Відео

Because [no code shown onscreen, though obviously there's code running because that's how the game works] doesn't fit in a title.

The Discord is a great community, and very helpful.

Brrraaaaaains...

Wow... just putting sunflower unlock earlier and making sure to keep up power, I gained 10 places on the leaderboard. Kicking myself now for missing that! 😆 Now to code up some maze solvers to try out.

I had some luck at Cactus by planting two at a time and kind of moving them together. There's definitely some room for improvement there. In the Maze, the algorithm shown here doesn't even try to remember where the walls are. You can make a better algorithm if you do. Checking a wall is cheap if the wall is still there, but expensive if not. Since the expensive cost for each wall is only paid once, I've gotten good results by checking every wall opportunistically, whenever I happen to be next to one.

@@zapakh6240 Oh, cool, I thought a `move()` call was 200 even if it failed, that makes things nicer.

@@danielrhouck I believe the cost of checking a wall and not moving is a cost of 1 while moving is 200

​@@zapakh6240this was my solution to maze as well. For cactus. I can see an ideal function. It has two parts. Initial pass where it opportunistically sorts any pairs memorizing each cactus position and size. Then run an algorithm that finds the minimum number of switches to meet sort condition for that configuration. Since it's on data via a helper fn, should be cheap. Then actually do it. This should technically be optimal. Hard part would be the helper fn. It wouldn't just need to sort rows. But consider switches across rows to tuly minimize steps. I can see a way to do it. If you take the sorted state and the initial state as comparisons. You can keep a list of bad matches between and graph positions by size of cactus. Reduce the list by turning the whole problem into graph traversal. This can be done with dictionaries or array of arrays since size is int. This should be fastest algorithm.

The overall sequence is hand-written, but the code moves on to the next segment whenever it has enough resources for whatever things it's supposed to be unlocking. So it's not "Do this so-many times and then buy Trees", but "Do this until you can afford Trees, and then buy Trees".

If you do, consider checking out the Discord. The community has been very active and helpful!

Omfg aaaaaa

Just did the same thing and I even saw this comment lol

very important comment here

I'm glad i was only a few minutes in before i read your comment.

Goddammit! 😂

There have been rebalancing changes since the video. Dino cost is higher now, but also cactus yield is multiplied by world size. In the video, I'm getting 192 Cactus per field, but today it would be 1536. It works out to require about the same number of Cactus passes.

I can help you out, but the new mazes have walls that are not connected to the other parts of the maze and that makes it 10x more difficult

Ah these mazes also have that

Try following the left hand rule - keep your left hand always on a wall. If you are able to turn left, do so. If you can't go left, go straight. If you can't go straight, turn right. Notice how I didn't say go West or East. "Turning left" depends on your previous direction. This description is if you always reset a maze so all of the walls are intact. Once the walls start to disappear, you'll need a different method.

I added a link to the description to get you started.

@@magefreak9356 jup had the same problems, but i think i found a decent solution where when it can't find the treasure after a certain amount of time it resets the maze. When it finds the treasure, measure it so you know the coordinates and then do an astar algorithm to now find every single treassure after that.

I have 200 hours now, but at the time it was more like 70.

Just re-read it also. Yep, you only need to buy leaderboard and then exit out with the same command to start and your done.

But it looks exactly like a following the left wall

@@Alexander.Kravchenko each time he reset the maze, a wall disappear, creating loops. this kills the left wall algo

You can also do A* meet in the middle which is way faster than this algorithm

@@Nichtdu-rt4ih nah A* isnt very good, and making it meet in the middle doesnt help. Meet in the middle BFS is much better but also needs an additional high depth algorithm, to make the drone move straight to the treasure, to actually get good times

​@MrBlobfisch Since comp time is different to event time. I'm wondering if you can't just map out all the walls, and assign moves to coordinates in a hashmap through exploration via left wall following a specific algorithm. Once you have the possible moves, run a bfs. Should always take shortest path. I don't know how the game checks for walls in the maze. But if it uses a move command rejection then this can't be done sexily. But if it's a simple bool check without movement. You can track walls such that as you move you check and update walls freely. This should almost always beeline towards treasure. I'm making assumptions though. If every "cycle" is a move then this is expensive and a left wall algorithm with backtracking is actually better.

When you harvest a pumpkin, it harvests the entire pumpkin. If your whole field is one big pumpkin, the whole thing gets cleared. It's a lot of pumpkin! I sometimes deliberately reset during Sunflower (15:10) because long Sunflower sections tend to get inefficient as you're repeatedly waiting for the newest plant to grow. Most of those resets ended up being removed because I opted for multiple shorter Sunflower segments instead.

@zapakh6240 That's so cool to come up with. Well done!

Maybe someday, when it's no longer competitive

@@zapakh6240 I'm struggling with the cacti, could you help with some sort of algorithm for them

@@zapakh6240 maybe now is the day

Oh hey! You found me.