That time of year: pipe delimiters are almost back

[pelagius]

Potential tweaks to the formula. As a public service announcement here is my plan for the formula this year (kind of bored with the old formula)

Old Formula:

Points = 20*winner + 
         40*winner*(1-MOVD/24) if abs(MOVD) <= 24, 0 otherwise +
         40*(1-d(predicted,actual)/28) if d( , ) <= 28, 0 otherwise

Winner = 1 if the winning team is correctly picked, 0 otherwise
MOVD   = margin of victory diferrential: Predicted (MOV) - Actual(MOV)
d( , ) = Euclidean distance

Here is the formula I think I will move to this season (but I may tweak it):

Points = 30*winner + 35*f(MOVD) + 35*g(d(predicted,actual))

Winner = 1 if the winning team is correctly picked, 0 otherwise
MOVD   = margin of victory diferrential: Predicted (MOV) - Actual(MOV)
d( , ) = Euclidean distance


f(x)    = 1 - 0.60(x/15)  if       x <= 5,
        = 1 - 0.80(x/15)  if  5  < x <= 10
        = 1 - 1.00(x/15)  if  10 < x <= 15
        = 0               if  15 < x

g(x)    = 1 - 0.60(x/20)  if       x <= 5,
        = 1 - 0.80(x/20)  if  5  < x <= 10
        = 1 - 1.00(x/20)  if  10 < x <= 20
        = 0               if  20 < x

Feel free to comment about the formula. I may take your advice (although probably not since I am definitely a curmudgeon when it comes to the prediction contest).

[PaloAltoCougar]

I have no idea what those formulae mean, but I have complete confidence in (a) your system, and (b) the high probablility that someone, anyone, will unseat Parrothead from his annual perch atop the standings.

Thanks for doing this.

BTW, my confidence in (a), above, was bolstered by the Olympic medals model that DDD mocked midway through the games, but proved to be pretty much dead on by the time Eric Idle was shot from a cannon.

[Coach McGuirk]

Me|Looks Great.

[Indy Coug]

I have no idea what those formulae mean, but I have complete confidence in (a) your system, and (b) the high probablility that someone, anyone, will unseat Parrothead from his annual perch atop the standings.

Thanks for doing this.

BTW, my confidence in (a), above, was bolstered by the Olympic medals model that DDD mocked midway through the games, but proved to be pretty much dead on by the time Eric Idle was shot from a cannon.

I think in a nutshell, it puts less emphasis on picking the correct winner and more emphasis on correctly predicting the score.

[pelagius]

I have no idea what those formulae mean, but I have complete confidence in (a) your system, and (b) the high probablility that someone, anyone, will unseat Parrothead from his annual perch atop the standings.

Thanks for doing this.

BTW, my confidence in (a), above, was bolstered by the Olympic medals model that DDD mocked midway through the games, but proved to be pretty much dead on by the time Eric Idle was shot from a cannon.

The formula is now concave with respect to MOV and distance. Before it decreased linearlly. So, for example, if MOV = 3 you would get 88% of the MOV points under the old and new formula. If MOV = 8 you would get 67% of the MOV points under the old and 57% of the points under the new. IF MOV = 13 you would get 46% under the old and 13% under the new.

Also, you can get point from MOV even if you pick the wron winner now. If BYU wins by 5 and you picked the oponnent to win by 4 the MOVD is 9 and you would get some points for that. Before you got zero,

[Jeff Lebowski]

Thanks, pelagius! As a token of my respect and esteem for you, I will forgo using smilies in this post.

And if anyone asks about the Euclidean distance part of the formula, don't tell them.

[pelagius]

The formula is now concave with respect to MOV and distance. Before it decreased linearlly. So, for example, if MOV = 3 you would get 88% of the MOV points under the old and new formula. If MOV = 8 you would get 67% of the MOV points under the old and 57% of the points under the new. IF MOV = 13 you would get 46% under the old and 13% under the new.

Also, you can get point from MOV even if you pick the wron winner now. If BYU wins by 5 and you picked the oponnent to win by 4 the MOVD is 9 and you would get some points for that. Before you got zero,

Also you used to get points for being way off on both MOV and distance. Points for those categories reach zero more quickly.

[pelagius]

Thanks, pelagius! As a token of my respect and esteem for you, I will forgo using smilies in this post.

And if anyone asks about the Euclidean distance part of the formula, don't tell them.

I am always happy when emoticon use is attenuated for any reason.

[Joe Public]

I think in a nutshell, it puts less emphasis on picking the correct winner and more emphasis on correctly predicting the score.

Are you kidding? He went from 20*winner to 30*winner, man!

I do have a question, though, about how "predicted" and "actual" are defined in the Euclidean distance part of the . . .

And if anyone asks about the Euclidean distance part of the formula, don't tell them.

. . . aw, rats!

[pelagius]

Are you kidding? He went from 20*winner to 30*winner, man!

No, Indy is exactly right. The explicit weight on winner is higher but effectively the old formula gave more weight to picking the winner. You only got points for MOV if you picked the correct winner. That has changed. Second, the shape of the points function was decreasing but much flatter. Thus being moderately off in terms of MOV still resulted in a relatively high amount of points as long as you picked winner correctly.

[Joe Public]

No, Indy is exactly right. The explicit weight on winner is higher but effectively the old formula gave more weight to picking the winner. You only got points for MOV if you picked the correct winner. That has changed. Second, the shape of the points function was decreasing but much flatter. Thus being moderately off in terms of MOV still resulted in a relatively high amount of points as long as you picked winner correctly.

Aw, rats!

[pelagius]

. . . aw, rats!

From last year's FAQ:

8. I keep looking at that refernce to "Euclidean distance" and assume that somewhere, someone is reading this and laughing so hard that a couple of pens fall out of their pocket protector. But, in fact, I have no idea if it is a joke or not. Is it a Joke?

No:

[old_gregg]

From last year's FAQ:

8. I keep looking at that refernce to "Euclidean distance" and assume that somewhere, someone is reading this and laughing so hard that a couple of pens fall out of their pocket protector. But, in fact, I have no idea if it is a joke or not. Is it a Joke?

No:

[YOhio]

Glad you fixed the formula. Parrothead was getting wise to your ways.

[Jeff Lebowski]

From last year's FAQ:

8. I keep looking at that refernce to "Euclidean distance" and assume that somewhere, someone is reading this and laughing so hard that a couple of pens fall out of their pocket protector. But, in fact, I have no idea if it is a joke or not. Is it a Joke?

No:

Dang it.