Hey Kamikaze, I can't quite remember what you asked us to do regarding the 1 / 2 split. I think it was get a difference of more than four right? If so I have just for it: - 1 6 48 - - - 96 - 1 24 6 1 1 - 6 (screngrab below) 5 ones, no twos. Hope that hasn't broken any theories I suspect it doesn't as the next move would make it 5 ones, 1 two. But I thought t was interesting anyway!
Let me ask you in response: did you achieve this situation after the tutorial? If so, then it's just a bug.
Nope normal play... Haven't don't the tutorial since day 1! Also now this, 5 ones no twos and the next is a white card! Edit - this is a carried on game, I started it last night and didn't finish it. Maybe I had a glitch...
Maybe I should clarify. The tutorial doesn't end in Threes! - it just converts into a normal game. If this is a game which began with the tutorial then it does not surprise me. If, however, this not such a game then it would really surprise me. Maybe something went wonky with loading the saved game? Did you at one point maybe close the game right after performing a move? Maybe that put the card draw algo out of whack.
Wish I could program like Nicola! I suspect if I actually stepped through simulated games like his, I'd eventually be able to see better options myself. I'm still stuck at 384 and usually am below 2000 pts.
It wasn't started from the tutorial in anyway! I have played loads I don't actually remember when I closed the app, but it could well have been! I would put it down to a glitch, until someone else comes up with a similar situation
Ah, I was wondering what that one-point achievement was! I'm guessing the previous four achievements (shown as 'Hidden' on my device) correspond to 12-lock, 24-lock, 48-lock, 96-lock? If so, it seems that no one in the real world has achieved any of these yet. I've managed both 3-lock and 6-lock, mostly by pure luck. (Luck-lock!) Perhaps there should be an achievement for "Color-Lock" as well. (checkerboard white and non-white tiles.)
Nice Fractals! (I'm a bit of a fractal nut myself: you can see my own app at http://fract.al/.) The relevant question I was trying to address with the 'atomic number' system was: What's the maximum number of simultaneous cards on the board (e.g. { 3 3 6 12 24 .. } ) that were required to create a given high card? To give the example, if you have { 1536, 768, 384, 192 } on the board and a 192 bonus card appears, then (if you ignore how the 1536 etc. was created, because that happened much earlier) it only takes 5 cards to construct the 3072. My system would give the 3072 an' atomic number' of 192 in this case. But if you don't get any bonus cards, then it would require a snake of { 1536, 768, 384, 192, 96, 48, 24, 12, 6, 3, 3 } to create the 3072: that's 11 cards, reflected by the atomic number 3. Mathematically: difficulty = max # of cards required = log2(high card / atomic number) + 1. Of course, for the 1536, you would know at the time of its creation what its own 'atomic number' was, and thus the level of difficulty involved in creating it. Each "critical phase" of the game (making a new big card) will thus have its own independent level of difficulty. E.g. you might have a crazy time getting the first 768, then luck out and get the 1536 and following 768 fairly easily, but then have a bear of a time with the 3072. The 'atomic number' system would provide a way to quantify that. This also reflects one of the fascinating paradoxes of the game: extremely high scores are only possible because of the bonus cards, but the bonus cards can also throw a massive wrench in the works when they appear! I expect that the '+' hints in the upcoming version will help greatly with this. I currently can achieve 768 regularly, and I suspect with the hints I will be able to crack 1536 regularly, or at least occasionally.
i have a couple quick issues, one if most of the times i go back into the app there is no sound, need to fully close through multitasker and relaunch to fix. Also many times opening the app, there is a 5-8 second delay (just a white screen). Running an iPhone 5S.
I can't compete with that but I do share your affinity for fractals. My current favorite (and wallpaper) is the dragon curve. Back to Threes! The audio thing is a Unity bug and out of the control of the devs. The sporadic long loading times are not new to me but I can't help you there either.
I just saw the alert on my watch list, it's not showing on my iTunes yet though, hopefully it won't be long.
Yay for the update Try and keep the high number in the corner, doesn't need to be but helps to start. Concentrate on combining 1's and 2's. try and line up multiple combines, so one slide combines several cards. Try and plan where the next card will come in and try to get it to match or be able to match quickly That's about all the tips I can give. I can get 384's fairly regularly but am stuck getting higher. Got a 768 once though!
To recap the most popular tips that can be found scattered around this thread and other resources. 1. Slow down. 2. Think before you move. (Use the slow drag and hold to see exactly what's going to happen) [or as Yoda would say: "do or do not, there is no undo"] 3. Learn the rules for where new cards come in, i. e. only in rows or columns where cards moved can a new card appear. Use it to your advantage 4. Plan ahead. 5. Keep your board as empty as possible. That especially applies to 1's and 2's. If you don't need them for intended blocks (advanced tactic), remove them immediately. 6. Just because you can merge two cards does not mean that you should. So long as you don't shift the two cards apart, they represent an opportunity that you can use later. 7. Know that 1's and 2's will always balance each other out. If there are more of one than the other, the corresponding color will come soon.
People, don't dismiss that link as just a throwaway curiosity. That app is mind blowing. If you have even a passing interest in the Mandelbrot set, you have to take a look at it.
Actually, I only played two real games with my bot too. I scored 88,908 and 202,542 respectively. Maybe I'm just lucky My bot plays via screenshots; I screenshot my iPod and the bot picks it up, OCRs it and tells me what move to make. It's slow but not too bad, because at least I don't have to manually copy in the board state. The bot uses a single 64-bit int to store the entire board state (4 bits per entry, for a maximum representable limit of 12288). Moves are just bitwise operations with a lookup table, so everything is meant to be blazing fast. I do search down into bonus cards, though I use a probability cutoff to avoid searching highly unlikely states (e.g. three bonus cards in a row). It also tracks the deck state precisely, which allows it to make inferences like "the next card must be a bonus card because the deck has no 3s" (and more generally, compute the probability that an upcoming white card is a 3 based on the deck state). Given that the "+" update *just* landed, I will probably attempt another run later this week using that new information.
Yeah, I'm starting to wonder what happens if we combine the longer lookahead with a better position scoring function. Currently, my bot also delays merging simply by virtue of seeing far enough into the future (it knows it can create many more open spaces by delaying a merge). I tried with the penalties but the scores didn't seem to improve (again, I'd like to do this more rigorously with a statistical evaluation of performance, but it would take much longer). I suppose the question is: assuming it's easy to look five or six moves ahead with nearly-perfect clarity, how would you best score a position in the future?
