Big news

[Yixin]

originally posted by VLM:
So, you know these shadowy guys in Singapore?

BTW, it is very doable to create a single score from a multi dimensional problem. One that is very predictive as well. Done all the time.

Just because these asshats don't do it doesn't make it a terrible idea.

No.

I used PCA for yield curve modelling, once upon a time. Very predictive, of course, except when it wasn't. I don't have the same fluency in statistics (and the underlying mathematics) that you do, but I submit the most interesting bits about wine (or more precisely, drinking wine) and yield curves are when observed relationships break down. There are the usual GIGO objections, of course.

Just because it's doable doesn't make it a good idea.

originally posted by Claude Kolm:
Whether it's one of those scales or 100 is immaterial...

I respectfully submit that it isn't (and that was the point of my sentence). The 5 star scale is quite different from a 100 point scale precisely because it is less granular. I sometimes help out in wine competitions, and judges are far more likely to be kicked out when we run through the test controls using a 100-point scale. You might not believe it, but not many people can reliably score 2 glasses from the same bottle of wine within a flight of 5/6 glasses. These 'inaccuracies' are elided on cruder scales, but cruelly exposed by their finer-grained cousins.

 

[Claude Kolm]

originally posted by Yixin:

originally posted by Claude Kolm:
Whether it's one of those scales or 100 is immaterial...

I respectfully submit that it isn't (and that was the point of my sentence). The 5 star scale is quite different from a 100 point scale precisely because it is less granular. I sometimes help out in wine competitions, and judges are far more likely to be kicked out when we run through the test controls using a 100-point scale. You might not believe it, but not many people can reliably score 2 glasses from the same bottle of wine within a flight of 5/6 glasses. These 'inaccuracies' are elided on cruder scales, but cruelly exposed by their finer-grained cousins.

You've never been to a wine tasting where you're supposed to rank the wines in preference from 1 to 8? You may be qualified to say that you can't do it, but I think most people who attend say that they can and that who are you to say that they can't? Even if they give the same score to more than one wine on a scale of 0 to five, they generally are able to say which ones they prefer, which would support further "granularity."

Moreover, zero to five stars almost always includes half stars, so we've got "granualrity" of nine, not really that much different from 80 (or 85 or 88, which seems to be the bottom range a lot of people on the internet use, even for terrible wines) to 100.

Moreover, scoring on 100 breaks down quite easily to assigning values to different elements fo the wine and then adding them up (or, I think, more easily, subtracting from a perfect score for each element and then adding them up). When you do that, also, the "granularity" of the 100 point system is much less because you are actually doing a composite.

Finally, rating a wine one time 88 and the next time 90 sounds to most people more consistent than 3-1/2 one time and 4 the next because with the lesser "granularity" you exaggerate the differences for wines on the cusp.

 

[SFJoe]

As Yixin notes, the reproducibility of human measurements is not that great, from day to day or from glass to glass. Critics (of whatever professionalism) really don't run 10 or 20 control glasses through the day to test this, at least none I know.

 

[Ian Fitzsimmons]

originally posted by Ian Fitzsimmons:

originally posted by VLM:
[...]

Haven't read Silvers book, but from my understanding, his methods are pretty basic OLS regression.

I haven't run across a 'methods' section, yet, but this sounds right. His deal, I think, is not so much sophisticated technique, but acquiring sufficient insight into the process of interest to count and properly weigh the things that best indicate outcomes. Also, to tolerate uncertainty, and play the odds over the long-run. Also, find ways to build up N (e.g., poll aggregation).

I realized after your point is Silver isn't doing multi-dimensional analysis. I was tunnel-visioning on the 'prediction' part when I wrote - apologies.

 

[Claude Kolm]

I find that 95+% of the time, when I taste a wine blind that I've had before (even double blind or forgetting that I've had it before only to later discover a note), I'm within 2 points of the previous ranking, often times dead on. Of the remaining times when they are different, it is often due to an externality such as a cork problem.

When I first got interested in wine I had serious doubts about the consistency of perceptions, but luckily for me, there were two blind tastings of California Cabernets and Chardonnays (this was late 1970s) that were going to be repeated in consecutive weeks. I went to all four tastings and found that my rankings were identical -- convincing me that there was indeed something consistent about perceptions of wine.

 

[Ian Fitzsimmons]

originally posted by VLM:

originally posted by Jeff Grossman:

originally posted by VLM:
Decide on the dimensionality. Get items to measure them. Item response or structural equation model. Expected a posteriori scoring on any scale you want ~N(µ,sd). Done.

Recently read an interview with Chomsky in which he did not agree that careful watching of the data will generate scientific results.

...ducking the flying blubber...

I don't get this.

I assumed this was in anticipation of flak from Rick Slicker.

 

[Peter Creasey]

originally posted by Claude Kolm: they generally are able to say which ones they prefer, which would support further "granularity."

Claude, Bravo!

I'm glad to hear people take a stand in favor of the rating system(s). People can ignore the rating system(s) if they wish.

On the other hand, for many folks, the rating system(s) add an additional dimension that tends to clarify whatever verbiage the reviewer may have seen fit to assign to the entity, in this case wine.

Many people use verbiage that, in context, can be taken as positive, neutral, or even negative. The rating system(s) put the comments into perspective IF the reader wishes to avail himself of that additional (valuable?) factoid.

. . . . . . . Pete

 

[Claude Kolm]

The thing to keep in mind about rating systems that they are just a mechanism for ordering preferences. They have no inherent meaning. So if one orders formally (with a score) or informally (without), it's really all the same.

 

[Yixin]

Claude, this is old ground (to wit, the precision of the 100-point scale, rather than points themselves) for both you and I, and perhaps better discussed in another thread.

Ordinal rankings and cardinal values are different.

I don't often see 0 or 1 stars on a 5 star scale; in fact most of them seem to start at 3 stars, regardless of whether half-stars are used or not. So there are fewer gradations of 'quality' than in a typical 100 point scale. And that is precisely why we tend to disqualify fewer judges when using a 5-star or even 20 point scale than when using a 100-point scale.

I'm quite sceptical about component scoring because it's a simple summation across various dimensions (or components, if you prefer), which brings up all sorts of problems with weighting and calibration.

But let a hundred points bloom, and don't let a Singaporean schnook get to you.

I wonder how much the other writers knew about the deal, and what they're thinking now.

 

[Claude Kolm]

originally posted by Yixin:

I wonder how much the other writers knew about the deal, and what they're thinking now.

He's 65. Anyone who didn't think it was going to happen sooner or later (and probably sooner) was being naïve or in denial. Fact is, he probably could have gotten a lot more out of the deal five or ten years ago when his brand was a lot more important.

 

[Yixin]

The use of score ranges (e.g. 91-95 instead of 93 to indicate a ±2 margin) to indicate some elasticity is something I would like to see more of, if one must score. Why use it only for barrel samples (aka young wine), as if the wine stops developing once it's bottled? One can still express a view about relative quality (e.g. 90-94 vs. 91-95) while being transparent about variability (of context, of taster, whatever). It strikes me as being more honest.

 

[Yixin]

originally posted by Claude Kolm:

originally posted by Yixin:

I wonder how much the other writers knew about the deal, and what they're thinking now.

He's 65. Anyone who didn't think it was going to happen sooner or later (and probably sooner) was being naïve or in denial. Fact is, he probably could have gotten a lot more out of the deal five or ten years ago when his brand was a lot more important.

There's no doubt things had to change (because of his age), but I'm not sure any of them (besides Ms. P-B) quite envisaged this.

But yes, he probably could have gotten more if he had sold earlier.

 

[Claude Kolm]

originally posted by Yixin:
The use of score ranges (e.g. 91-95 instead of 93 to indicate a ±2 margin) to indicate some elasticity is something I would like to see more of, if one must score. Why use it only for barrel samples (aka young wine), as if the wine stops developing once it's bottled? One can still express a view about relative quality (e.g. 90-94 vs. 91-95) while being transparent about variability (of context, of taster, whatever). It strikes me as being more honest.

I think that with a little thought anyone has to build that in inherently. Again, the scores have no inherent meaning, they're just an ordering mechanism.

 

[SFJoe]

originally posted by Claude Kolm:
I find that 95+% of the time, when I taste a wine blind that I've had before (even double blind or forgetting that I've had it before only to later discover a note), I'm within 2 points of the previous ranking, often times dead on. Of the remaining times when they are different, it is often due to an externality such as a cork problem.

When I first got interested in wine I had serious doubts about the consistency of perceptions, but luckily for me, there were two blind tastings of California Cabernets and Chardonnays (this was late 1970s) that were going to be repeated in consecutive weeks. I went to all four tastings and found that my rankings were identical -- convincing me that there was indeed something consistent about perceptions of wine.

I would say that the literature of human perception would pretty much brand you as unique.

Mazel tov!

 

[Tom Glasgow]

originally posted by Claude Kolm:

originally posted by Yixin:

I wonder how much the other writers knew about the deal, and what they're thinking now.

He's 65. Anyone who didn't think it was going to happen sooner or later (and probably sooner) was being naïve or in denial. Fact is, he probably could have gotten a lot more out of the deal five or ten years ago when his brand was a lot more important.

65 isn't so old, especially for those with sizable egos who run/own the show.
But of course, there's always the possibilty of undisclosed health issues.

 

[Claude Kolm]

originally posted by Tom Glasgow:

originally posted by Claude Kolm:

originally posted by Yixin:

I wonder how much the other writers knew about the deal, and what they're thinking now.

He's 65. Anyone who didn't think it was going to happen sooner or later (and probably sooner) was being naïve or in denial. Fact is, he probably could have gotten a lot more out of the deal five or ten years ago when his brand was a lot more important.

65 isn't so old, especially for those with sizable egos who run/own the show.
But of course, there's always the possibilty of undisclosed health issues.

He clearly never really enjoyed travelling (at least out of the US), has lost whatever palate he once had (in part due to the natural aging process), and had been rapidly loosing influence, perhaps with Bordeaux and Ch“teauneuf excepted, so what's in it for him to continue?

Who knows, he might even be loosing that exceptional memory of every wine he's ever tasted. ;)

 

[Tom Glasgow]

No argument with your points, but Parker wouldn't be the first to stay too long.

 

[Claude Kolm]

originally posted by SFJoe:

originally posted by Claude Kolm:
I find that 95+% of the time, when I taste a wine blind that I've had before (even double blind or forgetting that I've had it before only to later discover a note), I'm within 2 points of the previous ranking, often times dead on. Of the remaining times when they are different, it is often due to an externality such as a cork problem.

When I first got interested in wine I had serious doubts about the consistency of perceptions, but luckily for me, there were two blind tastings of California Cabernets and Chardonnays (this was late 1970s) that were going to be repeated in consecutive weeks. I went to all four tastings and found that my rankings were identical -- convincing me that there was indeed something consistent about perceptions of wine.

I would say that the literature of human perception would pretty much brand you as unique.

Mazel tov!

I don't know what literature you are looking at, but I'll guess the subjects were random sampling of the population, not people who daily did various evaluations (taste, sight, hearing, touch) according to a specific methodology. If that is so, then the studies are irrelevant.

I might add that the methodology gives me consistency in ordering preference of wines, but no special insight as a blind taster as to identity -- I'm not particularly perspicacious for that exercise.

 

[SteveTimko]

originally posted by Sharon Bowman:

originally posted by SteveTimko:
It captures the audience who are looking for a specific topic as opposed to someone who comes to a website every day and is too cheap to pay for a subscription.

But why is the random searcher a more valid reader than the avid cheapskate? Their money is of equal value.

Google puts them in direct competition with other web sites. For those people, the idea is, we'll let them in. That would be money lost.
But you want people who read the site every day to pay, since they actually generate revenue. The people who come in with Google clicks generate almost no revenue, but better our site gets almost no revenue than some other site.

 

[Jonathan Loesberg]

originally posted by Peter Creasey:

originally posted by Claude Kolm: they generally are able to say which ones they prefer, which would support further "granularity."

Claude, Bravo!

I'm glad to hear people take a stand in favor of the rating system(s). People can ignore the rating system(s) if they wish.

On the other hand, for many folks, the rating system(s) add an additional dimension that tends to clarify whatever verbiage the reviewer may have seen fit to assign to the entity, in this case wine.

Many people use verbiage that, in context, can be taken as positive, neutral, or even negative. The rating system(s) put the comments into perspective IF the reader wishes to avail himself of that additional (valuable?) factoid.

. . . . . . . Pete

Those of us who write tend to think that using verbiage is something different. Perhaps one wouldn't need to score if one wrote more and used verbiage (or, even worse, wordsmithed) less.