[IowaST8]

OK, Are the Bridge Hands on Bridgebase truly random? Or, are the hands sometimes padded to make the game more interesting?

[Gerardo]

Some tourneys have prepared (as in uploaded) hands. Randomness of those depends on the tourney organizer.
Players are informed when this happens, with a message "This tourney uses prepared hands" or something like that (sorry, didn't see the message in english in a while)

Instant Tournaments use hands recycled from old tourneys, but those were random the first time around.

Best Hand tournaments: South gets the hand with more HCP (or one of them if tied). Either by swapping or rotating, I don't remember.

No other hand is manipulated in any way.

Dealer's randomness has been tested for bias. No bias, truly random.

[pescetom]

If the hands look padded to you it may be because you have frequently played hands that were manually shuffled. Manual shuffling tends to produce less interesting hands than a genuine random shuffle.

[AL78]

It should be noted that streakiness is a property of randomness. I have streaks of getting poor hands and lower than average declarer play frequency, but when I have tested a sample distribution of the mean HCP I get dealt over several months, it comes out very close to the 10 HCP you would expect. If you ask someone to roll a dice and someone else to write down a sequence of random numbers from 1 to 6, it is possible to tell which sequence has come from rolling the dice, because what people think a random sequence should look like is different to how most random sequences really behave.

[thepossum]

I am wondering how you could test a single tourney's hands for randomness

But depending on how you look at randomness I think its quite obvious that sometimes the hands on BBO are random and sometimes not

I have definitely observed patterns in the hands from time to time

Also I am sure on occasion hands are dealt and balanced in someway across tournaments using a dealer script. So clearly not random at all

And who knows what shuffling algorithm they use

I have no idea how BBO deals their stuff

[smerriman]

thepossum, on 2021-March-07, 18:53, said:

But depending on how you look at randomness I think its quite obvious that sometimes the hands on BBO are random and sometimes not

I have definitely observed patterns in the hands from time to time

Also I am sure on occasion hands are dealt and balanced in someway across tournaments using a dealer script. So clearly not random at all

You seem to be fundamentally misunderstanding what 'randomness' means. If you *couldn't* spot patterns from time to time, that would prove beyond doubt that the hands were *not* random.

Unless you're referring to some Goulash-style tournaments, which are specifically handcrafted and thus not relevant to this discussion, it has been shown time and time again there is no bias in the deals, so you're just outright wrong there.

Perhaps this would be worth reading, which explains why you - and humans in general - may find it so difficult to understand this: https://cocosci.prin...papers/hard.pdf

[thepossum]

smerriman, on 2021-March-07, 22:22, said:

You seem to be fundamentally misunderstanding what 'randomness' means. If you *couldn't* spot patterns from time to time, that would prove beyond doubt that the hands were *not* random.

How do you know. Couldn't it be a very long sequence of "random" hands where nobody spotted any patterns at all. From your logic if occasional patterns in the infinity of randomness are indicative of randomness then so are infinitely long sequences without patterns. Maybe the person observing has a poor sense of pattern observation too. Who knows

Pattern matching is an attribute of each individual's brain. How many people have observed patterns etc

All I am hoping is that it was demonstrated that the number of people and the patterns they observed were also random etc

Quote

Unless you're referring to some Goulash-style tournaments, which are specifically handcrafted and thus not relevant to this discussion, it has been shown time and time again there is no bias in the deals, so you're just outright wrong there.

I'm not going to ask for details on how they assessed the hands for lack of bias, whatever that would mean anyway

Just wondering. From you above comments if the randomness consultant showed a set of hands and nobody spotted any patterns if the algorithm had to be changed to ensure there are patterns that you assure me guarantee randomness


Do I need to go on with this discussion

Its a fun topic though. I will see if I can dig up my musings on randomness in Bridge

Lets think. I know the number of hands dealt so far is a tiny fraction of infinity etc

[pilowsky]

You will be unsurprised to discover that there is quite a bit of research concerning people’s knowledge about probability as it applies to real life.
I mean, what are the odds that you can make money out of other people misunderstanding things.

Here is a table reproduced from one of these studies.
(Regarding) Independence of events
There is a Lottery number that has not come out in the last 10 draws.
Is it more likely to come out on the next draw?
Answer_________N____%___CI
Absolutely no:___759:_57.9:__54.6–61.2
Mostly no:______300:_27.0:__24.0–30.1
Mostly yes:______175:_13.5:_11.4–15.9
Absolutely yes:____23:_1.6:__1.0–2.5
(Tomei A., et al., (2017) Misbeliefs About Gambling in a Convenience Sample from the General Population. J Gambling Studies 33:899-906).

More than 40% of people do not understand that the chance of a lottery number being drawn is independent of the numbers drawn previously.
This belief system is not surprising. Elsewhere in this forum, we have discussed various strange beliefs that people hold for one reason or another, even when there is clear evidence to the contrary.
Medical students are taught that “common things occur commonly” and that “if you hear hoofbeats outside the window, think horse, not zebra.” - In Australia anyway.
Students are taught all kinds of mnemonics to help remember common patterns that appear in certain diseases e.g.: fair, fat, forty & female = gallstones.
They don’t tell you (know) the proportion of cases that the so-called typical presentation accounts for (I’m guessing about 10-15%).

Two interesting things about randomness. First, life requires a certain amount of variety. If your heart beats monotonically, you’re dead. All the same, the interval between beats is not entirely random - that would also be bad.
The second thing is that the human nervous system has evolved to see shapes and patterns: This ability is useful to bridge players.
A sheep has specialised neurons that activate when they see other sheep and entirely different neurons that fire when they see a sheepdog. If shown the image of a dog that is upside down the neuron fails to fire.
Monkeys are different; they don’t care about the orientation of the threat - but sheep spend minimal time swing around in the trees.

If Bridge hands were monotonically different, then as smerriman notes, something would be wrong.
It may or may not be ‘hard’ to learn, but as they say, if you don’t know, you don’t know.

To be fair, though, most of the people who write here enjoy collecting masterpoints, an activity that bears little apparent relation to skill.
But, there are other benefits to acquiring masterpoints. There must be. Why else would people bother?

[thepossum]

[hrothgar]

Few quick comments here:

I am not aware of any well established tests to determine whether or not a large set of bridge hands are sufficiently random. You can certainly test individual hypotheses about the hands (who is getting how many HCPs, are various shapes being dealt in the right proportions, etc.). However, I am not aware of bridge equivalent to the Diehard tests and the like that are used to validate the performance of PRNG's like the Mersenne twister and the the like. People have spent a whole lot of time thinking about how to prove that these are sufficiently random for various purposes.

The best way to validate the performance of the BBO hand generators would be inspection of the code base.

You want to verify that they

1. They are using a high quality PRNG
2. They are using a good method to provide this with seeds
3. The code that transforms the output of the PRNG --> deals is not introducing bias

Absent this type of inspection, you're (pretty much) forced to rely on testing a set of different hypotheses about some large number of deals

BBO hired a statistician to do just back a years back.
He was happy and content with the results.
BBO has not changed the dealing code base since then.

I think that they should publish their code.

[mythdoc]

Quote

I think that they should publish their code.

I agree. I also think there should be transparency on how the deals get to the various events, be it anonymous play, live games, deal pool, etc, since some (me included) assert that the deals differ among these. A while back I suggested one would notice a difference in hands between a series of “just declare” IMPs challenges and just declare MPs challenges. Probably no one tried this, because we are all stuck in our Rorschach test believing what we want to believe. (I include myself in this.)

[thepossum]

I was searching my hand archives for a clear example of a non random hand

It was like join the dots preschool Bridge where every play was a simple and fairly tedious sequence of covers round the table

Clearly not random at all. After I noticed it on the first few rounds I just played the whole hand that way

Just thinking of ways to test for bias. How about comparing the frequency of all hands dealt so far on Bridgebase with the theoretical uniform distribution of all possible hands

P = 0.5 X 10-38 or something

I can't remember tests and distributions. Maybe a simple Chi Square test

[smerriman]

thepossum, on 2021-March-08, 14:04, said:

I was searching my hand archives for a clear example of a non random hand

Thanks, this is now the best comment I've ever read on BBO.

[thepossum]

smerriman, on 2021-March-08, 14:23, said:

Thanks, this is now the best comment I've ever read on BBO.

😂

I was just thinking of a simple problem with one non random hand. Then we have to look at non random sets or sequences

I'm sure we've all had sets of hands which all seemed very similar

[thepossum]

Suppose there were 10^n hands dealt so far.

Let's hope for no duplicates

How does ChiSquare look

When I wake up I will calculate it. Degrees of freedom could be a problem

Of course I appreciate a uniform distribution is no guarantee of randomness

Sorry if I appear to be at all facetious about demonstrating lack of bias

I always used to get anxious about claims that there was almost zero chance of duplicate MAC addresses, and what would or could happen if there were

The level of discussion and appreciation of the issue reminds me of the old school level problem of how many people you need to get 50% chance of two with the same birthday. But in this case we have 10^38 order choices

And as far as I'm concerned in the infinity of true randomness it doesn't matter what any hand or set of hands are like and how they were dealt. The infinity of true randomness includes totally non random stuff too

But does that pass the Bridge pub test

In that case do things have to be engineered to ensure enough apparent randomness

Also without knowing I imagine hands for different levels of tournaments are designed. Much like golf. You don't want Tiger Woods playing a boring par 3 suburban course etc. They bore me to death and I can hardly hit the ball

[hrothgar]

thepossum, on 2021-March-08, 14:35, said:

I always used to get anxious about claims that there was almost zero chance of duplicate MAC addresses, and what would or could happen if there were

Back in the weird old days, MAC addresses were set manually on the cards using pins.
Even now, many drivers allow you to set a MAC address

Believe me, duplicate MAC addresses happen (and are a pain in the butt to handle)

This used to be a lot worse when stuff was all on a broadcast bus because you'd have two different IPs ARPing the same MAC

[thepossum]

hrothgar, on 2021-March-08, 16:18, said:

Back in the weird old days, MAC addresses were set manually on the cards using pins.
Even now, many drivers allow you to set a MAC address

Believe me, duplicate MAC addresses happen (and are a pain in the butt to handle)

This used to be a lot worse when stuff was all on a broadcast bus because you'd have two different IPs ARPing the same MAC

I imagined it must happen, never heard what happens it did.

And since the Internet of Things or whatever its called etc - how many people have accidentally controlled someone else's device

The other amusing things is that some of the people who keep having a go at me over randomness or other stuff here from time to time would assure me that each individual troll attempt were totally independent too

[hrothgar]

thepossum, on 2021-March-08, 14:35, said:

Suppose there were 10^n hands dealt so far.

Let's hope for no duplicates

How does ChiSquare look

When I wake up I will calculate it. Degrees of freedom could be a problem

Of course I appreciate a uniform distribution is no guarantee of randomness

Sorry if I appear to be at all facetious about demonstrating lack of bias

When folks are analyzing whether or not bridge hands are random, they normally look at things a bit differently that you are.

It is certainly true that you should NOT expect to see bridge hands occur more than once. However, unless you have a completely pathological hand generator, these sorts of issues normally occur because the process for seeding the PRNG is flawed and not the PRNG process or the hand generator itself. (Note: Because of this - and the way the PRNGs work - duplicate hands occur in runs. If you have one, you end up with a whole bunch in a row)

Normally, the sorts of tests that you run look at the distribution of HCPS or the distribution of hand shapes and the like and whether these are biased in some way. A more complex analysis would try and see whether there was some periodicity to these distributions. For example, the total number of HCPs is unbiased, but this is because odd number hands are biased in one direction and even number hands are biased in another.

Sadly, there's a near infinite number of ways in which a hand generator might be flawed and testing all of them is expensive. (Its for this reason that I think that code inspection is a much better way to handle these sorts of issues.)

If you're interested in taking a deeper dive on this subject, I recommend looking a discussion around the period of a PRNG. You also might want to look at the following tests that have been used to evaluate PRNGs. https://en.wikipedia...i/Diehard_tests

[pilowsky]

The problem may be more subtle than simply asking, "Are the hands dealt randomly?"
This is a fairly "easy" question. Given that so many hands are dealt and played, it is doubtful that a single individual (even those of us who play hundreds or thousands of hands a week) would detect any non-random pattern or if they did, be able to use it to their advantage.
Remembering that a non-random pattern (as pointed out earlier) is not a series of hands where patterns don't appear from time to time.
A true lack of randomness would exist if the coin turned up "heads" every single time. As Tom Stoppard explained in "Rosencrantz and Guildenstern are dead" when that happens, it is proof that you are dead because it's impossible (like the spade I suppose).

All kinds of illusory ideas about randomness pop into one's mind only to be quickly dismissed, either by ourselves or with a few caustic comments on the Forum.

In education, there is a different problem: the "quality of candidature".

It works like this. In a total population of students undertaking a final examination, it is necessary to generate a single final ranking.

This ranking is used in many ways, but one way is to allocate students to their university entrance preferences. The system may differ from jurisdiction to jurisdiction. Still, somehow it is necessary to "moderate" results so that the students that undertake modern European history are ranked in the same way as those studying advanced calculus, biology, legal studies or design.

How is this relevant to the topic at hand?
If we look at the Daylongs, there is an anti-cheating measure in place. Each contestant plays 8 boards (in the DL1, for example) played by roughly 30 other players.
This means that I am in a pool of ~240 contestants in a competition where about 1000 people finish.

Don't get me wrong - I enjoy my DL1 - even though the Zenith seems to have pinched some of the participants, but:

If 1000 people are playing 8 boards each that about 30 other people also play, then 1000 players in the DL1 compete against vastly more players than entrants.

This is where we run into the quality of candidature problem. It may not be the deals that are non-random, but it seems that the opponents are.

Reductio ad absurdum, imagine for a moment that you are playing a team match against another pair, and you are doing reasonably well when suddenly the pair that you are competing against leaves and Nige1 and shyams step in.
One could argue that as a test of your skill, this is not a bad approach. I'm not so sure.

My understanding of how a Bridge competition is meant to work is that the hands are randomised, not the opponents.
Elsewhere, barmar argued that although this may be true, The "good players" consistently do well. What about the rest of us? My results oscillate so wildly it's hard to know who I'm playing against.
Is it a bad result simply because my opponents on the day were all Nige1's, shyams' or kenberg's or are my good results happening through a judicious combination of luck and strangely incompetent opponents.

To make the competition fair, it would be necessary to use the same "quality of candidature" tools that the Board of Studies uses to ensure that all students are ranked fairly before the results are passed on to the next stage.

If this level of randomness of opposition is truly the case, the tournament might better be described as a race where the runners are completely blind.
In the end, you rank "x" depending on who else showed up on the day.

It's satisfying until you learn that every other blind competitor had a completely different pool of opponents: some of whom were using bicycles.

[mythdoc]

thepossum, on 2021-March-08, 14:35, said:

You don't want Tiger Woods playing a boring par 3 suburban course etc.

You also, certainly, wouldn’t expect a champion golfer to continue to visit your course if its design was so out of date that it sometimes (routinely) allowed bad players to beat him/her with bad plays. The GIB robots need all the help they can get, in my experience.