DM: Probabilistic prediction of football

[David L Dowe]

   Dear ``data miners'',

Monash University (Australia) -- home of Chris Wallace, Minimum 
Message Length
(MML) and the 1968 Wallace et al. MML mixture modelling paper on 
"Snob" --  has
been running a probabilistic football-tipping competition since 1995 
in which
tipsters tip a probability p for each match of who they think will 
win.

For each game, they score 1 + log_2 (p)   if they're right, and
                          1 + log_2 (1-p) if they're wrong.

The optimal long-term strategy in this competition is (if one knew 
the "true"
probability, q) to tip the true probability, q.

The competition is open to all comers, and it costs nothing to enter.
There are prizes totalling Aus$512 for the ten best Australian 
secondary
students.  Penalties for having started late amount to at most very 
little
or even less.  Prizes for post-secondary people are $0 and much kudos.

For those who like mining data, this is very much both a 
public-domain and
a high-profile data-set.

The competition gets weekly coverage in `The Australian' newspaper 
through
its mention of Dr. David Dowe's "expert" :-) computer tips.

For information on how to join and participate, see either
http://www.cs.monash.edu.au/~footy/  or  e-mail 
footy@cs.monash.edu.au .

Thank you.


Dr. David Dowe, Senior Lecturer, School of Computer Science and 
Software Eng.,
Monash University, Clayton, Victoria 3168, Australia   
footy@cs.monash.edu.au
Fax:+61 3 9905-5146         http://www.csse.monash.edu.au/~footy/