 friends. I would say in 90% of the games one side is favored. It is very annoying for the side which doesn't get points. I think the algorithm should distribute cards more evenly.
0

#17 User is offline   johnu 

  • PipPipPipPipPipPipPip
  • Group: Advanced Members
  • Posts: 4,833
  • Joined: 2008-September-10
  • Gender:Male

Posted 2023-January-06, 20:35

View Posttadyan, on 2023-January-06, 18:27, said:

I absolutely agree with the point raised (one sided distribution). I play 25 casual boards twice a week with 3 friends. I would say in 90% of the games one side is favored. It is very annoying for the side which doesn't get points. I think the algorithm should distribute cards more evenly.


There are 2 scientifically verified methods for reversing the observed hand bias. First one is that each of the players rotate their chairs and tables by 90 degrees whenever you feel that the cards are running in one direction or another. You also need to get a compass and ensure that tables and chairs are aligned directly N/S or E/W because there's no telling which way cards are going to go if you are randomly aligned. That way the good cards are distributed to the other side. This has been known to work at 73.5%.

The other involves using 2 different voltage converters. Suppose you are in a country on 120 volts. Get a 120 to 220 volt converter, and a 220 to 120 volt converter. Basically you want to go from 120 volts input, to 220 volts, then back to 120 volts. I don't know the details about how this works but you can check on the flat earth websites. This works 100% but it costs a fair amount of money for the converters. Glad I could help.
0

#18 User is offline   hrothgar 

  • PipPipPipPipPipPipPipPipPipPipPip
  • Group: Advanced Members
  • Posts: 15,372
  • Joined: 2003-February-13
  • Gender:Male
  • Location:Natick, MA
  • Interests:Travel
    Cooking
    Brewing
    Hiking

Posted 2023-January-07, 04:13

View Posttadyan, on 2023-January-06, 18:27, said:

I absolutely agree with the point raised (one sided distribution). I play 25 casual boards twice a week with 3 friends. I would say in 90% of the games one side is favored. It is very annoying for the side which doesn't get points. I think the algorithm should distribute cards more evenly.


this is an incredibly stupid idea
On many many levels

It would be stupid if the system knew in advance how many boards you were going to play
It would be even more stupid if the system didn't know in advance how many boards you were going to play
Alderaan delenda est
0

#19 User is offline   pilowsky 

  • PipPipPipPipPipPipPip
  • Group: Advanced Members
  • Posts: 3,620
  • Joined: 2019-October-04
  • Gender:Male
  • Location:Israel

Posted 2023-January-07, 04:44

I think tadyan makes an excellent point.
Justice and equity are vitally important.
BBO could play its part by changing the dealer so that all players get 10HCP on every deal.



2

#20 User is offline   Catfi 

  • Pip
  • Group: Members
  • Posts: 5
  • Joined: 2011-August-06

Posted 2023-January-10, 13:47

View PostAskehall, on 2020-November-09, 15:19, said:

Hi,

I play with 3 friends at our own table in the Casual bridge section. The table is set up as "Total Points". We play for about 1 hour, 3 times a week. I've noticed that on each session the hands seem to favour one side. Fine if it's your side with the points but demoralising for the opposition. Towards the end of the evening it is getting pretty obvious that the opposition has lost the will to play. Also I notice that we are presented with many more hands with 6 and 7 card suits. Far more than I would have got in an hours at our bridge club session.
Is this just my perception or are the hands pre-prepared and presented to favour one side or the other.
What has been a fantastic facility during this lockdown period,is becoming a bit tedious for one of the couples each night.

Thanks Mike

I do not care about the "demoralizing" part, but I do care about the randomness part. Many of us use principles of odds and probability in bidding and playing. Randomness is necessary for any meaningful uses of odds and probability. In the past 2 days, I have played on BBO. Today, N played 18 of the 20 hands and S played 1. Yesterday, E/W played 21 consecutive hands! I am having trouble assigning randomness to this distribution.
0

  • 2 Pages +
  • 1
  • 2
  • You cannot start a new topic
  • You cannot reply to this topic

2 User(s) are reading this topic
0 members, 2 guests, 0 anonymous users