Someone needs to if Sam is gonna keep the money.

"Paul Rubin" wrote in message
...
"StanB" writes:
We'll need your SS# for the 1099.

Do the sponsors also have to turn in 1099's?

I've sent out my $680 pledge and did not fill out a 1099. Didn't know
it was a requirement to giving away money. ;-)

StanB wrote:
Someone needs to if Sam is gonna keep the money.

"Paul Rubin" wrote in message
...
"StanB" writes:
We'll need your SS# for the 1099.

Do the sponsors also have to turn in 1099's?

Paul Rubin wrote:
Paul Rubin writes:
I'm still working on crunching the numbers according to the USCF
formula.

According to my latest calculation using the USCF formula, Brock is
favored by about 56-44. But having gotten it wrong so many times
before, I'm not enormously confident of this calculation either.
If anyone cares, I'll post the program I did it with.

How are you calculating this? The resulting odds I have calculated are
different than this.

"Tyrone Slothrop" writes:
According to my latest calculation using the USCF formula, Brock is
favored by about 56-44. But having gotten it wrong so many times
before, I'm not enormously confident of this calculation either.
If anyone cares, I'll post the program I did it with.

How are you calculating this? The resulting odds I have calculated are
different than this.

It was just a random simulation using this Python script:

http://www.nightsong.com/phr/chess/sloan-brock.py

However, there could easily be an error. I haven't looked over it
since right after I first wrote it.

Using what appear to be the same assumptions
in your program, (Pdraw=0.25, White worth
50 points, rating delta 111), I also calculate
different odds. (31% Sloan)

Neglecting color (PW=0.22, PL = 0.53, PD = 0.25),
PMatch (Sloan) = 30.4%


"Paul Rubin" wrote in message
...
"Tyrone Slothrop" writes:
According to my latest calculation using the USCF formula, Brock
is
favored by about 56-44. But having gotten it wrong so many
times
before, I'm not enormously confident of this calculation either.
If anyone cares, I'll post the program I did it with.

How are you calculating this? The resulting odds I have
calculated are
different than this.

It was just a random simulation using this Python script:

http://www.nightsong.com/phr/chess/sloan-brock.py

However, there could easily be an error. I haven't looked over it
since right after I first wrote it.

"David Kane" writes:
Neglecting color (PW=0.22, PL = 0.53, PD = 0.25),
PMatch (Sloan) = 30.4%

Thanks, I see where my program was wrong. With white worth 50 points
I now get about 32% Sloan. Sorry, Sam.

Updated program is at the same url:

http://www.nightsong.com/phr/chess/sloan-brock.py

I'm not a probability/statistics expert (far from it!), so what I'm
going to suggested is only a half-educated (at best -grin-) guess.

Assume Sloan's performance follows the Normal Distribution with a mean
of 1931 and standard deviation of 400. To break even against Bill
Brock he will need a performance of 2042 or better. The probability of
his doing so is roughly 0.390698 (using:
http://davidmlane.com/hyperstat/z_table.html)

So, based on this analysis (which may be totally flawed) I give Sloan
about a 39% chance.

Best regards,

George John

"George John" writes:
Assume Sloan's performance follows the Normal Distribution with a mean
of 1931 and standard deviation of 400.

I don't think ratings have ever been done that way. SD as used by
Elo was a lot smaller, like 55 points. These days the USCF uses the
logistic rather than normal distribution, with that 1/(1+10**(dR/400))
formula. I'm not sure what SD that corresponds to.

So, based on this analysis (which may be totally flawed) I give Sloan
about a 39% chance.

That might be close for a long match. For a 4-game match you have to
take other effects into account, like the draw probability and the
white-piece advantage (who gets white in round 1 doesn't seem to
matter much). I had nagging doubts about my earlier calculation since
among other things it was fairly insensitive to the draw ratio. It
turned out to be wrong due to excessive cleverness in mapping random
numbers to game results. But I think my current program (whose
results seem to match David Kane's) is doing the right thing or
something pretty close.

Paul,

I'm sometimes amazed by what I forget. -smile- The standard
deviation of individual performance is quite different from that of the
rating distribution, and will vary from player to player. I don't know
what the average is, but one number I saw thrown out is 200.

If we use my approach with a SD of 200, the probability of a 1931 rated
player having a performance at or above 2042 is roughly 0.289447 or
about 29%, which isn't too far from your 32%. FYI, if I use a SD of
240 I get something close to 32%.

BTW, I do not know if chess players' performance distributions follow a
Normal Distribution well or not, but they might.

Best regards,

George John

"George John" writes:
BTW, I do not know if chess players' performance distributions follow a
Normal Distribution well or not, but they might.

The normal and logistic distributions are pretty close when the rating
delta is small, I believe. The logistic distribution seems to be a
better predictor for actual chess players when the difference is
large. A 2100 player will occasionally score an upset over a GM and
this happens more often than the normal distribution would predict.

The normal distribution usually describes something that's the sum of
a bunch of smaller independent random variables (the central limit
theorem says that such a sum will always be approximately normally
distributed). But if a GM flubs a game against a 2100 player, it's
more likely the result of one big variable going the wrong way, than
the sum of a bunch of little ones ;-).