[Termtools] scores over 100 %

[Zantema, H.]

The termination competition seems to be over now after completing the
second round. Thanks to all participants, and congratulations to the
winners AProVE and JAMBOX (all winning tools have been written in
JAVA....). Thanks to Claude for running everything. 

About the scores I think it is strange that the second round is taken
into account here. Examples of two really undesired consequences:

* In the subcategory TRS modulo theory AProVE succeeded in proving
termination of 55 out of 53 terminating systems, resulting in a score of
103.8 %

* TPA could have increased its score of 422 to 424 only by building in a
10 seconds sleep: then for Zantema/z09 and ../z10 there would have been
a second round in which again TPA would have been able to solve these.

I think it is undesirable that solving a system both in the first and in
the second round counts for two points, for three reasons:

(1) according to the rules there is no second round if a system was
already solved by one tool in the first round;

(2) the total number of points does not reflect any more the total
number of systems that has been solved (resulting in scores over 100 %);

(3) it is not fair to be punished for fast generation of proofs.

I think it is fair to get one point for finding a proof, no matter
whether it is in the first or second round or both.

Claude: is it possible to repair this?

According to the rules all second rounds for which in the first round
already a proof was found should be removed. In case you think these
illegal second rounds contain interesting information, I can agree with
keeping them, but without affecting the results in the present undesired
way.


            Best regards, Hans Zantema.

_________________________________________________________
Dr Hans Zantema            
Technische Universiteit Eindhoven, Department of Computer Science 
P.O. Box 513, 5600 MB Eindhoven, The Netherlands 
e-mail: H.Zantema at tue.nl, homepage: www.win.tue.nl/~hzantema
office: Hoofdgebouw room 6.73, tel: (040)2472749
 

_______________________________________________
Termtools mailing list
Termtools at lists.lri.fr
http://lists.lri.fr/mailman/listinfo/termtools