A fun little piece of the source:

<!--[if lte IE 6]>

If you read this it's probably because the world has rebooted. I don't know how but you arrived here. I'm sorry you cannot enjoy this little game but you will be able to in the not too distant future. Browsers will become incredibly better. Just before the world turns into chaos due to Nutella production shortages. Scientists will realise too late that Nutella was to humans what pollen is to bees. Don't believe the hype, the end didn't (or won't) happen because the 'left-pad' package got removed from NPM. Good luck.

This is a great puzzle game and a very impressive demo web app. However, I wish it was available from the various app stores for two reasons:

1. I want to pay you for this.

2. It will be ripped off a dozen ways, and I would rather see you make money rather instead of some horrible knockoff app-vendor.

  Cirulli himself described 2048 as a clone of Veewo Studios' app 1024, who has actually said in the description of the app to be a clone of Threes.

Solved the first puzzle on "easy" in 53 attempts. The first 20 attempts were me figuring out that the 4 dots beneath each try are not in the same order as I drew the pattern, so I can't directly deduce if a single dot is correct or not. Then it took me another ~8 attempts to find the four correct dots and then a very long time to get the order right.

Fun game, interesting idea and nicely done! I'll probably play a few more rounds when I get home.

I found the display of previous attempts a little hard to decipher (1-2-3-4 labelling would be much easier to parse, for me), but other than that great fun, and exceptionally polished presentation!

There is some "overflow: scroll" that should be "overflow: auto" in CSS, IMHO.. So many empty scrollbars in firefox on osx...

A very fun game, it feels like a more "computationally intense" minesweeper, though I guess I compare everything to minesweeper[1][2]!

It seems the optimal strategy is to initially try to ferret out which dots are in the solution set and which are not in the solution set via broad non-overlapping sweeps followed by slight pattern deviations to determine membership of particular dots. Depending on luck and your errors, this might take a half dozen moves, it might take two dozen.

Once you know which dots are in the solution set and which are outside, I find it takes less than 5 additional moves to determine order. For this part, a pencil and paper for tracking the elimination of certain dots from certain orders made this phase consistently quick.

Is very fun, I was ultimately able to beat the medium difficulty setting in an average of 15 attempts per game.

[1] - https://github.com/lelandbatey/defuse_division

[2] - https://github.com/lelandbatey/minesweeper-solver

Maxwell, you have a spelling mistake on the Success page.

"Lock found in 20 attemps. You must hate this game by now."

I saw this game (Mastermind or originally called Moo or Bulls and Cows) in an article in Software Practice and Experience[1] in the 70's. A couple of years later I saw a very clear explanation of how to write an effective software opponent, written by Knuth[2]. It was a lot of fun to implement back in those days (first in Fortran and then in Pascal). I recommend the easy paper by Knuth.

It's surprising how effective Knuth's algorithm is. If I recall 4 or 5 guesses to solve 4 digit mastermind.

[1] Computer Recreations, Software – Practice & Experience 1.2, 201-204 (Apr-Jun 1971) [2] D.E. Knuth, The Computer as Master Mind, J. Recreational Mathematics Vol. 9 (1976/7)

Pretty neat. Is there some kind of "strategy guide" for mastermind-like games? I feel that I'm operating at close-to-brute-force levels (I spent 22 attempts at a 4-lock) and I'd love to get better

Great time waster, worth making lots of guesses just to see the performance reports. My first go took 65 guesses which included working out the use of the 4 dots, the game appreciated my determination. my second go was 3 guesses but I admit I got very lucky there.

It's interesting how straight forward the game becomes after you recognize the pattern. Fortunately android lock screens don't inform the users about how many correct points they choose in the pattern!

What is the algorithm to solve this in the shortest number of moves. I have got it down to 7, but there might be something faster.

Yay, break the hard mode in 9 attempts! Kinda like Sudoku but more flexible.

Reminds me of another phone game I've played, 0h n0. Quite fun.

I beat it in two moves out of sheer luck.

Lucky to do in 10 attempts only

I don't know the answer to your question, but I was curious to know if 22 attempts was anywhere near a bruteforce attack on a 4 point pattern.

This Quora thread contains some interesting breakdowns of the number of possible patterns that Android permits.

https://www.quora.com/Android-operating-system-How-many-comb...

If we assume that you don't know the number of points in use on a phone you're trying to unlock, the consensus seems to be there are in the region of 390,000 distinct patterns. One correspondent doubled that.

In the case where you know the pattern has 4 points, the number of valid permutations was calculated to be 1624. Your 22 attempt crack is well short of that figure so I think you are being harsh on yourself by implying you basically just got lucky. You're good at this!

I wrote a program and I'm getting sub 6 for every easy game. Think of the board as a grid of numbers.

0 1 2

3 4 5

6 7 8

Then an L shaped pattern would be the string 0367.

The main idea is that you first build a list of every possible combination of legal moves. Then you take a random guess from that list and you will get back the number of white and black dots.

Now, since you know your random guess and the right answer produces X white dots and Y black dots, you can remove from the list of possible combinations every combination that has: dots(myGuess, combinations[i]) != (X, Y)

you can keep taking random guesses from this list and filtering after each attempt until you get the right answer, I'm not sure of the math but I think worst case It'll be 6-7 moves to find the right answer.

I think trying to exclude nodes is more efficient than trying to find the correct ones.

After you excluded all "bad" nodes, you need to find the order. finding the order with the help of "bad" nodes was also faster for me

I went from 56 to a lucky 5 tries.

Yeah, as Spare_account said, 22 attempts is probably a pretty good result for a first try at the game. I needed 60 :).