Hey smokie,
I made 15 ciphers with 63 symbols etc, originating from a 332 character p1 plaintext. 5 with period 10, 5 with period 20 and 5 with period 30. Each has 8 nulls at random positions and a different substitution. The idea is that with 3 different periods and 5 ciphers each, any differences will average out and a trend should be visible between period 10, 20 and 30. Will run each with 10,000 hc iterations and 500 restarts.
Ciphers: https://drive.google.com/open?id=1TPi4i … 4XDl7e0NlW
smokie, I have added a shortcut to the hill climber, could be a setting, if the score is over 22500 or whatever specified the hill climber skips to the next restart since we know the cipher is solved anyway. Let me know if you want this.
All of the above sounds good.
smokie, I have added a shortcut to the hill climber, could be a setting, if the score is over 22500 or whatever specified the hill climber skips to the next restart since we know the cipher is solved anyway. Let me know if you want this.
I am o.k. with or without it. I wouldn’t want it to be mandatory though.
Hey smokie,
I made 15 ciphers with 63 symbols etc, originating from a 332 character p1 plaintext. 5 with period 10, 5 with period 20 and 5 with period 30. Each has 8 nulls at random positions and a different substitution. The idea is that with 3 different periods and 5 ciphers each, any differences will average out and a trend should be visible between period 10, 20 and 30. Will run each with 10,000 hc iterations and 500 restarts.
Ciphers: https://drive.google.com/open?id=1TPi4i … 4XDl7e0NlW
This is done and apparently 8 nulls at random positions is a breeze for the current state of the algorithm. This causes the results to be a bit unclear though it shows that it had a harder time on period 30 than on period 10 and 20.
Will do another run, 12 nulls & skips at random positions, start at period 1 and increase the period one by one.
I am glad that it was a breeze, but your solve percentages are so much higher than anything we had before I do not understand why.
There is a P20 8/0, 4/4 and 0/8 to see if any are easier than the others
I am glad that it was a breeze, but your solve percentages are so much higher than anything we had before I do not understand why.
Only nulls may be easier than nulls & skips combined and/or only skips. It should be because nulls & skips combined increase the search space since every position can be a null or a skips. I will run tests to determine this.
So far period 1 to period 15 is done, with 6 randomly placed nulls & 6 randomly placed skips per cipher. The plaintext is the same and the encoding is reasonably similar. Since the nulls & skips are placed at random positions, difficulty wise, one set of nulls & skips could be much more different than another. The number of nulls & skips weigh in only if they are actively contributing to the difficulty.
I am glad that you did that. Difficulty has a lot to do with where the nulls and skips are.
I am not doing so well with smokie 5/5, am approaching 500 restarts and no solve yet. Could you please take a look at this and see if I can adjust? I am using performance mode 5 grams practical cryptography. I tried the 6 grams, but it slowed down my MIPS a lot. I need a better balance of solves to MIPS now. Thanks.
For smokie 5/5, using 1,000,000 substitution iterations and 40,000 hc iterations should equal to about a 2% solve rate per restart. Using the Reddit 5-grams should increase the solve rate further.
reddit1805, right?
reddit1805, right?
Yes.
The Reddit n-grams are great for natural language but may not be so good for language that is more formal. The Wortschatz (standard) n-grams are better all around. I consider all of Zodiac’s communications to be written in natural language. With natural language I mean how people most commonly communicate or express themselves.
There appears to be an issue where a null can be at a negative position, such as -13, -2, 0 etc. Looking into it.
O.k. I never noticed that. Thanks.
It is fixed in this new update along with a few other things: https://drive.google.com/open?id=1okt93 … DuWzIw_siJ
– Added solver option: (Substitution + period + nulls & skips) Over skip . When this is set to 1 the hill climber will skip to the next restart after the score of Over is reached. This can make the hill climber run much faster especially when the cipher has a high solve rate.
