[nige1]

pescetom, on 2019-November-29, 15:36, said:

The really sobering thing is that the whole paradigm is about to change with the advent of online play, and 2027 will be much too late for WBF to put a foot in that door.

Pescetom is right. The on-line game provides a wonderful opportunity to rescue Bridge from oblivion especially in aspects of legal innovation BBO has done an excellent job

Reducing mechanical-error opportunities. (Admittedly, some of this would be possible, F2F, using tablets, phones, or bridge-mates).

• Illegal calls (e.g. insufficient, out-of-turn, illegal doubles and redoubles, extra passes)
• Illegal plays (e.g. leads out of turn, plays out of turn, revokes).
• Misscoring (e.g, misattribution of tricks, miscalculation).

Simplifying disclosure e.g. An application like full-disclosure can automatically display the meaning of your call to your opponents.

Simplifying claim law. You claim by stating a number of tricks and playing on, with your hand exposed. Opponents can accept the claim. Or they can dispute it, by continuing, double-dummy (with Declarer playing single-dummy) A theoretical draw-back is that this permits declarer to embark on a fishing expedition. In practice, that doesn't seem happen, perhaps because only particularly naive and tolerant defenders would allow it. On-line claim protocol seems to encourage claims, to save defenders from unnecessary pressure, to speed up the game, and to virtually eliminate claim-disputes. I would be interested in the experience of on-line directors.

Facilitating the prevention/detection/prosecution of cheats. e.g.

• Put S/W and N/E players in separated Faraday cages.
• Allow viewgraph scrutiny by several commentators and many kibitzers, augmenting TV and computer-records of bidding and play.
• Correlate hands with calls to deduce actual partnership methods and contrast them with those declared on system-cards/supplementary notes.
• Monitor selected deals where overbidding. underbidding, and anti-percentage plays are rewarded.

[barmar]

FYI, BBO considers Full Disclosure convention cards to be a failed experiment. The software for creating them has been abandoned and there are no plans to port it to the HTML version.

Unless someone can come up with a better design, it's incredibly difficult to use, and only really effective for simple bidding sequences (mainly the first couple of rounds of bidding, and simple conventions like Blackwood). You have to enter every bidding sequence, so the tree explodes as you get further into bidding sequences.

It might be possible to develop AI that examines a corpus of auctions by a pair and infers their agreements from it. But you'd need a huge number of deals so that you'll have enough examples of the less common sequences for the AI to detect the patterns. And having it turn that into human-understandable explanations would also be difficult.

[paulg]

Although I can understand why BBO thinks Full Disclosure was an unsuccessful experiment, it did a good job of highlighting many of the issues of automating the process.

My main complaint, in the competitive world of bridge, was that there was no good means of changing your methods depending on the meaning of the opening bid. For example, my bids after a one club opener differ depending on whether it is a strong club, a short club, a Polish Club, a Moscito club, or a Magic Diamond club. Which probably means more information needs to be passed in a defined way.

[hrothgar]

barmar, on 2019-December-10, 20:36, said:

FYI, BBO considers Full Disclosure convention cards to be a failed experiment. The software for creating them has been abandoned and there are no plans to port it to the HTML version.

Unless someone can come up with a better design, it's incredibly difficult to use, and only really effective for simple bidding sequences (mainly the first couple of rounds of bidding, and simple conventions like Blackwood). You have to enter every bidding sequence, so the tree explodes as you get further into bidding sequences.

It might be possible to develop AI that examines a corpus of auctions by a pair and infers their agreements from it. But you'd need a huge number of deals so that you'll have enough examples of the less common sequences for the AI to detect the patterns. And having it turn that into human-understandable explanations would also be difficult.

Few comments here

1. You hardly need any kind of sophisticated AI to generate convention files. Rather, having a system that looks at individual alerts / announcements, sees whether this sequence existing in the convention file, and then populations the text string that was written into the file is more than sufficient. The system would quickly converge onto something that handled "common" sequences.

2. If you compare this degree of coverage to, say, a standard ACBL or even WBF convention card I think it lines up quite well (at least for non competitive auctions)

3. As was already mentioned, competitive bidding becomes a lot more difficult. There are ways to skin this cat as well, but none of them are simple

[pescetom]

hrothgar, on 2019-December-11, 03:36, said:

1. You hardly need any kind of sophisticated AI to generate convention files. Rather, having a system that looks at individual alerts / announcements, sees whether this sequence existing in the convention file, and then populations the text string that was written into the file is more than sufficient. The system would quickly converge onto something that handled "common" sequences.

The system could even look at every call made and see if the current sequence is explained already, if not ask for an explanation - thus gradually populating the database. Explanations could be either free text (with easily selected repetition of recent/common texts) or selections from a tree of described rules (e.g."fourth suit forcing") distinguished by colour and clickable to open an URL with detailed description.
There should also be a way of editing the descriptions already entered - a text file in the frequently proposed markup language would be enough, although a purpose written editor would be better - and the ability to print out a top down system description (for study and use as convention card).

Our RA proposed to move convention cards into the app currently used to share results and hands, unfortunately nothing has happened yet.

[barmar]

hrothgar, on 2019-December-11, 03:36, said:

3. As was already mentioned, competitive bidding becomes a lot more difficult. There are ways to skin this cat as well, but none of them are simple

I could be wrong, but I suspect that most of the problems with inadequate disclosure involve competitive sequences. E.g. what's the meaning of double, is a pass forcing, etc.

But if you're just talking about automating the disclosure of relay sequences, for instance, I agree that it would be relatively easy to populate that dynamically.

[hrothgar]

barmar, on 2019-December-11, 09:22, said:

I could be wrong, but I suspect that most of the problems with inadequate disclosure involve competitive sequences. E.g. what's the meaning of double, is a pass forcing, etc.

But if you're just talking about automating the disclosure of relay sequences, for instance, I agree that it would be relatively easy to populate that dynamically.

Here's my view of things:

For non competitive sequences, I think that a system that used some kind of "learn as you go" methodology would relatively quickly converge on having all of the "common" sequences automated with a bunch of "one off" bids that needed to be populated.

I believe that the same holds true for competitive sequences, however, to use this type of methodology in this world you'd probably need to have a convention file that was specific to some opposing convention file.

So, I might have one version of my 2/1 file that I use playing against convention file foo and a second that gets used playing against convention file bar.
And, you probably need some kind of version numbers as well.

[axman]

Two of the things that particularly easy to adapt for online is a timestamp and unobtrusive smoothing of an individual's tempo. Smoothing tempo can be done with a simple algorithm that progressively measures how long a player takes, recalculates what his good** tempo is and holds his call or play when made early. A subroutine can create a report that can be analyzed for performance.


** good tempo as in the time that he comfortably maintains (not average tempo)

[pescetom]

Getting back to the OP...

nige1, on 2019-November-29, 17:42, said:

Facilitating the prevention/detection/prosecution of cheats. e.g.

• Put S/W and N/E players in separated Faraday cages.
• Allow viewgraph scrutiny by several commentators and many kibitzers, augmenting TV and computer-records of bidding and play.
• Correlate hands with calls to deduce actual partnership methods and contrast them with those declared on system-cards/supplementary notes.
• Monitor selected deals where overbidding. underbidding, and anti-percentage plays are rewarded.

You could also add:
- Monitor success in defence vs success in attack. A pair that does much better in defence than in (combined) attack is not necessarily cheating but is worthy of further attention.

[pescetom]

axman, on 2019-December-11, 10:48, said:

Two of the things that particularly easy to adapt for online is a timestamp and unobtrusive smoothing of an individual's tempo. Smoothing tempo can be done with a simple algorithm that progressively measures how long a player takes, recalculates what his good** tempo is and holds his call or play when made early. A subroutine can create a report that can be analyzed for performance.
** good tempo as in the time that he comfortably maintains (not average tempo)

Exactly.

I would generalise this to "Tempo Management" and enable definition of a profile among several parameters such as:
- degree of smoothing (0-100%)
- action when BIT is noticeable (BIT notification to opps and Director)
- budget (total seconds per board of thinking allowed for each player and/or the pair)
- credit per call (bonus seconds for each card held, bonus of 10s for jump or first call, carry-over bonus)
- maximum time per call (either fixed max or some percentage of remaining credit)
- timeout warning (seconds before maximum time)
- timeout expiry action (automatic bid of preselected call (Pass if none) or penalty and a further timeout).

Different profiles could be designated for different types of event and play.
An important policy decision will be whether to smooth heavily in important events (relying on delayed transmission to edit out delays and make spectating enjoyable) or make some kind of compromise to speed up play (at risk of cheating, difficult TD decisions and timeout mistakes).

[barmar]

In any case, this all seems highly speculative to me. Even if AI could be developed to solve these problems, it would be an expensive project -- who is going to pay for it? We can't create Laws that assume it until we have reasonably robust mechanisms in place. So it seems like this discussion is putting the cart before the horse.

If we want to revise the Laws for online play, the near-term version should just reflect the existing technology -- remove everything related to mechanical problems that are obviated by the software (plays/calls out of turn, insufficient bids, revokes, etc.). Laws related to unintended calls/plays probably don't need to be changed, but interpretation would reflect the possibility of misclicks. Probably the most significant difference would be in the area of disclosure, since self-alerting and explaining is the norm online.

In fact, there was an attempt to specify online laws about a decade ago, and that was pretty much the extent of it.

[hrothgar]

its unclear to me where the dividing lines should be between the "laws" of online bridge and the conditions of contest for individual events.

As a practical example, the existing laws for F2F bridge include the following:

Quote

B. The Deal
The cards must be dealt face down, one card at
a time, into four hands of thirteen cards each;
each hand is then placed face down in one of the
four pockets of the board. No two adjacent cards
from the deck shall be dealt into the same hand.
The recommended procedure is that the cards be
dealt in rotation, clockwise.

Arguably, this suggests that the Laws for Online Bridge should specify which algorithms should be used for hand generation.

I think that it would be worthwhile for the Laws to specify whether or not it is appropriate to add noise into the timing of bids. I am less sanguine that the Laws themselves should cover mechanisms around disclosure (even though I am very much in favor of migrating to online convention cards)

[pescetom]

barmar, on 2019-December-13, 10:33, said:

In any case, this all seems highly speculative to me. Even if AI could be developed to solve these problems, it would be an expensive project -- who is going to pay for it? We can't create Laws that assume it until we have reasonably robust mechanisms in place. So it seems like this discussion is putting the cart before the horse.

If we want to revise the Laws for online play, the near-term version should just reflect the existing technology -- remove everything related to mechanical problems that are obviated by the software (plays/calls out of turn, insufficient bids, revokes, etc.). Laws related to unintended calls/plays probably don't need to be changed, but interpretation would reflect the possibility of misclicks. Probably the most significant difference would be in the area of disclosure, since self-alerting and explaining is the norm online.

In fact, there was an attempt to specify online laws about a decade ago, and that was pretty much the extent of it.

Wasn't it in the previous century? In any case the world is a different place now. And drafting the Laws before the software is implemented is putting the horse in front of the cart, not vice versa.

I hope WBF will go beyond the minimal measures you mention and tackle issues like location/grouping/screening of players, broadcasting and privacy, intelligent disclosure, tempo management and claim resolution (not that I fully agree with nige's proposal, but it would be outrageous to maintain the current rules). Partly because these things can improve the game, partly because they are possible and people developing the software will tackle them with or without regulation and the results could be chaotic or not at all what WBF wanted. It's all too easy to imagine an internet platform that manages to oust all the others and impose it's own rules, ranking scheme and competitions on the game.

[barmar]

hrothgar, on 2019-December-13, 10:52, said:

Arguably, this suggests that the Laws for Online Bridge should specify which algorithms should be used for hand generation.

We already have computer-dealt hands in F2F bridge. This is authorized in 6E4:

Quote

The Director may require a different method of dealing or pre-dealing to produce the same wholly random expectations as from A and B above.

If you feel that the algorithm needs to be specified more precisely than just "wholly random expectations", it should be done across the board, not just for online bridge.

I think enshrining the algorithm in the Laws would be inappropriate. If a flaw were later found in the selected algorithm, we'd be stuck with it until the next version of the Laws. Specifying the objective is better, it allows flexibility in implementing it.

[pran]

hrothgar, on 2019-December-13, 10:52, said:

its unclear to me where the dividing lines should be between the "laws" of online bridge and the conditions of contest for individual events.

As a practical example, the existing laws for F2F bridge include the following:

Quote

B. The Deal
The cards must be dealt face down, one card at
a time, into four hands of thirteen cards each;
each hand is then placed face down in one of the
four pockets of the board. No two adjacent cards
from the deck shall be dealt into the same hand.
The recommended procedure is that the cards be
dealt in rotation, clockwise.

Arguably, this suggests that the Laws for Online Bridge should specify which algorithms should be used for hand generation.

I think that it would be worthwhile for the Laws to specify whether or not it is appropriate to add noise into the timing of bids. I am less sanguine that the Laws themselves should cover mechanisms around disclosure (even though I am very much in favor of migrating to online convention cards)

Why didn't you include reference to Law 6

Quote

E. Director’s Option on Shuffling and Dealing
1. The Director may instruct that the shuffle and deal be performed at each table immediately before play starts.
2. The Director may himself perform the shuffle and deal in advance.
3. The Director may have his assistants or other appointed agents perform the shuffle and deal in advance.
4. The Director may require a different method of dealing or pre-dealing to produce the same wholly random expectations as from A and B above.

which is even more important than your reference?

I wonder if Law 6B as such is ever applied anywhere except at informal (social) bridge games?

[pescetom]

barmar, on 2019-December-13, 11:03, said:

If you feel that the algorithm needs to be specified more precisely than just "wholly random expectations", it should be done across the board, not just for online bridge.

I think enshrining the algorithm in the Laws would be inappropriate. If a flaw were later found in the selected algorithm, we'd be stuck with it until the next version of the Laws. Specifying the objective is better, it allows flexibility in implementing it.

Maybe what should be enshrined in the Laws is the way the generating algorithm should be tested for randomicity and who is responsible for this.
In any case this is the sort of debate that should be going on.

[pescetom]

pran, on 2019-December-13, 11:12, said:

I wonder if Law 6B as such is ever applied anywhere except at informal (social) bridge games?

Decidedly so, not everyone has a dealing machine. Our club plays three formal competitive games a week and two of them are dealt manually at table (the other one is computer designated but still prepared manually).

[barmar]

pran, on 2019-December-13, 11:12, said:

I wonder if Law 6B as such is ever applied anywhere except at informal (social) bridge games?

In the US hand dealing is still the norm in team games.

[pescetom]

barmar, on 2019-December-13, 11:24, said:

In the US hand dealing is still the norm in team games.

Here too, except in the final stages of more important games.

[hrothgar]

pran, on 2019-December-13, 11:12, said:

Why didn't you include reference to Law 6

which is even more important than your reference?

I wonder if Law 6B as such is ever applied anywhere except at informal (social) bridge games?

I think the obvious reason is that Law 6 needs to be changed for Online Bridge games.
Having an individual player "deal" at their table is completely nonsensical.

However, specifying the way in which cards are dealt is analogous.