Show HN: I found a way to 'cheat'/solve 2048 (With live demo) | Hacker News Reader

Show HN: I found a way to 'cheat'/solve 2048 (With live demo)

4 pointsSchizoDuckie10 comments

After a loong long time trying to beat my girlfriends' 11.000+ highscore on the original 2048, I found myself working according to a standard method: move right, move down, repeat, until you're locked.

Then move left, move down and start from moving down and right again.

To test this theory to the max i've injected a little button into 2048 and it executes this method until you pause it.

http://schizoduckie.github.io/2048/

It works surprisingly well if I say so myself :)

View on HN

Yeah, I'm not getting any spectacular scores myself either. I ran four games and never had a tile bigger than 256. Thanks for sharing though - fun to see.

You should probably remove the "Note: This site is the official version of 2048." footer from your version.

Done :) I've also added a little tweak: It stops when the board is full up until the last 2 open blocks, from there on you have the chance to fix it manually :)

That's pretty much the method I used to solve it. You aim for the right hand corner using only right and down moves. Your goal is to keep the rightmost column fully merged. This allows you to make the occasional 'Up' move when you need to merge tiles from the third column into the rightmost column.

You should not make any 'Up' moves unless you absolutely have no other move. You should not try to merge large tiles from the bottom of the third or second column into the third. Just let them build up on their own and eventually the top of the columns will be able to merge together and you can then merge the rightmost tile and fill it up again.

First priority is to keep the rightmost column completely full.

Wow. I've just got the highest score i've ever made using this method. Very awesome! If i get the hang of this i'll try to wrap it into the autoplay >:) :P

I saw someone else post about this "corner" method and the highest score I've gotten is ~5000, with my highest square being 256. Before that I got to 16,696, with a 1024 square being my highest single square, just by moving it randomly, and I consistently get over 10k. My goal is to get a 2048 square, which is why I tried this method, but to no avail.

What if you change the direction? Move right, move down is focusing bottom right corner. How about switching with interval between corners?

I've experimented with this and it doesn't really matter. If you think about it this this works exactly the same way as an abacus would.

I ran it 5 times and the best it got was 3884, which is about what I expected. I submit that you just got lucky :)

http://ov3y.github.io/2048-AI/

That's actually trying to be intelligent. Mine just uses the simple down -> right -> repeat until locked rule.

Reply on news.ycombinator.com