Problem with the rules with particular grids

>The winner will be the script that after 30 complete runs has the highest score computed by finding the sum of all 30 runs.

Now, suppose that Matt puts the grid that gives the highest possible score (56903 points).

If my script is unable to find this solution, I guess it has no chance to finish in the first, even though it performs better scores on all other grids !
(for the moment, my program finds 22514 points in 30 seconds)

Instead of summing the points of all grids, I suggest the following:
for every grid, keep a list of the highest scores.
Your score for this grid will be:
you number of points/highest score of the grid

This way, even if your program doesn't find a given grid, it can still be in the finals.

JC

First, i dont believe he will do that, cause this will limit the competition only to the people that will find the max to this grid.

Second, it' far too late for the rules to be changed. Many people have already submit their scripts.

Third, instead of trying to fit the contest to your script, try to fit your script to the contest

My suggestion didn't change anything to the contest, except the way to compute the final ranking.
In this case, an entry achieving the best results on all grids will score 100*1.00=100 points, and program finding all optimal boards except one will score 99*1.00+(score/best) points. IMHO, this is much more fair !

The problem with summing scores of all grids is that it's like adding apples and potatoes.
One grid can give 10,000 points and another grid 100 points.
A winning entry could win with only a few good results !

Since I don't want to lose 34,389 points, I'll simply hardcode the highest scoring grid, and check at the beginning of my program, so that it will return instantly the higher score (this test will take only a few milliseconds).

I didn't want to use this kind of treachery, but the current rules force me to do that :-(((

About my entry, it's working now pretty good, no need to change the rules for my program :-)

JC

Dont expect to get higher scores than others in the other grids if you only find 22000 to this grid anyway.
Yes your method of scor calculating is more fair but Matt will be out of town for 3 days. Wait till he is back but i doubt he will change anything.

I'm not in the mind of Matt but if I want to make a suite of word files I choose a balnced set of files:
- real word (i.e. english) words (they will not have 3 consecutive letter and often the "Q"(ten points) is followed by a "U" ... etc)
- only 3 letters,
- only 4 letters,
- only 5 letters,
- only 6 letter,
- only 3 and 6 letters,
- set that contain the words of a perfect grid.
- balanced words (as suggested from someone, sorry I don't remember who), the letters are balanced with the possibilty of inserts in the score.
- totally random word file
- maybe I'm forgetting something
every one varying from 500 to 5000 non duplicated words.
Every script will perform better with some and worst with other.
If you have only one higher score with perfect, probably you cannot beat a script that perform better in all other 29.

fra: I agree with you about the wordlists, Xs0's balanced wordlists are excellent for testing the programs.

Xs0 posted previously a 2Mb file containing 200 wordlists (random and balanced).
http://dist.doticni.net/scrabble/inputs.zip

Can we compare our scores on his 5 first balanced inputs ?
With my last version, I get:
balanced/input00.txt: 5049
balanced/input01.txt: 4902
balanced/input02.txt: 4799
balanced/input03.txt: 4537
balanced/input04.txt: 5012

NOTE: The above scores are meaningless, since my computer is faster than the judging machine !

It's just to know how much efforts I have to spend on improving my program (we have still one week left, and I'm increasing my PHP skills consequently !).

JC

hi mcoder, theese are my results:
balanced/input00.txt: 5369
balanced/input01.txt: 6015
balanced/input02.txt: 5863
balanced/input03.txt: 5122
balanced/input04.txt: 5950

keep in mind that i've get then on athlon 2.0 with 1GB ram and a GCC 3.1 compiled php with all the optimizations on and not too many modules.
Also I've learned that php has very comfortable (and expensive) arrays and *very* fast strings.

This is gonna be a stiff competition. My scores are somewhere between jc and fra(closer to jc)!

After one night of sleep, I improved slightly my results:
balanced/input00.txt: 5012
balanced/input01.txt: 5598
balanced/input02.txt: 5215
balanced/input03.txt: 4946
balanced/input04.txt: 6516

(Good surprise for input04 !)
But I still feel miserable, compared to fra's results.

My configuration is P4 2.4Ghz, 1Gb RAM, W2K, EasyPhp 1.6.
Well, I guess I have to install a better PHP configuration.
What do you recommend for Windows ?

JC

I only have one comment on this thread, and that comment is to quote fractalbit: