The International Timetabling Competition is now well underway with a large
number of participants from many countries involved.

I am writing today in order to clarify some issues that have been raised by
participants, to make sure that we are all working to the same rules. The
web-site has already been updated to make these thing clearer.

a) Soft Constraints - "more than two events in a row". This contraint means
that students should not have more than two consecutive events. The meaning
is obvious if you think about real students. If they have more than two
consecutive classes then they stop concentrating. Having, for example, two
events together at the end of one day, followed by another at the beginning
of the next day, is OK. You should always use the constraint checking
program provided to make sure that you have understood the constraints and
that we agree with the results you are getting. The checker program also
makes sure that your results are in the right format.

b) Time to run the program. The benchmark program tells you how long you can
run your algorithm for on EACH instance.

c) Repeatability and random seeds. You must be able to repeat the runs that
produced your solutions. This may involve recording the random seed used.
You ARE allowed to experiment with different random seeds and submit the
best run you find for each instance. You ARE allowed to submit different
seeds for different instances.

d) Tuning the algorithm. The program you use should be the same one, with
the same run-time arguments, for each instance. You ARE allowed to have an
algorithm which analyses the problem instance (within the specified run
time) and makes parameter decisions based on the problem instance
charateristics (for example, a genetic algorithm might use a different
population size depending on how hard the instance looks). What you should
NOT do is to allow your algorithm to "recognise" an instance and then use
information from previous runs to help it. Remember that if you are a likely
winner then your algorithm will be run on new unseen instances. If it
doesn't perform well then we will look closer to see if you have stayed
within the SPIRIT of the competition. Generally I think you will be able to
predict what we would consider to be within the spirit and what not. But if
you are unsure about anything please do not hesitate to ask.

e) Checker program - Thanks to Andrea Schaerf for pointing out that the
check program contains a couple of instances of the expression "x=x++" which
has undefined meaning in standard c++. This should of course be simply
"x++". If you are having problems with the checker program then please get
the latest version from the web site in which this is fixed.

f) Programming language - you can use any that you like.

g) Algorithm type - this can be any type - not necessarily a metaheuristic.

If you have any further questions please do not hesitate to ask.

Good luck to you all!

Best Wishes

Ben Paechter