Join Bridge Winners
All comments by Richard Willey
You are ignoring the author of this comment. Click to temporarily show the comment.
Spearman's rho is our friend

arguable you might want to weight things based on the ordering. (Its more important to get the top seeds right rather than the bottom)
Sept. 12, 2013
You are ignoring the author of this comment. Click to temporarily show the comment.
I'm sure that the following suggestion will get me lambast, but the economist in me says that we should auction of seeds to the highest bidder. Use the funds collected to fund a prize pool.

Teams with rich sponsors might overbid, but they would (essentially) just be subsidizing the rest of the participants. Teams lacking sponsors simply don't bid, accept low seeds, and have a positive expected value for entering the tournament.
Sept. 12, 2013
You are ignoring the author of this comment. Click to temporarily show the comment.
I don't disagree. At the same time, I think that the primary advantage of this sort of system is matching people for short periods of time. How do I find a player who I can stand partnering for a dozen boards on BBO?
Aug. 29, 2013
You are ignoring the author of this comment. Click to temporarily show the comment.
Thanks for the thoughtful reply. FWIW, I'm not sure whether the existing Bridgewinner polls would (necessarily) provide the most interesting results. These polls certainly identify hands where there is some doubt regarding the best course of action. At the same time I'm not sure how common the hands are. In contrast, basic questions like

Would you chose a 1NT rebid or raise partner's major with three pieces?

Do you prefer to rebid 1NT or show a 4 card major with a balanced hand after 1C - 1D?

What do you prefer to open with 5-5 in the blacks (I thought this one was settled until I saw “In the Well” last weak)

What do you open with 5 hearts and a 5332 15 count?

What do you open with 5 clubs and a 5332 14 count?

What do you open with a 3451 15 count?





Aug. 28, 2013
You are ignoring the author of this comment. Click to temporarily show the comment.
The terms are placeholders. I'm fairly indifferent what expressions get used so long as there aren't any value judgements associated with them. I've seen far too many arguments start because people don't agree with their skill levels and insist that they are experts rather than intermediates. It seems a less likely that someone will get bent out of shape because they're classified as a meerkat rather than a bison…. (With this said and done, there was a particularly memorable episode of Better Off Ted in which Veridian Dynamics experimented with team building…)

With respect to the actual classifier… I haven't trained a classifier for bridge largely because I don't have the necessary corpus of hands to work with. I've done this for other problems which are very similar to this one. I'm quite convinced that the methodology is valid. However, I don't have the necessary raw data to work with…

Aug. 28, 2013
You are ignoring the author of this comment. Click to temporarily show the comment.
Here's my “pipe dream”:

I'd like to see the ACBL (or any NBO) partner with BBO to develop a mechanical Turk for analyzing questions involving UI and the like…

1. “Bob” is a tournament director in Newton, MA. Bob is directing a team’s event on Friday evening. Bob gets summoned to a table to deal with a bidding irregularity. There was a hesitation in the bidding. Bob needs to determine whether 6H is a logical alternative. Bob needs to generate a decision tonight, since the team that loses doesn’t want to show up at 9:00 AM just to find out they aren’t playing.

2. Bob would like to poll a bunch of players; however, he can’t interrupt the game in progress. Moreover, Bob is knows that most people are going to blow right out of the Marriott as soon as the game ends. Thankfully, BBO provides a solution…

3. Bob copies the auction and marches back to his laptop. Bob logs into his ACBL account on BBO and enters the hand record and the auction onto a web page. Bob also indicates that he wants 100 tables to duplicate “this” hand. The web page communicates with the BBO server and places the hand record into the queue.

4. BBO distributes the hand to a large number of tables. Each time the hand gets played, the information gets sent to a secondary process. This process (automatically) excludes each and every hand that doesn’t match the “Newton” bidding up to the hesitation. Eventually, the process notes that it has 100 hands that DO match the “Newton” bidding. The process then sends Bob and message and indicate that 75% of the pairs bid 6H, 20% of the pairs passed, and 5% of the pairs bid 6C.

5. Bob is now able to use this information to determine that 6H IS a logical alternative, despite the hesitation…



Aug. 28, 2013
Richard Willey edited this comment Aug. 28, 2013
You are ignoring the author of this comment. Click to temporarily show the comment.
The academic discipline of game theory differentiates between “pure” strategies and “mixed” strategies. Pure strategies are deterministic. Players choosing a pure strategy follow a predictable course of action. In contrast, mixed strategies deliberately incorporate random action. The simplest example of a mixed strategy in bridge is probably restricted choice where a defender will randomly chose which of two cards to play.

Mixed strategies can also be applied to the design of bidding systems. Players applying a “pure” bidding strategy will always chose the same bid bid with a given hand. In contrast, players employing a mixed bidding strategy allow deliberate randomization. Consider the following example taken from Bridge My Way by Zia Mahmood. You hold

S AQJ3
H K5
D 873
C A653

The auction starts

1H – 1S
3S - ???

and you need to chose a rebid. Zia advocates a bidding style in which players should randomize between 4C and 4D cuebids. Zia never goes so far as to discuss probabilities, but hypothetically he might chose a 4C cuebid 80% of the time and a 4D cuebid 20% of the time.

Few if any Zonal authorities incorporate mixed bidding strategies into their regulatory structures. Instead, regulators attempt to sidestep the issue using the concept of a psychic call. Regulators and players pretend that psychic calls are “deliberate and gross misstatements of honor strength or suit length”. In actuality, so-called psychic calls are a subset of a more complex meta-agreement involving mixed bidding strategies.

I argue that neither players nor regulators are served by this pretense. Complete disclosure can never be achieved unless the regulatory structure matches the actual strategies employed by players.
Aug. 23, 2013
You are ignoring the author of this comment. Click to temporarily show the comment.
Is the variance associated with a method reasonable grounds for restricting sanction?

Assume that you have two different 2H opening bids.

• Both bids have the same expected value per board
• The complexity of the suggested defenses is comparable
• The variance in the board results is significant higher for method A compared to method B

Is it reasonable to restrict approval for method A because its use can disadvantage stronger card players / defenders?
Aug. 23, 2013
You are ignoring the author of this comment. Click to temporarily show the comment.
Does it make sense to replace the legal construct of a “psyche” with the game theoretic concept of a mixed strategy? (This question applies to both disclosure as well as regulation)

On a meta level, can a legal structure operate effectively if it is using constructs that don’t accurately describe player’s behavior?
Aug. 23, 2013
You are ignoring the author of this comment. Click to temporarily show the comment.
> What happened to our ability to use our booking power
> to get fabulous room rates?

The market got more efficient…

Hotels (and for that matter airlines) are much better at selling “surplus” rooms. The rise of travel sites like Orbitz and the like provide a much more effective way for the hotels to price discriminate.
Aug. 22, 2013
You are ignoring the author of this comment. Click to temporarily show the comment.
A block bootstrap (If I knew how to parameterize it)
Aug. 21, 2013
You are ignoring the author of this comment. Click to temporarily show the comment.
It's unclear to me whether board results are independent events…

If I get a really bad result on board 15, it might throw off my game. If the opponents are down by a significant amount with 16 boards left to go, they might start swinging.

If board results are independent of one another, the math is a lot easier.
Aug. 21, 2013
You are ignoring the author of this comment. Click to temporarily show the comment.
>Still, I'm still not convinced that these sorts of >probabilities are the best way to think about it.

My biggest concern is auto-correlation across boards…
Aug. 21, 2013
You are ignoring the author of this comment. Click to temporarily show the comment.
Even if there were no seeding, I'd still expect later rounds to be closer than early rounds.

Maybe the real problem is that the standard deviation is too high relative to the mean.
Aug. 21, 2013
You are ignoring the author of this comment. Click to temporarily show the comment.
Are these figures constant across all the rounds of a major tournament? Assume once again a 5 round KO. This time assume that there is a reasonable attempt at seeding.

On average, the difference in the skill level between teams should be smaller in later rounds than in early rounds. I would expect that mu would decrease as the tournament progressed…
Aug. 21, 2013
You are ignoring the author of this comment. Click to temporarily show the comment.
Hi Bernard

Sorry for any confusion. I should have been more clear in my original post.

When I was discussing optimizing the number of boards, I as implicitly assuming trading off boards between different rounds of a multi-round event.

Lets assume I am running a five KO round event, with a total of 320 boards. How should I distribute these boards across the total set of rounds? (I'd be shocked if 5 rounds of 64 boards turned out to be optimal in anything other than an extremely tight field)
Aug. 21, 2013
You are ignoring the author of this comment. Click to temporarily show the comment.
> My (unpopular, I'm sure) solution would be to
> shorten matches to 56 boards, or even 48.

It would be relatively easy to model different types of tournament designs.

Treat each hand in a long team match as an individual sample. Determine the mean and standard deviation. Repeat for all of the matches across all the rounds of the event.

Next, determine how many boards each round should be to maximize the chance that best team wins (subject to some set of external constraints - the event can't last for more than 5 days, players can't compete for more than 10 hours a day, whatever)
Aug. 21, 2013
You are ignoring the author of this comment. Click to temporarily show the comment.
It will be interesting to see if you are correct…
Aug. 20, 2013
You are ignoring the author of this comment. Click to temporarily show the comment.
That's a curious assertion given that the C&C has repeatedly stated that Midchart legal openings are not allowed unless there is an explicit defense listed in the Defensive Database…

Then again, consistency was never their strong suit…
Aug. 19, 2013
You are ignoring the author of this comment. Click to temporarily show the comment.
All well and good, but until they manage to publish a suggested defense to the a 2D opening showing the majors or a 2H opening showing 5+ hearts and 5+ cards in another suit neither bid is legal…
Aug. 19, 2013
.

Bottom Home Top