Metanet Forums

because this thread just had to be created

http://gabrielecirulli.github.io/2048/

arguably the most productivity-killing time-waster i have ever played since n v2 came out

still, my life is complete now

I have this many posts now.

▲▪►▪▼◄▲ wrote:I have this many posts now.

Please ignore this person.

I hate you two so much ... >< This is my single biggest in-class time sink ever now.

This was way more addictive than expected, just completed it, a little more optimized than yours mwahahaha, I now challenge you to go for sub-20k points, its technically possible but it relies on luck.

Full pic: (http://i.imgur.com/KH8uvkj.png)

EDIT: Even more optimized!

Full pic: (http://i.imgur.com/ebhyJjY.png)

I was testing this 2048-undo version, but I didn't really require to undo any movement in this particular match, though in case you are reticent to believe it, I might go for a better one in 2048.

Regarding score possibilities:

If only 2 tiles were to appear, the minimum score needed to build a 2048 tile would be 20480 (2048 + 2*1024 + 4*512 + ... = 2048*10, because the initial 2's don't count towards the score), but that would be perfect ensamblation, however since you have a 10% chance of having a 4 tile to appear, you can substract 10% of the 4's row that will appear on their own, that is 204.8, hence 20275.2 should be the average minimum score required. Seems like I was lucky with my initial 4's :P

18432 would be technically minimum with 100% initial 4's, which is obviously very unlikely to happen (at least 1/10^510 probability, since 510 is the minimum number of movements you'll need to make to make it to the 2048, as each time you make a move, the sum of tiles on the board increases by at most 4, actually always by 4 in this ideal situation, but in a real situation it would take almost double as most tiles are 2's).

Now I'm too lazy to calculate the chance of getting a sub-20k average, but it doesn't seem too low, with a clean technique you would probably get it after 3 wins or so.

I'm simply amazed at the intelligence levels of some of our members. You guys stun me every time.

Wow thanks Aidiera, I have fun doing those little analyses. Anyway, I kept playing that second match I showed, and managed this:

Full pic: (http://i.imgur.com/UAtXSqd.png)

Now, I did use the Undo button in several occasions to get that second 2048 piece and merge it into the 4096 one, so its not completely legit. I guess the next goal is to do it in the original version without any undo option. Its perfectly doable though, as most of those undos were just because I lost concentration and moved the main row of tiles, and if you have played 2048 you know that just fucks everything up haha, most of the times its just unrecoverable.

Anyway I'm not continuing this match, mainly because the next step is way too far and I'm running out of time, and also because it now wouldn't be a fully legit game anyway so its not worth it. But damn its addictive.

Total time was about 30 mins for the first 2048 tile, 40 for the second one and hence for the 4096 tile, and 20 minutes more for the other smaller tiles that you see, so a total of 1h 30min :P

psh. i beat 4096 LAST MONTH. the ball is in your court, eddymatagallos. :P oh, and without any undos. let me take this very rare opportunity to point and laugh at you.

EddyMataGallos wrote:Regarding score possibilities:

If only 2 tiles were to appear, the minimum score needed to build a 2048 tile would be 20480 (2048 + 2*1024 + 4*512 + ... = 2048*10, because the initial 2's don't count towards the score), but that would be perfect ensamblation, however since you have a 10% chance of having a 4 tile to appear, you can substract 10% of the 4's row that will appear on their own, that is 204.8, hence 20275.2 should be the average minimum score required. Seems like I was lucky with my initial 4's :P

18432 would be technically minimum with 100% initial 4's, which is obviously very unlikely to happen (at least 1/10^510 probability, since 510 is the minimum number of movements you'll need to make to make it to the 2048, as each time you make a move, the sum of tiles on the board increases by at most 4, actually always by 4 in this ideal situation, but in a real situation it would take almost double as most tiles are 2's).

Now I'm too lazy to calculate the chance of getting a sub-20k average, but it doesn't seem too low, with a clean technique you would probably get it after 3 wins or so.

ok i'll do it for you. i haven't checked the source code but i'm going to trust you when you said the 4's appear with 10% probability.

to get a score of under 20000, you said 18432 is the minimum score with all 4's, so the number of 2's you'd be limited to at most would be (20000-18432)/2 = 784. then to make 2048 you'd need the rest of them to be 4's, so that's (2048-2*784)/4 = 120 fours. keep in mind this is a lower bound for the number of fours and you'd most likely need several more 4's than this, since even assuming perfect play, when you do the 2->4->8->...->1024 thing at the end that's 10 more steps that give you tiles that merge together and give you more points, but i'll ignore those.

to get the chances of getting at least 120 fours with a 10% probability of each occurring within the first 902 tiles you use a binomial calculator which tells you there is a < 1/1000 chance of getting 120 or more fours, and it drops rapidly if you ask for even a few more fours.

so yeah, i really doubt you'll ever beat it with sub-20000. :P

of course, using the undo function on 2048-undo a lot of times until you get however many 4's you wanted would obviously work, but that's cheating.

999_Springs wrote:psh. i beat 4096 LAST MONTH. the ball is in your court, eddymatagallos. :P oh, and without any undos. let me take this very rare opportunity to point and laugh at you.

ok i'll do it for you. i haven't checked the source code but i'm going to trust you when you said the 4's appear with 10% probability.

to get a score of under 20000, you said 18432 is the minimum score with all 4's, so the number of 2's you'd be limited to at most would be (20000-18432)/2 = 784. then to make 2048 you'd need the rest of them to be 4's, so that's (2048-2*784)/4 = 120 fours. keep in mind this is a lower bound for the number of fours and you'd most likely need several more 4's than this, since even assuming perfect play, when you do the 2->4->8->...->1024 thing at the end that's 10 more steps that give you tiles that merge together and give you more points, but i'll ignore those.

to get the chances of getting at least 120 fours with a 10% probability of each occurring within the first 902 tiles you use a binomial calculator which tells you there is a < 1/1000 chance of getting 120 or more fours, and it drops rapidly if you ask for even a few more fours.

so yeah, i really doubt you'll ever beat it with sub-20000. :P

of course, using the undo function on 2048-undo a lot of times until you get however many 4's you wanted would obviously work, but that's cheating.

Haha well played :P

You have a point on the undo thing, though thats not how the undo works, when you undo a movement, and do it again, the very same piece appears on the very same place, it kind of saves it you can test it out by yourself, so you can't use it for that, but rather to fix things up if you mess them. I'm not trying to justify this playing mode though, its just for clarification, I do consider it cheating as well.

I'm impressed at how low the chances are haha, my 20128 run required no undos and it was my second completion only, thats why I assumed that after some more tries and by polishing the method it would be reasonable to attain that, but it just seems that I was lucky (or either I'm wrong with the 10%, but all sources seem to agree on that, I haven't checked it by myself either). Plus sub-20 would mean 32 more 4's, which now that I see it that way, would have been pretty unlikely.

EDIT: Anyway, here you have an even more optimal game, this time done in the actual 2048:

Full pic: (http://i.imgur.com/UzSUf7v.png)

I don't get what's so engaging in this game?

so i have these 4 exams which are worth almost 100% of my entire undergraduate degree, i had one yesterday, and the other three are next week. so instead of revision, i chose to spend my time on something that i consider more productive, which as a result means i'm probably going to fail my degree.

Hah nice one. Thats the curse of the procrastinator I guess.

Damn, so, the other day I was playing this, not in my computer though but in an iOS device, so I'm not sure whether the conditions change (I'd say no), and managed a sub-20k completion. I unfortunately didn't consider leaving that until being able to picture it at home, so I kept playing, but I swear I got it man, I wish I had thought on taking a picture.

ios/android devices have a way higher 4 spawn rate. this is pretty obvious if you're actually playing it for a while. also it makes it much harder to get 4096 or 8192, because random 4 spawns that don't go with the 2s on the board are often what kills you. i have a screenshot of 2048 on android with <20000 which i'll edit in this post as soon as i find my usb cable

i almost got 16384 on pc :(

999_Springs wrote:ios/android devices have a way higher 4 spawn rate. this is pretty obvious if you're actually playing it for a while. also it makes it much harder to get 4096 or 8192, because random 4 spawns that don't go with the 2s on the board are often what kills you. i have a screenshot of 2048 on android with <20000 which i'll edit in this post as soon as i find my usb cable

i almost got 16384 on pc :(

Indeed it is, that was my first completion there and I thought that it might have been big luck, but my second completion was also sub-20k and its not particularly clean:


So yeah, it becomes clear. Just did a sub 19.9k now.

hello

I WIN AT LIFE



this was my 11th try getting to 8192+4096+2048; i messed up the first 10 times i got there

eddymatagallos i'd like to see you have a go at beating this :P

edit: i just looked back at my first post back in march and it really frustrates me how much my crypto investment has failed. the prices in the tabs were 624.4 usd/btc and 0.02786 btc/ltc and 128 sat/doge back in march and the current prices are 378.5 usd/btc and 0.00958 btc/ltc and 63 sat/doge

Haha damn, that's pretty crazy. I'm quite sure I must be missing something, because there are some situations that happen every now and then in which I just have to move everything out of the corner because its the only allowable move. Clearly, the possibility of screwing everything up is strong here :P I should develop some strategy to avoid that.

I never even got to the 8192 piece due to that. I managed the 4096 a few times though.