[DNO] Unoffical Results

OK, here we go. The unofficial results are ready. Hit the ftp at:

ftp://ftp.codewalkers.com/

all scripts are there. all results are there. all inputs are there.

At the moment Script 8 (crisp) is the leader. Now is your time to scutinize his script .

I'm second !!!!

Crisp is first, Fractalbit is third and Skrol29 fourth !

JC

BTW, who won the award of the best interface ?

It's a pity there wasn't a 50x50x3 in the inputs; I would have liked to have a score of over a million on one of the boards ;)
input13.txt clearly shows that my algorithm doesn't always produce better results than just playing color by color, but I have the highest score on 22 boards out of the 50 which are mainly the bigger ones

I'll closely examine the other scripts; always nice to see how other people have gone about it.
I already seen some scripts, and from those I think JC has the nicest interface (looks especially nice on smaller boards; large boards are sluggish and don't show the individual tiles so well).

I tried to run all the scripts now. What bothers me is the fact that 8 out of the 15 are throwing numerous notices when running with error_reporting set to E_ALL (which is default on my server).
Script number 1 is producing results, but in the wrong format (times out on large boards, same as number 3).
For script number 5 somehow the flash display output doesn't work on my machine.
Script number 12 (as discussed in the other thread) doesn't work on machines running windows.
Script number 15 (gatopeich) also didn't give me any output; don't know why yet....

Darn you JC, not again! I am somewhat happy and somewhat dissapointed now :-| Congrats to crisp.

I'm working on that one

as you can see in the results, script 1 got 0's al the way down and script 3 got lots of 0's.

Fractalbit: it's weird, since Crisp has 22 records, you have 14 and me only 3.
The difference between Crisp and me is 5%, and between me and you 3%.

In general, Crisp's script reached 94.24% of every board, you 93.46% and me 89.10%. I guess the difference has been done on the 3-color boards.

JC

I know; I was just explaining why

>I guess the difference has been done on the 3-color boards.

As expected indeed the big boards have determined the ranking. If I look at the boards that could get you scores >200,000 (input12, input20, input22 and input25) you will see that these make the main difference between our scores. If I leave those boards out of the totals fractalbit's totalscore is higher than JC's

Just for the record: I have a fallback built in that switches back to the fast algorithm when 24 MattSecs have passed. According the results I can see that this happened on the following boards: input28, input34, input44, input48 and input49. You may get different results on these boards because on other machines the switch may occur later or sooner.
I should have spent more time finetuning, because changing the $nodes <= 60 into $nodes <= 50 on line 347 would have prevented the fallback giving me better results on these boards ;)

Meanwhile I also figured out why gatopeich's script isn't working on my pc; simply because ob_start with a user-defined function doesn't flush the buffer after timeout and the function isn't called when running PHP under IIS (same problem as in Marcel's script).

Thanks for the analysis (crisp also) as i didnt have the time to do something similar. Between the 2 of us i have the record on 34 files and you on 15 files (1 is a tie!). But the point is, that
on the files that crisp mentioned, i am far behind from you (ex. input22. me 223000, you 268000 - 45000 difference!). That explains why you have a better overall scor than me. It seems i should optimize better for 3 color tiles. Anyway with the analysis i feel somewhat better. Besides, i am on the first three this time as i wanted

I'm happy to be second, but I would have been even more happy to be first !
Perhaps for the next time.
BTW, it doesn't seem that xs0 (Mitja Slenc) and Andy Uhl worked on this contest.
Not enough time ?

Crisp: I think you can improve a little bit your solver as follows: when you reduce the 50x50x3 boards to 2 colors, you always do it from bottom left to top right. This is not the best way. Instead, remove the grouped tiles (that is: not unique) from the top left to bottom right, and the unique blocks from the bottom right to top left.
Also, if you can find the biggest chain of tiles, it's better starting by removing them first. I did some testings in C and discovered these 2 simple principles.

JC

crisp: This came up last contest too. I have a bug report in but they aren't ever going to touch it I don't think. Maybe if we vote for the bug someone will get around to fixing it. Here is the link:

http://bugs.php.net/bug.php?id=22401