Join Bridge Winners
All comments by Richard Willey
1 2 3 4 5 6 7 8 9 ... 132 133 134 135
You are ignoring the author of this comment. Click to temporarily show the comment.
> “After reading many, many old world championship deals, I agree
> with the author. My own view is that neither Avarelli, Chiaradia,
> D’Alelio or Pabis-Ticci were good enough to be on an ordinary
> open Swedish national team – if they played bridge. They
> needed a partner who “helped them”. Otherwise, they
> were simply too weak.”

My understanding is that the strength of field for National teams has changed dramatically over the past 50 years. Standards for competitive bidding especially have improved dramatically. The rise of professional players has allowed our best bridge players to develop their game enormously.

I'm not at all surprised that many of the Blue Team player's skill level doesn't stand up.
6 hours ago
You are ignoring the author of this comment. Click to temporarily show the comment.
FWIW, I was able to find a copy of “The game of Bridge: a challenge for ILP” on google and found it an interesting read.

I'd be curious to compare the results of your work with what some kind of evolutionary algorithm might created. For example, assume for the moment that we have two populations of algorithms.

Population 1 consists of a set of 1H opening bidders, each with its own probability density function that defines that set of hands that might get opened 1H. (Lets assume that we're locking down the “shape” of the hand, so the main dimension of variance is what constitutes a minimum and a maximum opening hand)

Population 2 consists of a set of limit raisers. Each of the limit raisers has its own belief about what does / does not constitute a limit raise to 3M. (All limit raises will promise 4+ hearts however, strength is pretty much unbound.

Our game plays out in three stages

1. A hand is dealt to the 1H bidders and they decide whether or not the hand is suitable for a 1H opening.

2. Each time a hand is opened 1H, the 1H opener passed a corpus of hands to the limit raisers that define the probability density function that constitutes a 1H opening along with a specific hand held by the limit raiser. The limit raiser uses this information to determine whether or, alternatively to pass.

3. If the limit raiser decides to make a limit raise, it passes a corpus of hands to the opener that are consistent with a limit raise opposite the original definition of the 1H opening. The opening bidder then uses this information to decide whether or not to raise to game or pass (probably easiest to pretend that slams don't exist for now)

Once the final contrast has been determined, we lock down opener and limit raisers hands, deal a whole bunch of hands to the opposing pair, and play things out.

Algorithms that do relatively “well” in the bidding have lots of children, algorithms that do relatively “poorly” have fewer. Ideally, the system will eventually converge on a consistent set of agreements about the range for a 1H opening bid, coupled with the strength that a limit raise should show.
8 hours ago
You are ignoring the author of this comment. Click to temporarily show the comment.
FWIW, if you feel like travelling and Boston seems too far I'll be in Barcelona for a 5 days or so in Mid October, followed by another 5 in Marrakech…

And if you want to go further afield, I'll probably be spending 3-4 weeks in Cambodia in late November December.

First round is always on me!
15 hours ago
You are ignoring the author of this comment. Click to temporarily show the comment.
Bother! Wish I had known of this earlier. I might have scheduled a trip to Florida.

Regardless, good luck and let me know if you find yourself in/near Boston…
15 hours ago
You are ignoring the author of this comment. Click to temporarily show the comment.
Quick comment: From a practical perspective, an awful lot of computer security exploits boil down to some combination of

1. Social engineering (compromising an individual)
2. Physical issues (breaking into something)

The more people have access to a given piece of information and the longer that they have this info, the greater the chance that something can go wrong.

Ultimately, this is all going to boil down to some trade off between convenience / reliability and security. Given the goals of the common game, folks are probably making the right call in focusing on the former rather than the later…

At the end of the day, you have people playing the same boards at very different points in time. Given that the barn doors are wide open, not sure whether it makes sense to worry about gates on the windows
17 hours ago
Richard Willey edited this comment 16 hours ago
You are ignoring the author of this comment. Click to temporarily show the comment.
Thanks for posting these…

Interesting to see approaches that used to be more popular
Sept. 17
You are ignoring the author of this comment. Click to temporarily show the comment.
> Does anyone seriously imagine that playing in separate
> rooms on tablets is c***t-proof?

I don't recall anyone claiming that placing players in separate rooms is cheat proof. I'm certainly don't believe this to be true, and I am one of the strongest advocates for making this change that I know of….

The fundamental mistake that you are making is assuming that we are attempting to secure a system against all possible threats.

The way in which type of threat modeling normally works is that you assume different sets of adversaries who possess different types of powers. You then build in controls that are designed to thwart different types of adversary powers.

In many cases, you will explicitly decide to exclude certain classes of adversaries and adversary powers from consideration. (This typically occurs because there is not a cost effective way to do so). For example, I would be surprised if the ACBL were able to put a set of controls in place that could successfully block a nation state actor from compromising the results of a tournament. However, this doesn't mean that there is not value in implementing controls that would make it significantly more expensive for lesser classes of adversaries from achieving the same end.
Sept. 17
You are ignoring the author of this comment. Click to temporarily show the comment.
> In “Blue team system” they may respond 2/1 in a 3 card suit.

I BELIEVE the Neopolitan 2C/2D convention permitted a 2/1 from 1M to 2m on a two bagger if your intention was to rebid 4m…
Sept. 17
You are ignoring the author of this comment. Click to temporarily show the comment.
Nic, I didn't talk to a single person on your precious little committee… Don't even know who the members might be be, nor do I care.

FWIW, on this end, I saw an interesting article at work stating the following

“Felix Doerre and Vladimir Klebanov from the Karlsruhe Institute of Technology discovered a flaw in the mixing functions of GnuPG's random number generator. An attacker who obtains 4640 bits from the RNG can trivially predict the next 160 bits of output.”

This got me thinking about some of the work that Hans Van Stavern had done way way back implementing Big Deal in a secure manner and I started asking myself whether or not the ACBL had done anything similar. As I recall, in that same thread, you ended up posting details about the ACBL's implementation which made it patently obvious how bad things really were. (Using an LCG, the ridiculously small space, …)

At that point in time

A. I made some public postings to a crypto list to get third party opinions
B. I asked some of my cronies at work to help operationalize a crack

Now I readily admit, you beat us to the punch… The high school student that you farmed this work out to had an operational crack long before we did. (Personally, I suspect that this has something to do with the fact that you had a copy of the ACBL's dealer code which translated numbers to deals which is the only really hard part of this process)

And you know what, maybe we should have bothered to follow through with the work to completion, but at the time there really didn't seem to have been much need to do so.
Sept. 16
You are ignoring the author of this comment. Click to temporarily show the comment.
Sad the the judges didn't find this convincing…

It should have been significant to the finding
Sept. 16
You are ignoring the author of this comment. Click to temporarily show the comment.
Nic, statisticians work with public data sets all the time. And then still divide their data into training sets, test sets, and validation sets. And the good ones don't cheat by using the validation sets to train their models. Even if we can't conclusively prove that we didn't cheat, its still a good idea to make and effort and follow the appropriate norms.

With this said and done, its interesting to note that the judges didn't care about the brand new data set. Was the results of this analysis presented at the hearing?
Sept. 16
You are ignoring the author of this comment. Click to temporarily show the comment.
> The law does not require a pattern. One murder does not
> require a pattern. One theft does not require a pattern.
> One assault and battery doesn't require a pattern.

Steve, you are creating a strawman.

Assume for the moment that it was possible to use a single board to prove with absolute certainty that a player foo has cheated. I don't think that anyone would argue that that is not sufficient for a conviction.

With this said and done, I think that many people doubt that a single board is able to provide this type of conclusive evidence.

Equally significant, I expect that there are a lot more occasions in which murderer foo commits a single act of murder than there are cheater bar played a single hand of bridge which can't help but shape the surrounding discussion.
Sept. 16
You are ignoring the author of this comment. Click to temporarily show the comment.
Maybe there are lots of female professional killers, but they're very very good…
Sept. 16
You are ignoring the author of this comment. Click to temporarily show the comment.
> You forgot to mention stable…..

Far too many people know me here to get away with a claim like that…
Sept. 16
You are ignoring the author of this comment. Click to temporarily show the comment.
> RW, it is annoying that you continue to think the process
> of collecting all the data, and using an AI approach to
> create a statistical answer, rather than looking at a dozen
> or two dozen hands, for “impossible” decisions that
> worked out, is better for the ACBL and the WBF to
> work out how and who cheats.

Hey Gene,

Its really sad that someone who is as cynical and bitter as you are about anyone else's contributions can't see the obvious problems with your own suggestions.

Given how little faith you have in any individual or any institution, why do you believe that whatever Star Chamber that you put in place will get anything right? Moreover, by denigrating attempts to improve record keeping or apply actual math to studying these problems, they only thing that you're doing is furthering divorcing your proposed judiciary from any kind of accountability.

> You want to be smart, and viewed as a bridge expert?

For the record, I have no pretensions of being a bridge expert nor do I have any expectation or even real desire to be viewed as such. I am, however, very very smart. And I have a lot of experience with areas like security and statistics that are (or at least should be) directly applicable to the game.

And oh, BTW, be glad that I'm not motivated by something as petty as winning bridge games because if I wanted to I'd be robbing you all blind. (And if you don't believe me, think back a couple years when I initially started commenting about the security of the ACBL hand generators. I didn't have to tell you all…)
Sept. 16
You are ignoring the author of this comment. Click to temporarily show the comment.
Normally, when people say that they are planning to buy your book and give you money, one responds with the words “Thank you” rather than yet another gratuitous insult…
Sept. 16
You are ignoring the author of this comment. Click to temporarily show the comment.
Yes

I have a copy of The Roman Club System of Distributional Bidding on my bookshelf. (Along with four or so books on Blue and Neapolitan and some weird old pamphlets on Vienna…)
Sept. 16
You are ignoring the author of this comment. Click to temporarily show the comment.
FWIW, I just tried to buy a copy and couldn't yet find this on either Amazon or Baron Barclay. (I did find a reissue of Enterprising Tails by Marc Smith which I found well worth the purchase price 25+ years ago)
Sept. 16
You are ignoring the author of this comment. Click to temporarily show the comment.
> Perhaps the creation of complicated artificial systems,
> from Roman Club through Neapolitan and Blue, was
> a deliberate attempt to conceal cheating.

1. Canape does not equate to artificial

2. In North American at least, the elimination of that old black magic (arguably) lead to an increase in system complexity

3. If I have a side channel, the last thing that I want is a rigidly defined system that will draw attention to deviations
Sept. 16
You are ignoring the author of this comment. Click to temporarily show the comment.
> From R.W.'s point of view: Exactly what does it take to
> bring as proof so that everybody is satisfied?

I sincerely doubt that it is possible to bring proof such that everyone will be satisfied. At the same time, I don't believe that this is necessary.

I will note in passing that, back during the debates around analyzing hands and the like, I was one of the only people arguing from the get go that I do not believe that it is necessary to “break the code” that is being used to convict a pair of cheating or to punish them for this. Rather, I argued that a long and consistent history of strange card placements or the like should be sufficient in and of itself. However, I also recognized that this type of approach requires enormous care with respect to record keeping and analysis to avoid casting and overly wide net.

I find it highly amusing that I was attacks back in the day for this line of reasoning and that I am currently being attacked for the converse.
Sept. 16
1 2 3 4 5 6 7 8 9 ... 132 133 134 135
.

Bottom Home Top