BBO Discussion Forums: player skill rank - BBO Discussion Forums

When mainly playing in a closed group of players (main hall, team matches) the rating of bboskill can at most show you the rating within this group. Other habits like playing as sub or when being tired or coaching or only playing with a good partner or playing while the baby is around or ... will have a big influence too. It is too simple to just take the results of my-hands to create rating of skill.

billw55, on 2014-May-27, 07:24, said:

Unfortunately, skill is not normally distributed. Think e.g. about the distribution of Elo ratings among chess players. There are very few high ELO ratings and many low Elo ratings (and even more without an Elo rating).

The same in soccer. There is Ronaldo, Messi, ..., a few very good players, more good players, many more quite allright players, lots of ok players, tons of so-so players, masses of poor players.

I do not know any field where skill is normally distributed.

Rik

Trinidad, on 2014-June-02, 07:49, said:

You should rethink that.
If you take the ELO to the x-axis and the number of player with a specific ELO number as y-axis you will get a normal distribution.
At the right end of your x-scale there are only few player with a high number.
The bulk of player with have some mediocre number.
Of cause you have to leave those people out, who don't have a number at all.

I don't think the question whether skill level is normal distributed or not makes so much sense. It likely depends how it is measured.

Say it is measured as expected IMPs per board that you would win playing in a large crossimp indy with participation of a representitive sample of bridge club members from all over the World. It would't surprise me if you would get something close to a normal distribution. OTOH if you measure it in terms of price money won it will be heavily right-skewed.

helene_t, on 2014-June-02, 08:37, said:

How something is distributed always depends on what plotted along the x-axis. Any continuous distribution can be corrected to a normal distribution with a normalisation function.

Take, as an example, the distribution of the diameter of the particles in a powder (sand, powdered sugar, etc.), measured in meters (mm, µm, whatever). In most powders, this distribution is heavily skewed: there are many small particles and few large ones. However, if you plot the log of the particle size, then many natural particle size distributions are normal. This is typically not true for the powders you have at home since they are sieved, cutting off the distribution.

I think that if you could organize a pairs tournament where every single bridge player participates (let's not get into the problem of making pairs), ranging from the aunt Millies of this world to the national team of Monaco, I would think that the bulk of the field would end up with 49.9% and the Monaco players with 80%. In practice, most bridge tournaments are "sieved" and the distribution is cut off: Aunt Millie doesn't play in the Bermuda Bowl and Geir Helgemo doesn't play in Aunt Millie's kitchen. But I think there are many more kitchens than Bermuda Bowls.

Rik

Of course, in total there are many more very bad players than very good players. Many players who only barely leaned the game and rarely play, etc.

I was thinking more of the player population that seeks out and participates in competition. At tournaments for example - most of the bad players don't go, and most who do go are least marginally competent (insert jokes and stories if you like, it might improve the thread). So this population might look more normally distributed.

Trinidad, on 2014-June-02, 07:49, said:

Among United States Tennis Assn league tennis players (330,000 strong), skill ratings are pretty close to normally distributed. Of course, this does not include the masses of people who hit tennis balls around once a year and would (almost) all be included at the lower end of the skill range, but it's a good indication for people who might call themselves "tennis players".

If you only selected players from a population of those who have reached their equilibrium state of competence, then perhaps the resulting curve might look close to a normal distribution.How you would achieve that selection is of course the big IF.Where I fit in that spectrum is a little odd. Against human opposition I am certainly now in decline. Against robots I regard myself as ascending.

It is not you jack - senility is setting in for the robots.

Suppose you are captain of a team of four strangers, two advanced ones and two intermediate ones. What would partnership compatibility mean in that context? If "intermediate" means "understanding one's own system" while "advanced" means "understanding a wide range of systems", probably you should pair each intermediate with an advanced partner. Unless, maybe, if there is a local standard which both intermediates can be assumed to play.

helene_t, on 2014-June-03, 07:58, said:

If there is a national standard established. When I refer to the clubs where I play I can tell you that there is a wide variety of systems and no standard system (at least 4 versions of Forum, Große-Kleine-Treff, Kleine Treff, so called Standard American, Acol, Swiss Acol, 2/1 Turkish style, Precision, even Truscott found the way to our club last week). So where is the compatibility bonus coming from?

helene_t, on 2014-June-03, 07:58, said:

That's kind of what I was getting at earlier. An advanced player knows enough that they can play comfortably with a variety of different partners -- he can usually adjust to their system. An intermediate player knows a limited amount of stuff, and is still building up his repertoire.

But this isn't necessarily the only way to measure skill. Consider someone who has been playing kitchen bridge for a lifetime. They may know just a handful of conventions, but they could be extremely successful with them. Furthermore, cardplay is the same for everyone -- it's harder to measure what you "know" about it (although something like Bridge Master 2000 could be a good way to assess this).

helene_t, on 2014-June-03, 07:58, said:

This doesn't seem like a very useful measure. Even a non-player could understand a wide range of systems.

Vampyr, on 2014-June-03, 15:10, said:

What does that mean? It sounds like saying someone speaks French because they've memorized an English-to-French dictionary, although they've never actually had a conversation in the language.

Trinidad, on 2014-June-02, 09:08, said:

It is funny that you did not rethink the assertion that "any continuous distribution can be corrected to a normal distribution ...", after your powder example.

First of all, by definition a Gaussian (normal) distribution is defined as a function over the domain of all real numbers.

In any event, the universe is full of non-Gaussian distributions - The energy density as a function of frequency of a black body and the energy levels of electrons in an atom spring immediately to mind. In fact, any measurement that is bounded for any reason, for example temperature, can't fit a Gaussian distribution. The age of all of the members of any particular species - yes, non-Gaussian.

That said, I can't imagine a reason for imposing a gaussian distribution model on a player skill ranking at anything - except that perhaps the person making the measurement plans for the results to be Gaussian and is willing to skew the measurements in some way to "make them fit" his postulated distribution function.

barmar, on 2014-June-03, 15:55, said:

Understanding bidding systems does not indicate that a person is even a good bidder, judgment being more important in this area; nor does it indicate that the player can declare or defend competently.

Vampyr, on 2014-June-03, 21:06, said:

Quite so. When I learned to play bridge I did so alone from books. It was many years before I actually played with people. During that time I wrote out the bidding systems and understood how they worked, indeed even combined ideas I liked from various places and created my own. It was probably similar to the way computers operate bidding rules. But with no practical experience you could say that I had no judgement whatsoever and was truly awful at card play.

Now I still think about system structures and have the computer-like rules in mind but hopefully at least have added a little bit of judgement. And improving the system rules helps the judgement since the hands are better defined. Sadly the card play is still pretty terrible but perhaps one day the posts from BBFers will rub off on me.

I don't think the criteria are intended to be applied indiscriminately. They assume the typical way that players advance in experience. They start off with little playing experience and instruction in a few basic conventions. Over time, they get more experience playing, and usually also learn more conventions to improve their bidding, or at least be able to play competently with and against players who want to play those conventions (e.g. you might not agree that Flannery is a good convention, but being able to play it means Hamman is a potential partner).

So knowledge of many conventions doesn't necessarily mean that you're a great player, but being able to use them effectively is a strong indicator.


Thus, my analogy with languages. It's not enough to know the meaning of words in a language, you need to be able to have meaningful, fluent conversations to be considered to "know" the language.

Zelandakh, on 2014-June-02, 07:01, said:

pfft http://bboskill.com/user/phantomsac

larlar, on 2014-June-07, 22:49, said:

Maybe I'm a little dense. Why is that funny? Because of 0 adjustment?