PDA

View Full Version : KCBS weighting coefficients


comfrank
05-14-2009, 11:09 AM
The KCBS weighting coefficients for appearance, taste, and texture are, respectively, 0.5714, 2.2858, and 1.1428. If you do the arithmetic, you will find, as I am sure most of you already know, that these numbers imply that texture is worth twice as much as appearance, and taste is worth twice as much as texture (and thus four times as much as appearance). Well...not quite. Taste is worth 2.000175009 times as much as texture. If it were to be worth exactly twice as much, the fourth digit after the decimal in the taste coefficient should be a 6, not an 8. Why did they make it an 8 instead of a 6? I'm guessing so that a set of three nines would be 36 rather than 35.9982.

I don't have a problem with weighting taste twice as much as texture which is in turn twice as much as appearance. What I have a problem with is the coefficients implementing this weighting scheme. Two point two eight five eight?!?!? Come on, you've got to be kidding me. Why on earth would anyone choose such ridiculous coefficients? I've got a better set. How about these: 1, 4, and 2.

With the current system, scores ramge from a total of 0 (for an unjudged entry) to 180 (for five sets of perfect nines). The problem is that to rank order entries in between these two extremes you need to compute scores to four digits past the decimal. And even if a computer does the actual computation, that's still a lot of numbers on the score sheets. If you used 1, 4, and 2 as weights, then a judge giving three 9s would contribute 63 points to a team's score. Five sets of perfect 9s would equal 315 as the top possible score for an entry, and you would never need to go past the decimal point to rank an entry. Moreover, taste could now be *exactly* twice as important as texture.

Even if you don't like 1, 4, and 2 as weighting coefficients, you could certainly choose more sensible ones such as .5, 2, and 1, or .25, 1, and .5.

Could someone who has more history with the organization please explain to me the reason for these goofy coefficients?

Thanks,
--frank in Wilson, NY

Jorge
05-14-2009, 11:14 AM
I'm not seeing the flaw in the system. As it exists, a tie is much less likely than the system you are proposing.

KC_Bobby
05-14-2009, 11:19 AM
I like it because taste is king. Taste points will slightly edge tenderness and appearance points this way - a 798 will beat a 989 instead of being tied (33.7144 compared to 33.7142)

Jacked UP BBQ
05-14-2009, 11:23 AM
Run a fake comp on paper with your numbers with around fifty teams and see if you get a lot of ties, if not, should work nicely.

crd26a
05-14-2009, 11:27 AM
I'm not seeing the flaw in the system. As it exists, a tie is much less likely than the system you are proposing.


Winner winner chicken dinner

By having the multiplier out to the four decimal places, it helps to eliminate potential ties along the way, not to mention, scores would be crazy high, as 3 9's would become 9 + 18 + 36 = 63.

Whereas I would agree it would be easier to do the quick math in your head, this allows for an easier selection and tie breaker setup.

BBQchef33
05-14-2009, 11:35 AM
yup.. reducing the chance of ties is the key. Do a mock contest to 0 decimal places and see what happens.

Alexa RnQ
05-14-2009, 11:41 AM
Having been tied to four decimal places, let's just say I have no problem with the system which allows that to occur as infrequently as it does.

ique
05-14-2009, 11:43 AM
Could someone who has more history with the organization please explain to me the reason for these goofy coefficients?

Thanks,
--frank in Wilson, NY

Why do the goofy coefficients matter? As long as the scoring algorithm is weighting the right things (taste most important, appearance least) and not producing ties who cares that the cofficients are not nice round numbers?

Jacked UP BBQ
05-14-2009, 11:46 AM
Why do the goofy coefficients matter? As long as the scoring algorithm is weighting the right things (taste most important, appearance least) and not producing ties who cares that the cofficients are not nice round numbers?

Algorithm - such big vocabulary:lol:

stlgreg
05-14-2009, 11:48 AM
Exactaly, There is 81 different scores in the scoring system you are proposing.
In the KCBS system there is 354 different combination of scroes a judge can assign to an entry.

In Edwardsville this past weekend, GC was decided by .0002. That means if they were using your system they would have tied. However, by weighing taste just a smidge more, they were able to declare a winner - the one that tasted just a bit better than the next guy.

PatioDaddio
05-14-2009, 11:52 AM
I don't have a problem with the weighting, but the goofy and unintuitive 2-9 scoring scale makes zero sense to me. The look on people's faces when you describe the scale speaks volumes. Almost every person that I've explained it to looks at me like a dog that just heard a strange noise.

Here's an idea... How about, oh I don't know, 0-10 with zero being, you guessed it, no score/DQ.

Just venting,
John

Scottie
05-14-2009, 12:05 PM
In Edwardsville this past weekend, GC was decided by .0002.


It was .0008, but who's counting.....:eek:

stlgreg
05-14-2009, 12:08 PM
Oops, just ran through the math. There is only 40 combinations of a score doing weighing the scores with a 1, 2 and 4. Logistically, there is even fewer. How many times does 111 or 222 get used? Some really you are looking at a pool of 34ish scores that will actually get used. Take that to any contest with more than 30 teams and see how many ties really show up.

stlgreg
05-14-2009, 12:08 PM
It was .0008, but who's counting.....:eek:
Sorry. I didnt mention names :-P

stlgreg
05-14-2009, 12:11 PM
I don't have a problem with the weighting, but the goofy and unintuitive 2-9 scoring scale makes zero sense to me. The look on people's faces when you describe the scale speaks volumes. Almost every person that I've explained it to looks at me like a dog that just heard a strange noise.

Here's an idea... How about, oh I don't know, 0-10 with zero being, you guessed it, no score/DQ.

Just venting,
John

It is really 1-9 but thats neither here nor there. I broke down the system mathimatically last year (maybe I should change my name to NerdQ). It may not be intuitive but I could tell each step sure was thought out and just not thrown in willy nilly.

My thought....I think 10 is not used for entry issues. It is easier on the reps and probably the program itself if they can keep it to 1 digit.

KC_Bobby
05-14-2009, 12:17 PM
Platte City was .0002

Jorge
05-14-2009, 12:20 PM
Platte City was .0002

Andy likes to make it close, and interesting:razz:

nthole
05-14-2009, 12:31 PM
Like Greg said, between the printing sheets, alignment of scores, entry system, etc, keeping the scoring to 1 number is probably pretty important. And the reason they don't use a 0 is dq or not (which most are accidental), no one spent heaps of money and time to show up and get a 0. It's just kind of insulting. I see no issues with 1 through 9. It's pretty clear when it's explained.

71-South
05-14-2009, 03:44 PM
I think it should be a series of happy faces... For example...

:icon_shy :icon_sick :cry: :icon_frow :icon_smil :| :-? :icon_wink :icon_cool

That would solve it real good.

watertowerbbq
05-14-2009, 09:54 PM
This is just my thoughts on how the factors came to be. I might be close or I might be way out in left field.

Since a perfect score of all 9's yields a score of 180, I think this part was intentional, although I can't think of any reason it was chosen. So if you take 180 and divide by 5 (number of scores counted out of 6), you get 36.

Now make the assumption that the original idea was that taste was to be twice as important as tenderness, which in turn was to be twice as important as appearance. In effect, you get the 4, 2, 1 numbers others have described. Solve (4x + 2x + 1x) * 9 = 36 where x is the weighted value of appearance and you get x = 0.5714.

So the appearance weighted value of 1x = 0.5714

If you go back to your assumption that the tenderness is worth twice as much as appearance, you get 2x = 2 * 0.5714 = 1.1428, or the weighted value of tenderness

Now although the value of 1x is determined to be 0.5714, it actually has some additional digits beyond that. Therefore, to solve back to the original equation we have to rearrange it a little bit. Solve (y + 1.1428 + 0.5714) * 9 = 36 where y would be the weighted average of taste and you get y = 2.2858

So the taste weighted value of y = 2.2858

I think that makes sense. Appearance is 0.5714, tenderness is 1.1428 and taste if 2.2858. Isn't algebra wonderful! :-D:-D

I think the original premise that the coefficients are random is not accurate, but I could be wrong.

To remove the 0.0002 difference in scores the whole fractions (36/63), (72/63) and (144/63) would have to be used.

Hope this makes sense.

71-South
05-14-2009, 10:10 PM
Matt, that's brilliant. I hadn't thought about it that way.

It's not as brilliant as the happy faces, but it's brilliant nonetheless.

Plowboy
05-14-2009, 10:38 PM
It was .0008, but who's counting.....:eek:

In Platte City it was .0002... and I'M COUNTING!!!! :cry:

Plowboy
05-14-2009, 10:40 PM
Andy likes to make it close, and interesting:razz:

It was more close than interesting.

Springfield was decided by .0002 as well. WAIT, Andy was on that team, too!

Jeff_in_KC
05-14-2009, 10:47 PM
<sarcasm>At least with comfrank's idea, there would be fewer numbers and thus KCBS would use less ink and save money on score sheets.</sarcasm>

Plowboy
05-14-2009, 10:48 PM
<sarcasm>At least with comfrank's idea, there would be fewer numbers and thus KCBS would use less ink and save money on score sheets.</sarcasm>

How very Green! I knew you were a tree hugger deep inside, Jeff.

Slamdunkpro
05-14-2009, 10:51 PM
This is just my thoughts on how the factors came to be. I might be close or I might be way out in left field.

Since a perfect score of all 9's yields a score of 180, I think this part was intentional, although I can't think of any reason it was chosen. So if you take 180 and divide by 5 (number of scores counted out of 6), you get 36.

Now make the assumption that the original idea was that taste was to be twice as important as tenderness, which in turn was to be twice as important as appearance. In effect, you get the 4, 2, 1 numbers others have described. Solve (4x + 2x + 1x) * 9 = 36 where x is the weighted value of appearance and you get x = 0.5714.

So the appearance weighted value of 1x = 0.5714

If you go back to your assumption that the tenderness is worth twice as much as appearance, you get 2x = 2 * 0.5714 = 1.1428, or the weighted value of tenderness

Now although the value of 1x is determined to be 0.5714, it actually has some additional digits beyond that. Therefore, to solve back to the original equation we have to rearrange it a little bit. Solve (y + 1.1428 + 0.5714) * 9 = 36 where y would be the weighted average of taste and you get y = 2.2858

So the taste weighted value of y = 2.2858

I think that makes sense. Appearance is 0.5714, tenderness is 1.1428 and taste if 2.2858. Isn't algebra wonderful! :-D:-D

I think the original premise that the coefficients are random is not accurate, but I could be wrong.

To remove the 0.0002 difference in scores the whole fractions (36/63), (72/63) and (144/63) would have to be used.

Hope this makes sense.

My head hurts:tongue:

HoDeDo
05-15-2009, 05:11 PM
It was .0008, but who's counting.....:eek:

Yea, it was Todd and I that lost in Platte city by .0002 last weekend :rolleyes:

And in Springfield there was .0002 between Quau and I for Reserve. Just goes to show how close it can get.... and it can't get any closer than that! :tongue:

Vince RnQ
05-16-2009, 06:54 PM
This is just my thoughts on how the factors came to be. I might be close or I might be way out in left field.

Since a perfect score of all 9's yields a score of 180, I think this part was intentional, although I can't think of any reason it was chosen. So if you take 180 and divide by 5 (number of scores counted out of 6), you get 36.

Now make the assumption that the original idea was that taste was to be twice as important as tenderness, which in turn was to be twice as important as appearance. In effect, you get the 4, 2, 1 numbers others have described. Solve (4x + 2x + 1x) * 9 = 36 where x is the weighted value of appearance and you get x = 0.5714.

So the appearance weighted value of 1x = 0.5714

If you go back to your assumption that the tenderness is worth twice as much as appearance, you get 2x = 2 * 0.5714 = 1.1428, or the weighted value of tenderness

Now although the value of 1x is determined to be 0.5714, it actually has some additional digits beyond that. Therefore, to solve back to the original equation we have to rearrange it a little bit. Solve (y + 1.1428 + 0.5714) * 9 = 36 where y would be the weighted average of taste and you get y = 2.2858

So the taste weighted value of y = 2.2858

I think that makes sense. Appearance is 0.5714, tenderness is 1.1428 and taste if 2.2858. Isn't algebra wonderful! :-D:-D

I think the original premise that the coefficients are random is not accurate, but I could be wrong.

To remove the 0.0002 difference in scores the whole fractions (36/63), (72/63) and (144/63) would have to be used.

Hope this makes sense.

"Thank you, Mr. Know-It-All!"

http://www.hhcc.com/wp-content/uploads/2006/12/rabbit1.gif