2

I've noticed that bridge hands in many top tournaments, are often extremely uncommon combinations of distributions and types. For example a single hand at the 2021 EBU Premier League contained a hand with AKJ 9 times clubs (East), a 7-4-2 distribution (S), and a 21 point hand (N).

This seems far from unique, or particular to that tournament.

Clearly these aren't random distributions, or else we would see these kinds of deals regularly at ordinary clubs.

Are they chosen this way, to test players on more extreme hands, because top players usually won't make mistakes or show much difference on routine hands? Or for drama and entertainment value, to attract onlookers? Or for other reasons?

Also, how are such hands dealt? Are there set criteria for hands, in a match, or set biases towards certain kind of unlikely hand being more common, or more usual hands filtered out on computer dealing? What's deemed "too boring" or "too extreme", or however it works?

3
  • 3
    Clubs rarely shuffle well, as that is more challenging than usually given credit. Expert and advanced players prefer the accurate distributions of computer-randomized hands; and if there was any large-scale systematic bias to the hands generated someone would have noticed, and piped up, long before now. Dec 14, 2021 at 17:54
  • Agree re: hand-shuffled boards, but most clubs are also using computer dealt hands these days. Having said that, there are many many people who have piped up in the last 20 years. And they've all been unable to prove anything statistically - they just know it "looks cooked". Of course, if they really knew how the boards were cooked, their scores would improve, ja-nie?
    – Mycroft
    Dec 15, 2021 at 19:09
  • 1
    The odds of all 4 players being dealt 13 cards of a single suit each are exactly the same as the odds of any other specific deal you've ever seen. It's not less likely; it's just more noticeable by humans.
    – GendoIkari
    Dec 16, 2021 at 1:05

1 Answer 1

9

All bridge players "know" that the hands are cooked. They all have their own ideas how. Each person's ideas are different. They're all wrong. Mostly this is because humans, and bridge players in particular, are "pattern-matching" machines, and in a 24-board session, they can always find some coincidence (for instance, yesterday I had 4 voids in 24 boards, including 2 diamond voids on consecutive hands! Monday I had either 5 or 6 5-5 or better hands! ...)

Bridge hands for high level tournaments are always created on a computer with zero human interaction, and almost all use some frontend on top of Big Deal. Big Deal basically does two things:

  1. Generates a true random 160-bit integer. This is not easy, but is much easier than when the documentation was written in 2000. This becomes 36 96-bit integers through a cryptographic PRNG.
  2. Converts each 96-bit integer into a bridge hand using a one-to-one mapping [*].

It then outputs the hands in whatever formats are required (PBN, hand record,...) for purpose.

As explained in the documentation,

as far as possible the software should view the operator as an enemy in the struggle to generate unique sequences.

So, far from "selecting interesting hands", the program does its absolute best to ensure that no selection takes place at all. 100% random.

You can see why this is critical - almost always the weak point of a security system is the jellyware. Humans do things wrong, can be bribed, cheat on process (and make bad process)... You never want "selection" in creation of bridge hands for high level tournaments, because that's one more point of failure (so, the selector is friends with the English team captain, so he selects in favour of weak NT hands. Or Chinese, and "influenced" by the Party, so makes good Precision hands. Or a particular pair just gets told about a few hands before the session, bypassing any need to play games with the placement of the board after the tray was removed...)

In fact, in a high-level event, nobody ever sees the hands at all until they get pulled out of the board. The dealer program outputs a seed and a PBN file. Sure you could read it - it's basically plaintext - but nobody does. The PBN file gets sent to the Dealer4 or Duplimate software, and they rattle out the N sets of boards, with "show boards" off [**]. The boards are played, and the reporting software creates and displays the PDF of the hand records automatically from the PBN. Only then does anybody actually look at the hands [***].

There have been some poor dealers in the past (and some still exist).

  • The ACBL's dealing program was famously "presumed to be" insecure based on the described process, and in 2016, it was finally (openly) cracked. Turns out all you needed was 3 consecutive boards to uniquely identify the entire set. I'm sure none of the multi-millionaire sponsors ever decided to hire a few cryptanalysts and a bunch of computer time to do that before 2016 and didn't tell anybody, but who knows? Needless to say, they use Big Deal now.
  • BBO's dealer is known to only be able to deal 80% or so of all bridge hands (because its randomness is <96 bits). That's not really a bad thing - you'll see a repeat in a few fewer million years on average - but if it turns out that the hands that can't be generated are skewed somehow by the missing bits, it could tweak the percentages slightly. Statistical tests done on the hands however do not show this.

The joy of bridge is that you don't have to cook the hands to have a great game (or, to cheat on the famous math joke, "most boards are interesting. If it's boring, that's unusual enough to be interesting, isn't it?")

[*] Not all 96 bit integers map to a bridge hand, so some have to be thrown out and the next one created. So maybe as many as 40 96-bit values have to be cranked out to get 36 boards.

[**] It's actually faster to deal with "show boards" off, never mind better for security vs shoulder-surfers. It's just a distraction.

[***] Of course, the directors should have access to the hand records to resolve issues, from use of UI to "I found a card on the floor".

4
  • This is good info. What I was wondering was more, whether the random computerised dealing software was configured to reject or prefer certain types of hands/deals, more often than pure chance. Meaning for example that perhaps deals with feature combinations prone to passing out are filtered out to increase chances all boards played? Or it internally deals 10000 randomised deals and preferentially selects those with specific features or widely differing bidding possibilities, based on internal logic? Something like that. Still black box and not seen by people, but selecting for trickier deals?
    – Stilez
    Dec 16, 2021 at 10:33
  • 1
    100% random. There is zero filtering. I tried to explain in my answer why it has to be to be safe. Please read the Big Deal announcement I linked for all the nasty details. As a side note, it's much harder to tweak for "interesting" than you might think, for a large number of reasons. But for one, I play Precision with a 10-12 1NT NV. So we open "all" 10 counts at the 1 level. So a "boring flat hand" that might be a passout in most of the room, is a very interesting hand at our table. But seriously, THERE IS NO FILTERING, there's can't be to be safe, it has been frequently tested.
    – Mycroft
    Dec 16, 2021 at 19:02
  • Frankly, there are so many statistical tests that if you pass them, you can't be filtering for anything. And they get very regularly run, because if it was found that the hands were tampered with in any way, it would cause the kind of stink in that second link about the old ACBL dealing program. And the experts would very quickly tweak their systems to match the hands they actually get rather than "pure random". We know that because we watched the systems evolve back in 1990 when "computer-dealt hands" started to become standard.
    – Mycroft
    Dec 16, 2021 at 19:04
  • Interesting!! If there's a web resource on this aspect of bridge history, I'd be interested to read more, too!
    – Stilez
    Dec 16, 2021 at 23:16

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .