PathFinder Scores

[Electroman]

Here you go, Congratulations Alkatran :D, StevenHickerson was very close though :).

Name L1 L2 L3 L4 L5 L6 L7 L8 L9 L10 | Score
---- | -----
Merri - 3040 (4) 2739 (7) 1320 (3) 5408 (6) 7637 (4) 7448 (3) 1179 (2) 3010 (4) 4628 (4) 5891 (3) | 40
BodwadUK - 1836 (8) 75915 (1) 508 (7) 55626 (2) 5147 (5) 1328 (7) 295 (7) 69766 (1) 2274 (6) 2037 (7) | 51
NotLKH - 4608 (1) 9221 (5) 510 (6) 1650 (8) 10605 (2) 3424 (6) 667 (4) 2888 (5) 1428 (7) 2323 (6) | 50
BuggyProgrammer - 2340 (5) 62839 (2) 2818 (1) 59986 (1) 84391 (1) 42868 (1) 1501 (1) 15264 (2) 5432 (3) 241617 (1) | 18
Alkatran - 1892 (7) 9135 (6) 506 (8) 7238 (5) 9265 (3) 1316 (8) 387 (6) 4778 (3) 1354 (8) 2523 (5) | 59
StevenHickerson - 2256 (6) 2125 (8) 1066 (4) 11280 (3) 4697 (7) 5970 (4) 199 (8) 1714 (8) 2968 (5) 5127 (4) | 57
Arun.NS - 3316 (3) 23007 (3) 1056 (5) 8324 (4) 2525 (8) 9262 (2) 395 (5) 2676 (6) 7060 (1) 923 (8) | 45
bpd - 4368 (2) 13305 (4) 1664 (2) 3110 (7) 5127 (6) 4528 (5) 1137 (3) 2366 (7) 5542 (2) 5911 (2) | 40

Edit: Final markings with Cyborg's entry : http://www.vbforums.com/showpost.php?p=1809423&postcount=8

[Electroman]

BTW here are the maps I used for the testing...

[Merri]

Too bad the codes didn't get a rating for generally giving low results, I'm the only one whose code was ALWAYS BELOW 8000! :D

Looks like my code wasn't able to have any lucky run this time. Oh well, one shot is one shot.


Edit Heh, I ran the level 7 once as it was my worst level and got the same score StevenHickerson had in the level. Life is cruel! :)


Edit #2 Just for interest: total points ranking!

1. NotLKH: 37324
2. StevenHickerson: 37402
3. Alkatran: 38394
4. Merri: 42300
5. bps: 47058
6. Arun.NS: 58544
7. BodwadUK: 214732
8. BuggyProgrammer: 519056

Makes me wondering: shouldn't the one with the least points be a winner? ;) It says this in the contest rules: We will be running your code with a variety of different maps of sizes ranging upto 300x300 and will be recording the scores. Whoever gets the lowest score overall will be the winner. - so isn't the current ranking system against the rules given in the beginning?


Edit #3 Yeah, I just HAD to :D

[Electroman]

Posted by Merri
Too bad the codes didn't get a rating for generally giving low results, I'm the only one whose code was ALWAYS BELOW 8000! :D

Looks like my code wasn't able to have any lucky run this time. Oh well, one shot is one shot.


Edit Heh, I ran the level 7 once as it was my worst level and got the same score StevenHickerson had in the level. Life is cruel! :)


Edit #2 Just for interest: total points ranking!

1. NotLKH: 37324
2. StevenHickerson: 37402
3. Alkatran: 38394
4. Merri: 42300
5. bps: 47058
6. Arun.NS: 58544
7. BodwadUK: 214732
8. BuggyProgrammer: 519056

Makes me wondering: shouldn't the one with the least points be a winner? ;) It says this in the contest rules: We will be running your code with a variety of different maps of sizes ranging upto 300x300 and will be recording the scores. Whoever gets the lowest score overall will be the winner. - so isn't the current ranking system against the rules given in the beginning?


Edit #3 Yeah, I just HAD to :D Yea I remember yours on L7 managed to visit every block I think before it got there, just spot of bad luck though.

Your code had real problems on L5, it made losts of bad choices and had to keep going back to the grid in the middle and took ages.

I guess I should let NotLKH win too. I did say part way though the Contest thread the new Marking system that would be more fair, to avoid things like BuggyProgrammers 200,000+ score ruining the overall result.


Bodwad's code on L2 was funny too, I thought it covered the open area really well then once it had visited it all it then danced around in the space for ages. I was going to stop it when it got to 100,000 but it stopped just short. I was gonna do the same for BuggyProgrammer's on L10 but I then thought I'd leave it to 200,000 and the phone rang, when I came back it had finished like that, his had got stuck in the top part and everytime it got close to coming out went back in.

Try running Arun.NS's on L10, it acts like it it was made for that map, it goes straight to the end no messing and doesn't even do the lookAbout every move.


BTW Alkatran, when your code is compiled it doesn't work, just does a Look and thats it :confused:.

There was also another one that didn't compile because of a enum but ran in the IDE ok.

[cyborg]

What the ****?!

I thought the contest whas due today! I didn't send my code! Is it ok to send it now?

[Electroman]

Posted by cyborg
What the ****?!

I thought the contest whas due today! I didn't send my code! Is it ok to send it now? If you submit your code I'll add it to the scores.

The deadline was today, (Mid-day ;))

[cyborg]

Ok!

Thanks alot! :)

My code sucks, but it's pretty good on some of the maps!

[Electroman]

Heres the scores with Cyborg added on:

Name L1 L2 L3 L4 L5 L6 L7 L8 L9 L10 | Score
---- | -----
Merri - 3040 (5) 2739 (8) 1320 (3) 5408 (6) 7637 (4) 7448 (4) 1179 (2) 3010 (5) 4628 (5) 5891 (3) | 45
BodwadUK - 1836 (9) 75915 (1) 508 (8) 55626 (2) 5147 (5) 1328 (8) 295 (8) 69766 (1) 2274 (7) 2037 (7) | 56
NotLKH - 4608 (2) 9221 (6) 510 (7) 1650 (8) 10605 (2) 3424 (7) 667 (4) 2888 (6) 1428 (8) 2323 (6) | 56
BuggyProgrammer - 2340 (6) 62839 (2) 2818 (1) 59986 (1) 84391 (1) 42868 (1) 1501 (1) 15264 (2) 5432 (4) 241617 (1) | 20
Alkatran - 1892 (8) 9135 (7) 506 (9) 7238 (5) 9265 (3) 1316 (9) 387 (7) 4778 (4) 1354 (9) 2523 (5) | 66
StevenHickerson - 2256 (7) 2125 (9) 1066 (4) 11280 (3) 4697 (7) 5970 (5) 199 (9) 1714 (9) 2968 (6) 5127 (4) | 63
Arun.NS - 3316 (4) 23007 (3) 1056 (5) 8324 (4) 2525 (9) 9262 (3) 395 (5) 2676 (7) 7060 (2) 923 (8) | 50
bpd - 4368 (3) 13305 (5) 1664 (2) 3110 (7) 5127 (6) 4528 (6) 1137 (3) 2366 (8) 5542 (3) 5911 (2) | 45
Cyborg - ---- (0) 16004 (4) 729 (6) 1167 (9) 2806 (8) 15369 (2) 389 (6) 14996 (3) 9114 (1) 916 (9) | 48

[Electroman]

Digesh, I've seen you've posted yours too, I've not got the time at the moment to add its scores too so will do tomorrow night as long as I get time :).

[DIGESH]

Thanks Electroman, though I must confess the algorithm sucks. It was worth a try...

[BuggyProgrammer]

Lol I did horrible. :p

[alkatran]

Thank you! Thank you! The wall crawlin algorythm wins!

BTW Alkatran, when your code is compiled it doesn't work, just does a Look and thats it :confused:
Weird.

[NotLKH]

I got fourth on the first set of results, and tied for third on the second, and apparently fewest points overall!

My code just did the standard! Didn't even optimize for backtracking!
It did do a random seed for initial orientation.

Cool!

:wave:

-Lou

[StevenHickerson]

I'm happy with that lol only missing lowest overall by less than 100 points and doing farely well on most maps :) Gonna have to download the maps and find out what happend on level 4 though.. eek that one is horrid!

Was a very fun contest, thanks!

EDIT:

You know what I noticed thats sad though lol .. I'm between both winners lol.. beat Alk in the overall.. and beat NotLKH in the points per map *sigh* I bet lvl 4 caused all this! arg darn that level :)

EDIT2:

Arg I knew it.. see my algo has a chance on going pretty much 2 different ways each map but once it goes that way its score is consistent for that way. Well the 11k is my bad way on this map .. my other 990 can you believe that.. I would've had first in both sides of the contest had it gone that way when Electro tested it.

[alkatran]

I could have had mine decide to follow the left or right wall randomly at the start... but... why? Mine had to cover entire maps many times.

I think you should have had more than one ending.

[Electroman]

Posted by alkatran
I could have had mine decide to follow the left or right wall randomly at the start... but... why? Mine had to cover entire maps many times.

I think you should have had more than one ending. You might notice on one of the maps I put four ending but I didn't really think too much of that map. I just wanted to create a big map but then didn't have time to add lots of detail so just t5horugh a few lines all over the place. (Just checked and that was lvl 4)


BTW if anyone is interested in the Map Editor I made I will post it.

[Electroman]

Posted by StevenHickerson
I'm happy with that lol only missing lowest overall by less than 100 points and doing farely well on most maps :) Gonna have to download the maps and find out what happend on level 4 though.. eek that one is horrid!

Was a very fun contest, thanks!

EDIT:

You know what I noticed thats sad though lol .. I'm between both winners lol.. beat Alk in the overall.. and beat NotLKH in the points per map *sigh* I bet lvl 4 caused all this! arg darn that level :)

EDIT2:

Arg I knew it.. see my algo has a chance on going pretty much 2 different ways each map but once it goes that way its score is consistent for that way. Well the 11k is my bad way on this map .. my other 990 can you believe that.. I would've had first in both sides of the contest had it gone that way when Electro tested it. BTW Steve on Lvl 5 I thought yours had to be the best at getting though the grid in the middle, for most the algos they left little spaces and had to keep coming back but yours fills the whole thing without retracing. I noticed this at the time but only just realised who's it was cos I couldn't remember when I posted the scores and was gonna mention it.


My Fav. map has to be L9 but I didn't get time to make more blocks :(. All the algos that stuck to the walls went straight though with it though :(.

[alkatran]

Originally posted by Electroman
BTW Steve on Lvl 5 I thought yours had to be the best at getting though the grid in the middle, for most the algos they left little spaces and had to keep coming back but yours fills the whole thing without retracing. I noticed this at the time but only just realised who's it was cos I couldn't remember when I posted the scores and was gonna mention it.

Mine left little spaces. :( But it was a consequence of the wall-tracing goodness. I suppose if I had it switch wall following directions it would have filled it perfectly...

[Electroman]

He's Digesh's Scores: Name L1 L2 L3 L4 L5 L6 L7 L8 L9 L10
----
Digesh -62829 (_) 151713 (_) 7734 (_) 39492 (_) 6555 (_) 7920 (_) 765 (_) 17697 (_) 25686 (_) 4059 (_)

[alkatran]

-62829

Wow! How'd he get a number so low as to be negative?? :rolleyes:

[Electroman]

Posted by alkatran
Wow! How'd he get a number so low as to be negative?? :rolleyes: That was cos his score was a digit bigger than the rest and that made it touch the coloun ;).

[wossname]

When / what is the next contest?

I'm eager to pwn you all again.

[wossname]

Originally posted by Electroman
That was cos his score was a digit bigger than the rest and that made it touch the coloun ;).

Why was he touching your colon with his big digit?

[Electroman]

Posted by wossname
Why was he touching your colon with his big digit? Wossy is Banned from taking part in any forum contests (untill the next one) for that :p.

BTW I jsut found out I have the ability to Ban members so watch out wossy :evil:. (Unfortunetly I can't unban members otherwise I'd keep banning and unbanning NoteMe :lol:)



As for the next contest its gonna be a bigger one and we need some time to set it up (mainly to make sure we have both VB6 & VB.NET ability). Might be about a week but depends, there was talk of doing one in the meantime but we've gone against that and your just gonna have to wait a little but the next one should be a good one. BTW I go on holiday for 10days starting the 21st so I'd definitly hope to get this one started before then otherwise it'll have to wait till after I get back.

[wossname]

Originally posted by Electroman
otherwise I'd keep banning and unbanning NoteMe :lol:

Just the first one would do. Muppet puppet.

[alkatran]

So... what was teh strategy behind everyone else's code?

I designed mine to look only as much as necessary, and always try to turn to the left (or was it right?). I also used a floodfill algorythm to back up to the last space in the move list with an adjacant empty space (I stored where I had been)

Also, I made my own Look and move functions which made sure I didn't move into a wall or look somewhere I'd already checked (just returned from the array isntead).


In the end the behavior was to slide along walls, slowly working it's way to the center. It did pretty well on open areas, covering the whole thing then taking the shortest route back. I found it outperformed most of the other ones by far when it came to pathing... yikes.

The only place it had trouble with was the grid's with single spots. It wouldn't cover them perfectly... :cry:

[Merri]

I got and stored all the information I could by using Look. If the code hadn't checked some position before, it got the info and stored it. The code also marked how many times it had been in a certain spot, thus avoiding going to directions where it had already been at. If there were more than one equal directions to go to (say, all three roads hadn't been visited), then it randomly first chose if it continued to go straight and then if it wouldn't go straigth, it would randomize to turn left or right. I also marked areas where it should never go back again, basically the dead ends in mazes. This was the initial idea I had for the thing. It worked pretty nicely, though it did end up walking too much in the same places on open areas.

Later on I added simple pathfinding: basically it marked how long it would take time to go until the closest free position. I guess I even had a small bug in it, since it sometimes seemed to take a bit longer road than it should have. I couldn't debug it properly and I was about to change the code - I just happened to find a girlfriend and that took all my interest out of the project. Nasty?

What I was about to do, was that it would rely more on the pathfinding - the code I submitted relied too much on the "avoid certain position" code, sometimes overriding the pathfinding. It caused some extra steps to happen here and there. I also would have liked to do speed optimization, but since it didn't matter in the contest, I submitted the code as-is. As you can read from before, Electroman really loved it unoptimized ;)

[alkatran]

Originally posted by Merri
The code also marked how many times it had been in a certain spot, thus avoiding going to directions where it had already been at.

You weighted it? I did a pure "if you've been there and you can go where you haven't GOGOGOGOGO otherwise backtrack to the closest open spot and GOGOGOGOGO"

My program never looked at an empty spot without moving onto it.

[StevenHickerson]

Mine was a lot like Merri's up till a couple days before the end. I did have it randomly choosing which open direction to go to but the scores were to unreliable. So I changed it so at the beginning of the maze it randomly chose if it would go left to right or right to left. It would move through looking and recording all directions in a memory matrix but following its original random.

If it found itself in a position where there was no open spaces next to it, then the pathfinding code kicked in and it found the nearest unexplored space that it knew wasn't a wall.

In my original design I marked deadends and stuff like Merri, but in my final it was unecessary to mark dead ends. Because of the way I worked it, it would never visit the same place twice unless it was pathfinding back out of a dead end or something like that :)

I did leave some redudant code in, in my memory matrix it recorded the direction taken to leave a particular square, I had originally used this in order to backtrack out of a dead end and to be able to go down an unexplored path, but once I put the pathfinding in it was obsolete but I never removed it.

All in all my final design was nice.. considering it would've won had it not been for that one bad choice at the beginning of map 4 hehe ;)

[alkatran]

[i]I did leave some redudant code in, in my memory matrix it recorded the direction taken to leave a particular square, I had originally used this in order to backtrack out of a dead end and to be able to go down an unexplored path, but once I put the pathfinding in it was obsolete but I never removed it.[/B]

That's what i was doing at first, too. Then I switched it so it recorded the move index (how many moves had been used to reach that square), which let me backtrack etc just as easily. Come to think of it, I could have just stuck with the directions... bytes instead of longs... grrr

[Merri]

In my original thoughts I was planning to mark deadends in a more complicated matter. I figured out how I could do it, but it would have been too complicated for the code structure. Or maybe not... I'm still not sure. Making it to work properly with open areas would've been the hardest: it should've mark where never go again and yet keep up fastest possible roads out of the dead area... so, making a real pathfinding was easier.

Anyways, my final code if I ever did it would've been a lot different from Steven's on detail level, making it interesting to compare how it would have competed.