## Game Predictions

Expand Messages
• There are a lot of methods for predicting games out there. Probably the best is just Vegas line. I posted a method at
Message 1 of 4 , Feb 22, 2001
• 0 Attachment
There are a lot of methods for predicting games out there. Probably
the best is just Vegas' line. I posted a method at

http://www.tsoft.com/~deano/articles/kalman.html

This is similar to what Ken Massey uses at

http://www.mratings.com/rate/nba-p.htm

and Massey uses the same system for college football, where he is a
contributor to the infamous BCS.

Someone did some work for me recently to study what are optimal ways
to use the methods and to look at how good the methods could be. He
generally found that they can only predict games at about a 65% clip,
which is about what Massey is getting this year.

Has anyone seen other systems that legitimately do better?

How does Bob Chaikin's simulator do?

One of the good things about these systems is that they do come up
with a strength of schedule evaluation. One of the teams I am
working with is very interested in using such a thing for evaluating
college players. I know the RPI has an approximate strength of
schedule method. Are there others?

Dean Oliver
• ... ways ... He ... clip, ... This is an area of interest for me, have looked at this for several sports, and some academic conclusions seem to be that a
Message 2 of 4 , Jul 4, 2002
• 0 Attachment
--- In APBR_analysis@y..., "Dean Oliver" <deano@t...> wrote:
>
> Someone did some work for me recently to study what are optimal
ways
> to use the methods and to look at how good the methods could be.
He
> generally found that they can only predict games at about a 65%
clip,
> which is about what Massey is getting this year.
>
> Has anyone seen other systems that legitimately do better?
>

This is an area of interest for me, have looked at this for several
sports, and some academic conclusions seem to be that a variety of
methods will get around the same results. The amount
of 'information' in the scores used to produce the next prediction
has an 'upper limit'. Seems to be pretty much right. So 60-70% is
• ... I may not answer all of your posts today. I m swamped. Let me point you to Ken Massey s website where he makes predictions. He is at
Message 3 of 4 , Jul 8, 2002
• 0 Attachment
--- In APBR_analysis@y..., "aussievamp2" <rnmscott@b...> wrote:
> --- In APBR_analysis@y..., "Dean Oliver" <deano@t...> wrote:
> >
> > Someone did some work for me recently to study what are optimal
> ways
> > to use the methods and to look at how good the methods could be.
> He
> > generally found that they can only predict games at about a 65%
> clip,
> > which is about what Massey is getting this year.
> >
> > Has anyone seen other systems that legitimately do better?
> >
>
> This is an area of interest for me, have looked at this for several
> sports, and some academic conclusions seem to be that a variety of
> methods will get around the same results. The amount
> of 'information' in the scores used to produce the next prediction
> has an 'upper limit'. Seems to be pretty much right. So 60-70% is

I may not answer all of your posts today. I'm swamped. Let me point
you to Ken Massey's website where he makes predictions. He is at

http://www.mratings.com/

or for NBA predictions

http://www.mratings.com/rate/nba-p.htm

You can see he was 64-65% right. He did a pretty good job getting
the percentiles right, too.

Do I think 70% is possible with basketball? Yes, Massey does over
72% for college men's hoops and about 76% for college women's hoops.
It really depends upon the level of competition and the nature of the
game. Massey posts predictions for all sorts of sports and all sorts
of levels, making it kind of neat to compare.

DeanO
Message 4 of 4 , Jul 8, 2002
• 0 Attachment
sort of thing, and having seen tons of them, that that is sort of the basic
limit. To move a little past that you need to start taking other things
into account - injuries, scheduling, motivation, statistics, etc., etc.

You can get higher percentages when you have something with huge mismatches
like say, college basketball - e.g. if there are a lot of as good as 100%
victories in there, then the 'correct prediction' percentage will go up,
compared to any professional sport in general where the talent levels are
much more even.

-----Original Message-----
From: HoopStudies [mailto:deano@...]
Sent: July 08, 2002 6:51 PM
To: APBR_analysis@yahoogroups.com
Subject: [APBR_analysis] Re: Game Predictions

--- In APBR_analysis@y..., "aussievamp2" <rnmscott@b...> wrote:
> --- In APBR_analysis@y..., "Dean Oliver" <deano@t...> wrote:
> >
> > Someone did some work for me recently to study what are optimal
> ways
> > to use the methods and to look at how good the methods could be.
> He
> > generally found that they can only predict games at about a 65%
> clip,
> > which is about what Massey is getting this year.
> >
> > Has anyone seen other systems that legitimately do better?
> >
>
> This is an area of interest for me, have looked at this for several
> sports, and some academic conclusions seem to be that a variety of
> methods will get around the same results. The amount
> of 'information' in the scores used to produce the next prediction
> has an 'upper limit'. Seems to be pretty much right. So 60-70% is

I may not answer all of your posts today. I'm swamped. Let me point
you to Ken Massey's website where he makes predictions. He is at

http://www.mratings.com/

or for NBA predictions

http://www.mratings.com/rate/nba-p.htm

You can see he was 64-65% right. He did a pretty good job getting
the percentiles right, too.

Do I think 70% is possible with basketball? Yes, Massey does over
72% for college men's hoops and about 76% for college women's hoops.
It really depends upon the level of competition and the nature of the
game. Massey posts predictions for all sorts of sports and all sorts
of levels, making it kind of neat to compare.

DeanO

To unsubscribe from this group, send an email to:
APBR_analysis-unsubscribe@yahoogroups.com

Your use of Yahoo! Groups is subject to http://docs.yahoo.com/info/terms/
Your message has been successfully submitted and would be delivered to recipients shortly.