Join Bridge Winners
All comments by Richard Willey
You are ignoring the author of this comment. Click to temporarily show the comment.
The USBF only exists to determine which teams represents the United States in internation competition.

The only events that it holds are various team trials.

The comparison that you are suggesting really isn't “apples to apples”
Aug. 20, 2014
You are ignoring the author of this comment. Click to temporarily show the comment.
Hate to always be playing the cynic, however, who precisely is asking for these changes to be made? I hear a number of folks on this forum complaining about masterpoints, however, I don't hear any clamor from the ACBL or the USBF.

The ACBL is a geriatric organization with a dying user base. The folks in charge aren't interested in rocking the boat. They are simply trying to extract as much as possible before the membership roles collapse.

Its fine and dandy to discuss what folks would like to do in theory, just don't expect anything to change before the board gets swept clean…

If you do want to try to do something in the short term, focus on getting the USBF. Ask them whether they need a better seeding system. See whether they will start collecting board records for all USBF matches.


Aug. 20, 2014
You are ignoring the author of this comment. Click to temporarily show the comment.
Maybe I have grown old and cynical. Maybe I have spent too much time working for large companies. However, I am highly skeptical that any third party “project” will get any traction.

I've learned the hard way not to engage on projects unless I see a clear ask from any proposed partner and strongly prefer to be able to require them to pony up something of value to demonstrate that they are serious.

I don't see the ACBL asking for a rating system.

I don't see the USBF asking for an algorithm to seed their events.

I don't see a partner willing to provide the raw data necessary to do a good job.

Doesn't sound like a recipe for success…
Aug. 19, 2014
You are ignoring the author of this comment. Click to temporarily show the comment.
If I were appointing a committee to try to attract more young players to the game, I wouldn't let a single person over the age of 30 on it. (Anyone else remember “Bridge is Cool”)

I'm also highly skeptical that a free one year membership to the ACBL is the way to go.

1. I don't believe that this would be considered valuable to the players

2. I don't think that the Bulletin represents the game the way you'd want to motivate and retain young players.
Aug. 17, 2014
You are ignoring the author of this comment. Click to temporarily show the comment.
> The crux of the problem is the image of bridge as “a boring
> game for old people.” No, I don't have the answers,
> but it's not true, and the image needs to be changed.

Take a look at the ACBL’s membership demographics. For better or worse, bridge in the US has become a game for old people. I’ll be 48 years old next week. I am younger than 97% of the ACBL membership. Half of the members are older than 72. I went to an ACBL regional three weeks back. I found the event extremely depressing to play in. Large numbers of the participants were infirm or ill or bewildered. You aren’t going to be able to attract young players to events like that.

If you want to try to salvage something, I don’t see many options other than segregating the game. Get serious about creating a game that will be attractive to young players

• Liberalizing the convention charts is an absolute necessity
• Recognize that 20 years aren’t going to want to socialize with retirees

Sponsoring events at gaming conventions and the like is a great idea, however, this won’t do a bit of good unless you have a critical mass of young players who already know the rules.
Aug. 17, 2014
You are ignoring the author of this comment. Click to temporarily show the comment.
> What is “forcing sponsors to use specified
> convention charts” mean?

I am willing to speculate on this one: The C&C is trying to authorize different sets of conventions for different types of events. For example, they don't want you to play Multi ins 2 board rounds, but do allow it in 7 board team matches.

However, within the ACBL there has been a long standing tradition of event organizers ignoring the Convention Charts. (This has been most noticeable with defenses to 1NT openings)

I suspect that the C&C is trying to crack down on this stuff. I don't wish them luck…

Aug. 16, 2014
You are ignoring the author of this comment. Click to temporarily show the comment.
Sorry, I might not have expressed myself well.

Let's assume that that you have two completely separate populations (no intermixing) with significantly different average strength.

While your ranking system might be very accurate within your population, its unclear whether it has any real predictive value between populations.

Aug. 16, 2014
Richard Willey edited this comment Aug. 16, 2014
You are ignoring the author of this comment. Click to temporarily show the comment.
Lior made some good observations that I was hoping to gloss over. In an ideal world, it would be nice if we could treat all draws as independent events drawn from the same distribution.

However, think about common situations where folks start swinging (for example, when a team is behind in the late stages of a match). Arguably, those teams might change their behavior to adopt strategies with a lower expected value but a higher variance.

Aug. 16, 2014
You are ignoring the author of this comment. Click to temporarily show the comment.
I wouldn't necessarily use results in the local club towards developing a rating system to predict performance in the Spingold.

Note the earlier comments regarding “network analysis”. A rating that is conditioned on your performance in one environment won't necessarily be accurate in another.



Aug. 16, 2014
You are ignoring the author of this comment. Click to temporarily show the comment.
Non-parametric learning systems work best with large data sets. They can be downright dangerous if your data is sparse.

Conceptually, you can think of developing a rating as a sampling problem. You have two pairs, each with unknown strength. You have some population that represents the outcomes if these two pairs were to play an infinite number of hands against one another. Each board represents a sample, drawn from this distribution. Your goal is to use these samples to draw inference about this distribution and then develop a rating that lets you describe your expectation.

Yes, you could just record the results of the overall match, but in doing so you're throwing away a lot of information and making your life harder.

Aug. 16, 2014
You are ignoring the author of this comment. Click to temporarily show the comment.
Bit of background information.

I'm a mathematical economist by training, with a lot of experience with predictive modeling. Up until a couple years back, I was the program manager for MATLAB's statistics system and spent lots of time teaching folks to use various types of stats and machine learning systems.

I also have a lot of experience playing online bridge dating back to the early days of OKB. Arguably, online bridge is the best environment to introduce “real” rating systems. (You have near perfect information about hand records and a player based that is younger and more tech savvy than the ACBL’s)

Based on my experience on OKB and BBO, I strongly believe that formally introducing any kind of ratings system would be highly problematic. FWIW, I view the problems as largely being social/political rather than technical.

Here are the big issues as I see them:

1. I believe that it’s possible to design a rating system that is accurate. I also believe that it’s possible to design a rating system that is easy to explain. I don’t believe that you can do both. Any accurate rating system is going to have some very complicated stuff hiding underneath the hood. (Kalman filters, a neural network, boosted decision trees, or some such). Your end users are never going to be able to understand how the algorithm operates. They are going to need to take things on faith. Which takes us to point number two…

2. Any accurate rating system is going to tell half your users that they are below average. Those users are going to get upset. They are going to look for ways to argue about the rating system and they are going to seize on the fact that it’s a black box as a source of discontent.

Simply put: If you use the rating system for anything serious, you’re going to have a never ending serious of arguments about the accuracy of the system. If you’re not going to use the system for anything real, why bother doing this at all.

With all this said and done, if you do decide t go down this path, here’s a few pieces of practical advice:

1. If you are going to build a rating system, start by doing a network analysis. Ratings are only going to be accurate within sub-populations that compete against one another. I’d want to start by understand what these sub-populations are and determining why I’d want to use a rating system for this group.

2. Focus on rating pairs rather than players. It’s a lot easier to evaluate a pair than individual players. If partnerships are long lived, rating partnerships is equivalent to rating players. If partnerships are short lived, there’s going to be too much noise and flux to generate a meaningful rating. (Please note, if you’re able to do a good job rating partnerships, you can decompose it to rate individual players)

3. Hire Mark Glickman. Glickman is one of the best at this sort of work, has an interest in it, and is experienced with bridge. He also has a lot of good code available that can be modified for bridge ratings systems. (He's been running the ratings system for the Chess folks for decades)

4. Garbage In, Garbage Out. The better your data sources and more complete your information, the more accurate your ratings.

If it were me, and I were starting up a ratings project, I’d focus on developing a seeding algorithm for top level events like the Spingold and the USBF team trials. Before doing anything else, I’d start by collecting a large corpus of hands that can be used to develop the ratings systems. Ideally, you want to be able to collect the scores for each and every board played in these events over the course of a year or two. If you can’t muster the institutional will to collate and store this information, than there’s much point investing in developing the methodology. In a similar vein, if you can’t get the organizations to agree to use your ratings to seed these events, what’s the point in doing the work…



Aug. 16, 2014
Richard Willey edited this comment Aug. 16, 2014
You are ignoring the author of this comment. Click to temporarily show the comment.
> Richard, when I served on the ACBL Board, someone
> repeatedly argued that ACBL membership would fall off a
> cliff when the average age hit 70? 72? – were you that
> person? If so, you have been forecasting the end of the
> world for the past couple of decades.

You are quite right. I have been saying for quite some time that the ACBL has an enormous demographic bulge and that membership numbers would contract precipitously when that bulge exits the system.

Here are some figures that Kevin Lane provided six months back.

10% decile = 60
20% decile = 65
30% decile = 68
40% decile = 70
50% decile = 72
60% decile = 74
70% decile = 77
80% decile = 80
90% decile = 84

Median = 72
Mode = 71

We're now at the point where the average age is past 72 and I think that the end times are near. I know that the average life expectancy of a 72 year male is only 12.75 years and that 80 years olds don't get out of the house much.

More formally, if you look at this chart, you'll see that 10% of the membership falls between the ages 72 and 74. This range is half that of 80-84, indicate a significant drop off once folks hit the age of 80. Indeed, the width of these bands starts increasing as soon as you pass the modal age.

So, let's check back in another 10 years and see what has come to pass. (Of course, I'm only 48. On average, I'll have another 31.17 years to see how this one plays out)
Aug. 14, 2014
You are ignoring the author of this comment. Click to temporarily show the comment.
> Most people who play frequently online think that compared
> to typical club games these days online is “real bridge”,
> that the competition is serious and challenging, and that
> it helps hone their bridge skills in ways that club games
> can't. That it's every bit as legitimate and serious as
> club bridge. And that's the position the ACBL should firmly
> support.

During the last year, I've played in more than a few local club games and tournaments in and around the greater Boston area after taking a decade or so off. I was horrified at the changes in the game over the past decade.

The standard of play seems to have fallen enormously. You can still find good players in Flight A, however, the number and proportion of such players is dwindling dramatically.

Simply put, club bridge isn't “real bridge” any more.
Aug. 14, 2014
Richard Willey edited this comment Aug. 14, 2014
You are ignoring the author of this comment. Click to temporarily show the comment.
> Because, a private corporation shouldn't try to strong-arm
> a non-profit, in my opinion. Or, attempt to bully
> individuals who are volunteering their time serving on its
> Board of Directors.

I don't see incorporation status as a relevant distinction.

I do know that over the past decade, BBO has done a hell of a lot more to promote and sustain the game of bridge than the ACBL. BBO is significantly larger than the ACBL and arguably a lot more significant to the future of the game than the ACBL.

We're I on the BoD, I'd be focused efforts on

1. Trying to strengthen my relationships with the BBO
2. Trying to clean up more pressing issues like ACBLScore, the web site, the C&C committee, and a demographic base that is fast circling the drain….

Aug. 14, 2014
You are ignoring the author of this comment. Click to temporarily show the comment.
Three quick points:

1. The reason that many players open offshape NTs partnering GIB has less to do with confusing GIB on defense as it does with GIB's relatively poor bidding. Its a lot safer being in the well defined NT bidding structure than it is risking a cue bidding sequence or some such.

2. These tactics are in no way, shape, or form unique to partnering with GIBs. Many of us bid precisely the same way opposite live partners in Indy's and the like

3. I would far rather partner a GIB than a random BBO player in an Indy




Aug. 12, 2014
You are ignoring the author of this comment. Click to temporarily show the comment.
As I recall, regulations regarding psyches typically hinge on whether or not partner is more likely to expect the psyche than the opponents.
Aug. 12, 2014
You are ignoring the author of this comment. Click to temporarily show the comment.
> it is not discrimination to recognize that F2F bridge play
> is significantly different from online bridge play and to
> keep separate masterpoint lists for both categories of
> play. Different environment, different controls, different
> skills required to win. So why not celebrate the
> differences but keep separate records/masterpoint lists?

I propose that we call this “Separate, but Equal”
Aug. 12, 2014
You are ignoring the author of this comment. Click to temporarily show the comment.
For better or worse, the online version of the game is missing a wide number of infractions that are endemic to the online version. Think revokes, leads out of turn, misdeals, etc.

I think that it is almost definitional that a F2F player would have more familiarity with these aspects of the game.
Aug. 12, 2014
You are ignoring the author of this comment. Click to temporarily show the comment.
To my knowledge, the robot races don't award ACBL master points.
Aug. 11, 2014
You are ignoring the author of this comment. Click to temporarily show the comment.
Grossly dysfunctional. We'd be much better off outsourcing things to the Brits.
Aug. 10, 2014
.

Bottom Home Top