Those Totally Random Deals

[bixby]

Just finished playing in the Daylong MP (A) game. On deals 5, 6, and 8, where I had 9-card trump fits (deal 7 was at notrump), the adverse trumps broke 4-0 three times in a row. It doesn't even bother me any more. I just expect it when playing here.

[HeavyDluxe]

So, a couple things...

 

First, this is not a sufficient sample size to prove anything.

 

Second, remember that, if the format is 'best hand' or anything like that, it stands to reason that things would be less evenly distributed than normal. After all, no random deal sets would have you end up with the best hand.

 

More to the point, IMO, the other players against whom you're comparing are dealing with the same hands, suit divisions, and other challenges. One of the great things with duplicate is that the variance is norm'd out between comparing players. So, what does it matter if the suit divisions are abnormal? If you're right and they're abnormally so, plan around it and watch your scores go up against the field since they're playing for more 'random' breaks.

[hrothgar]

Same reply as the last time you advanced this nonsense

 

Out of curiosity, why do you think that BBO would ever bother to implement a feature like this?

 

Regardless, I have an offer for you...

 

Let's play 8 X 16 board challenge matches (That way we can each see the complete set of hands and make sure that no one cherry picking the hands...)

 

At the end of the match, we'll run a hypothesis test to see whether the hands split unevenly or not.

(95% confidence level)

 

If the null hypothesis can not be rejected, I'll pay to have $100 BBO bucks put in your account.

If the null hypothesis can be be rejected, you pay to put $25 into mine...

[nige1]

Totally random is virtually impossible.

Suppose you're given a sequence of "random digits" but when you examine them they are all threes.

It's possible that the sequence is genuinely random. A sequence of threes has the same likelihood as any other specific sequence of digits.

Similarly, when you examine a set of bridge-hands, you are quite likely to notice peculiarities.

It's more interesting and significant if the sample is large and you predict specific peculiarities, before you examine the sample.

[steve2005]

Just finished playing in the Daylong MP (A) game. On deals 5, 6, and 8, where I had 9-card trump fits (deal 7 was at notrump), the adverse trumps broke 4-0 three times in a row. It doesn't even bother me any more. I just expect it when playing here.

playing at the club with computer deals at one table got 4-0 trump breaks on both hands.

Plus i was sitting against the club owner!

I assure you it was just a a coincidence. These things just happen and you tend to remember them.

[steve2005]

Second, remember that, if the format is 'best hand' or anything like that, it stands to reason that things would be less evenly distributed than normal. ... After all, no random deal sets would have you end up with the best hand.

BBO deals hands randomly and then if the human doesn't have the most points, hands are switched.

So the distribution of all 4 hands should be random.

Now, the human having the best hand will on average have less shortness as those suit's cards don't contribute as many hcps or 0 in case of voids; which is what decides best hands.

[johnu]

Just finished playing in the Daylong MP (A) game. On deals 5, 6, and 8, where I had 9-card trump fits (deal 7 was at notrump), the adverse trumps broke 4-0 three times in a row. It doesn't even bother me any more. I just expect it when playing here.

Just by chance, I also played in that same tournament. On those 3 deals, 5-6-8, I got 2-2 breaks on all 3 deals. Of course, I paid premium prices to get BBO's Double Platinum membership. That means the default standard BBO members have to get worse suit splits to even out the overall suit distributions.

[hrothgar]

So, for kicks and giggles, I decided to throw together a quick sim to examine what level of impact the "best hand" format had on expected shape.

 

I generated two pairs of Sims.

 

Sim 1A shows the result of a best hand style Deal with North having a hand whose HCPs are >= the other seats

Sim 1B shows the result of an unconstrained Deal

 

Sim 2A and 2B repeat this same simulation

 

If you look at the numbers, you'll see that the "best" hand is very slightly biased in favor of balanced patterns with short suits and biased against holding voids.

(The chance that you hold a balanced pattern is ~ .3% more likely)

 

Converse, the chance that East (an arbitrary non best hand) holds a void is even more slightly biased (~.05%) in the positive direction.

However, I'd be shocked if this was really noticeable across any reasonable number of hands

 

Sim 1A

Void_North: 0.0498356

Balanced_North: 0.480438

Void_East: 0.051725

Balanced_East: 0.474331

Generated 10000000 hands

Produced 2759034 hands

Initial random seed 1577142813

Time needed 8.000 sec

 

 

Sim 1B

Void_North: 0.0510673

Balanced_North: 0.475933

Void_East: 0.0510549

Balanced_East: 0.475924

Generated 10000000 hands

Produced 10000000 hands

Initial random seed 1577142854

Time needed 9.000 sec

 

Sim 2A

Void_North: 0.0498041

Balanced_North: 0.480475

Void_East: 0.0517035

Balanced_East: 0.473993

Generated 10000000 hands

Produced 2755153 hands

Initial random seed 1577142972

Time needed 8.000 sec

 

Sim 2B

Void_North: 0.0511672

Balanced_North: 0.475817

Void_East: 0.051189

Balanced_East: 0.476197

Generated 10000000 hands

Produced 10000000 hands

Initial random seed 1577142996

Time needed 8.000 sec

 

 

 

Best_Hand = hcp(north) >= hcp(south)

and hcp(north) >= hcp(west)

and hcp(north) >= hcp(east)

 

Void_North = spades(north) == 0

or hearts(north) == 0

or diamonds(north) == 0

or clubs(north) == 0

 

Balanced_North = shape(north, any 4432, any 4333, any 5332)

 

Void_East = spades(east) == 0

or hearts(east) == 0

or diamonds(east) == 0

or clubs(east) == 0

 

Balanced_East = shape(east, any 4432, any 4333, any 5332)

 

condition Best_Hand

 

action

average "Void_North" Void_North,

average "Balanced_North" Balanced_North,

average "Void_East" Void_East,

average "Balanced_East" Balanced_East

[johnu]

I recently played a hand where I had

 

[hv=pc=n&s=sk72ha952dq85cjt4]133|100[/hv]

 

I remembered this hand because spot cards are my cell phone number and it has exactly 10 HCP. I then researched this and found that is 635 trillion to 1 that I would get this exact hand. What are the odds that I would notice this hand on a random deal? :rolleyes:

[nige1]

M * E * R * R * Y * * * * C * H * R * I * S * T * M * A * S

[HeavyDluxe]

BBO deals hands randomly and then if the human doesn't have the most points, hands are switched.

So the distribution of all 4 hands should be random.

Now, the human having the best hand will on average have less shortness as those suit's cards don't contribute as many hcps or 0 in case of voids; which is what decides best hands.

 

Sorry... You're right, of course. I was being imprecise in an attempt to make a different point, though.

 

I was trying to say that, sitting south at a random table, I wouldn't expect to have the 'best hand' 8 deals in a row. So, it's a little disingenuous to complain about non-standard distributions (of cards, hcps, or whatever else) in a tournament whose foundation principle is a manipulation of HCP to benefit one particular player.