Small change, but 2 less walks and 1 less kick now.
Also I was trying to beat my record on 2-3, thinking maybe a perfect there is possible, I tried hundreds of times and tied 8.16 over 10 times, but no dice. I had a rule that if I ever tied a non perfect record I would keep trying until I beat it, but I think I'm going to have to give up on that one. Also got a good time in 3-1, don't know how, because most of the time I was consistantly way slower than that.
Edit:: Sorry Gorash to steal your 10-3 cheaply (I used my program to get the optimal solution) If I find any more optimal solutions with it I won't be cheap again though, well maybe once more if I get a level where there is no BisqBot results yet, but only once more :)
new record for 11-4. It was fun going back and forth on 11-3 and we improved the level more than I thought imaginable. Hopefully we can do the same for some other level. This level is improved alot and I don't know if it can be improved more but some parts don't seem clean.
Writing a program to find the optimal solution to me isn't any less "cheap" than finding it yourself. That's why programs are written. Writing the program I'm sure wasn't easy. And what if someone else plays the level using the same route? Are they more justified because they discovered it in a different way?
<Swordless> Go hug a tree, you vegetarian (I bet you really are one)
Hmmm, k flagitious.
I tried about 4-5 different changes in 11-3, but none is any faster than the current version.
I also took a glimpse at you 11-4 but I haven't spent much time on that level after solving it initially, so I can't really rate your path. Looks good to me though.
I'll go back to some of my old records I think. The 8-x ones are still untouched. Seems like noone likes those levels. :)
About the solver issue:
I also think it's cheap to use them. I'd be fine with a solver that finds better pathes heuristically by changing the current path (which is what we do). But the perfect path basically removes the level from any further competition.
Not that it matters though, there are only 2 levels at the moment where the optimal path is computable but not accessable.
I can't find any further improvements :(
Edit::
Gorash, don't know if you had worked on 8-1 because Samurai G. had the record, the other world 8 levels are too long for my time attack taste, but were good puzzles when just playing
Time 2:18.03 2:32.96
Steps 479 545
Pushes 131 121
Kicks 19 24
Yes I was the original solver of all of the 5-x and 8-x. The previous record was already not bad, and Goroh did a decent job on that I thought.
Let's see where you found 15 seconds.
EDIT: Watched it, completely different approach, but has good ideas in it. You also discarded one idea from the previous movie, which I will try to combine with the current strategie. Will ake me some time to analyse, which I don't have now. This evening I hope.
I think it was me who really set the level 8-1 record & gorash improved it, but I recuperated it... (now I lost it again :( )
Your strat looks great, but I think you can save like 2 steps when the time is around 45 secs (when you push a block left then up), think that instead of pushing that block up, go to the one in the right & push it up to set it closer when you return for it...
BTW, have you tried to beat my level 4-5 record? I set this one long time ago & never got beaten :D (But I wanna see how much can it be improved...)
Also flagitious, how many levels do you still miss, if any? (I still miss around 10)
I tried pushing that last bottom block up before pushing the other block but it didn't seem to save any steps, maybe I don't understand.
I'll try to beat 4-5, but it looks pretty complicated so it will take me a while to analyze. I am still missing 3 levels, in world 10.
Also slight improvement made to level 4-1 saving about 18 frames.
Errr, if Goroh says so he'll have solved it first... whatever, it still is one of the best levels.
I found 6 frames in 8-1 so far, but I want to try a few things before I implement it.
Well after looking at 4-5 I don't really see any major improvements possible. Some minor things that might save a few seconds are:
At the start push block down, and then push right, so you can put that first block down in that group of 4 goals on your way all the way around, not sure, but seems faster.
Before going back to the top left around 2:30-2:45 move blocks in the line into the goal, so that when you come all the way around again you can kick up from bottom right.
Around 3:00 you go afk or something :)
Other than that it looks good, it would take forever to try out all the little things that might or might not save time like I try for other levels. So for now I will bow down to this level. BTW it has the highest bounty of any level.
Gorash, I look forward to seeing the improvment in 8-1. Mean while I'll be checking out 3-5, it is pretty short and looks like it has potential time savings.
It would take forever to try out all the little things that might or might not save time like I try for other levels. So for now I will bow down to this level. BTW it has the highest bounty of any level.
Meh, it's been so long & think that no one else had that record, so I wanted to see a video from anyone else... (though, it would be really hard for anyone to improve this level...)
flagitious wrote:
Mean while I'll be checking out 3-5, it is pretty short and looks like it has potential time savings.
Really? I though I perfected the strat... In which part do you think it may be improved?
BTW, think I can sub 5 min my level 7-2, the current strat has some notable blunders...
Well I never got to see the 6 frame improvement in 8-1, but I think I can guess what it was. Was it ordering the 4 blocks in the left area so that the new block can be brought in sooner to avoid wasting 2 steps going around it?
I improved the level by 18 frames, but in the process I think I made it impossible to do that 6 frame improvement, but hopefully you can find a way to still incorperate it.
For 3-5, I am certain of 2 improvements (1 is only 1 frame, the other 1-2 seconds), but the level is somewhat complicated so I'm still searching.
7-2 gave me some headaches first time through. I remember that I would be dumb and not realized a mistake until wasting several minutes, good luck with it, fun to break minute barriers :)
Edit::
Ah crap about to fall asleep and I realize how to save 4 steps in 8-1 I think, will get it in morning hopefully.
Edit2::
Nope, I could only make it a 2 step improvement, and it cost 5 frames, so savings = 7. (Change was pushing top left block in last to take better advantage of where peter is after putting blocks into finish area. I don't see how, but it might be possible to save 2 more steps taking better advantage of this.
No, the 6 frames where letting the 2 blocks in the lower right be, to kick out the blocks from the left.
I also tried to put a kick blocker stone above the goal entry, but that costs way too much in all variations I've tried.
The third thing I was trying was to prefetch the blocks in the left area better, but again, nothing saved, always lost.
Yeah the kick blocker there cost about 50 frames to setup and saves 5 frames each use. It was nice that at least 1 kick blocker could be setup at no cost though.
Level 3-5 beat by 5 seconds due to organization of placing blocks in goal so that less trips are needed, plus minor other changes.
Level 3-3 beat by 4 seconds by pretty much the same change in strategy. For some reason that glitch 'where when you click right as you kick and you don't walk anywhere' happened to a ton on this level, it was my number 1 cause of aborts, I'm going to try to figure out better why it happens. This level now has the lowest time of any level where the optimal solution is not known.
Edit:: 4-2 improved a little, 3-3 no longer fastest non optimal :(
Edit2:: 11-1 and 3-2 beat by a couple seconds each, but using very different routes.
Edit3:: 4-2 improved more
Edit4:: 8-5 beat by almost 1:30, still lots to improve in terms of execution, probable that minor route changes might help too.
I just mean the current best time is no longer the fastest of the non optimal known levels. It is possible that optimally it would be faster than all the other levels optimally that aren't optimally known now, but not probable. Ah, that sounded more confusing than it should be, but I don't know how else to word it.
Well it has been a while, I don't think I will be working on my peterbox program anymore. I am sad that I wasn't able to find optimal solutions for any new levels. But I think if Bisqwit and I combined our methods a couple more levels would fall. From what I know of how BisqBot works there are 3 improvements that aren't too hard to implement, 1 minor, 1 major, the other soso. I could be wrong though, maybe you already do some of the things. Anyways if anyone is interested I will elaborate and give code/pseudo code.
Has anybody played lately? Seems that trying to find better strats it's a lot harder now...
Hope that when my parcial exams finish, I can get some records & beat the levels I still miss...
Have not been playing. A while ago though I found this site with other sokoban puzzles: http://www.js-games.de/eng/games/sokoban
The implementation is much much worse than bisqwit's, but I like the puzzles more. They are smaller, so at first you will think meh that can't be hard, but most of the puzzles there were harder than any level from the original peterbox in my opinion. The puzzles there were actually taken from some Japanese dude's site, which is linked to from there. The site has ~60 puzzles, 20 of which are used at the jsgames, I have not tried to do the other 40, but I can image they are just as good.
