The Online Life-Like CA Soup Search

[Sokwe]

The highscore for longlived patterns in 2x2 are messed up

I've confirmed all of the current long-lived patterns, but it will probably be messy again in a few days.

@Nathaniel
There is currently a long-lived pattern in the database with an estimated lifespan of 25120 generations, but its actual lifespan is 27986 generations. This is the only time I have ever seen an underestimate of a lifespan in the database. Do you have any idea why this is?

Speaking of long-lived patterns, there is a known 20-cell methuselah that lasts for over 30000 generations that has been neglected for the last 1.5 years (See here and here).

Code: Select all

x = 16, y = 10, rule = B3/S23
14bo$14bo$15bo$9bo3b2o$7bo4b3o$7bo3bo$9b2o$bo$3o$obo!

[dvgrn]

Speaking of long-lived patterns, there is a known 20-cell methuselah that lasts for over 30000 generations that has been neglected for the last 1.5 years (See here and here).

30046 ticks from 20 cells... yikes, the Soup Search only very recently managed to improve on that survival time, and has yet to beat the bounding box or cell count. Guess that posting last year on comp.theory.cell-automata must have slipped past me somehow --

I can adjust the pattern to get 30047 ticks with just some random bit-twiddling in WinLifeSearch (or actually Karel Suhajda's JavaLifeSearch beta 3, but WLS would probably be better, since the cell-count contraints in the JLS beta don't seem to work...)

Very likely there's some more room for improvement in there somewhere, but it will take a more organized search than I'm willing to attempt this evening --

Code: Select all

#C Eve's Stepmother --30047 ticks, 20 cells.
#C The daughter pattern is 19 cells -- Eve's Little Sister?
#C All following generations are the same as Eve's descendants.
x = 17, y = 9, rule = B3/S23
15bo$14bobo$8bo4bo$7bo5bo$7bob2obo$9b3o$9bo$3o$o2bo!

[Nathaniel]

There is currently a long-lived pattern in the database with an estimated lifespan of 25120 generations, but its actual lifespan is 27986 generations. This is the only time I have ever seen an underestimate of a lifespan in the database. Do you have any idea why this is?

Wow, that's the first time for me too -- I wanted to avoid underestimates at all costs. The reason for the underestimate is because the only movement after 25k generations is taking place so far away from where the pattern started -- about 250 cells down and to the right. The script only searches for movement within an (expanding) box located at the starting point of the pattern, though (to allow for things like moving gliders to escape without being thought of as new movement/growth). Since no movement is seen in the "area of interest", the script thinks it's done.

Bleh, yet another thing to fix

[Macbi]

There is currently a long-lived pattern in the database with an estimated lifespan of 25120 generations, but its actual lifespan is 27986 generations. This is the only time I have ever seen an underestimate of a lifespan in the database. Do you have any idea why this is?

Wow, that's the first time for me too -- I wanted to avoid underestimates at all costs.

If we were doing a systematic search then underestimates would be a problem. Since we aren't, I don't think they really matter. I suspect that the probability that we miss a record breaker due to a bug like this is small, say about 1%, but that the speed decrease to estimate more reliably would be large, e.g. 10%, so that updating the script would decrease the rate at which we find new Methuselahs. In fact, I would seriously consider making the script less accurate, but faster. For instance, you could check at 1000, 5000, 100000, and whatever the current record is. This would be tuning the script just to find methuselahs, but I'm not sure if I care about the sixth-decimal-place-of-the-frequency-of-blinkers enough for the censusing feature to interest me much. (In Life at least, where probably all novel natural patterns are known).

[Sokwe]

@Nathaniel - Does the database still contain long-lived patterns that are not among the confirmed top 50?

I looked through a few of the longest-lived patterns in the database to see if any could be made into relatively small methuselahs. The only reasonably sized pattern I found was this 21-cell methuselah that lasts for 28128 generations:

Code: Select all

x = 24, y = 16, rule = B3/S23
11bo$10bo$11bo$12b2o2$14bo3$bo$ob2o18b2o$o21bo$b2o2$3bo2bo2$4b3o!

It is based on this pattern which stabilizes after 28158 generations (currently sixth in the database):

Code: Select all

x = 20, y = 20, rule = B3/S23
o4b4o3b2obo$o2b2o4bo2b2o2b2o$4bo2b2o2bobob2o$7b2obobo5bo$3ob2o12bo$o5b
o2bob3o2bo$bo2b2o3bobo2b3o$bo2bo7bo$o3bo4bobo3bo$b2o4bo2b2o5bobo$4bo4b
2o2bo3b3o$3bo4bob4o3bo$2obo7b2obob2obo$bo3b3o3bo2bo4bo$bob2ob2ob3obo2b
obo$b10ob2o3b2o$5bobo8bobo$2o3b2o2b2o2b2o2bo$2bo2bo3b2obo4b3o$bo5bobo
4bo!

[Nathaniel]

@Nathaniel - Does the database still contain long-lived patterns that are not among the confirmed top 50?

Yep, it keeps anything with a lifespan above 20,000. I'll see if I can figure out an easy way to upload them here or make them publicly-viewable.

[schneelocke]

.

[Extrementhusiast]

There are two rules that I am requesting a census for: B34/S267 and maybe B3/S013 (because of a lot of extra clutter). They both usually settle down within about 500 generations.

[Sokwe]

@Nathaniel
The total number of soups searched is not displaying correctly (it just says '0').

[Macbi]

New record methusala, lifespan 35426 if I'm not mistaken. Congrats to Schneelocke

Code: Select all

x = 20, y = 20, rule = B3/S23
5bo8bo2b2o$bo5b6o2bobo$3b6ob2o2bobo$o2bo2b2obo3bo4b2o$obob3o2bobo6bo$b
ob4o2b3o5b2o$bo2b3ob2obob3o$o2bo4bo3bo6bo$b2o2b3o3bo3bo$obo3b2obo3bo2b
obo$bo6bobo5b2o$6o2b3o2bo$2bo4b3o5bobo$2b2ob2o5bo3b3o$3bo2b4obobo2bob
2o$o4bo6b2obobo$o2bo2b2obobo$bo5bo2bo2bobobo$b3ob2obo$ob3obo6bo3b2o!

Slight optimisation, I'm sure others can do better:

Code: Select all

x = 25, y = 25, rule = B3/S23
19b3o$18bo2bo$18bo3bo$16bobobobo$18bo3bo$14bo3bob2o$16b2o$3bo11b2o$b2o
$obo$2o17bo$15b3obo$15b4o2$8bo$6b2o4b2ob2o$6b2ob3o3b2o$7bob3o3bo5bo$7b
o2bob2o5b2ob2o$8b2o9b2o$24bo$4bo7bo$3b3o5b2o$4bo6bo8bo$20b2o!

This methusala also beats the one that Sowke uncovered recently, meaning that it's the longest one currently known.

[schneelocke]

New record methusala, lifespan 35426 if I'm not mistaken. Congrats to Schneelocke

Thanks!

Since we already have Edna, I propose that this one be named "Fred".

[Sokwe]

Wow, that's over 4000 generations longer than Edna!

This also beats the one that Sokwe uncovered recently

If you are referring to the 35201-generation pattern found by Andrzej Okrasinski in 2008, It was Lewis that noticed it, not me (see this post).

[Nathaniel]

Wow, congrats schneelocke!

Sorry for my absence recently around these parts -- I'll have a look at cleaning up some things related to the Soup Search in about a week's time. The soup search counter is displaying zero because ASP is awful at handling large integers (so basically it doesn't know what to do with a number larger than 2.1 billion).

[calcyman]

The soup search counter is displaying zero because ASP is awful at handling large integers (so basically it doesn't know what to do with a number larger than 2.1 billion).

Shouldn't it display Two's Complement, jump to -2147483648 and start counting up to zero? That's what any normal 32-bit signed register would do.

[ErikDeNeve]

Schneelocke, congratulations on discovering Fred !
The race is on for a 40K+ methuselah ! I may need to build a faster PC now, lol...

[schneelocke]

Schneelocke, congratulations on discovering Fred !
The race is on for a 40K+ methuselah ! I may need to build a faster PC now, lol...

Thanks Nathaniel and Erik!

A 40k+ methuselah would be great - it'll be interesting to see who can find one first, or when it will be found. Maybe I should invest into a faster PC, too (or a second one).

Thanks again everyone for the kind words! I took the liberty of creating an entry for Fred on the LifeWiki, too.

[Sokwe]

The only reasonably sized pattern I found was this 21-cell methuselah that lasts for 28128 generations

I am somewhat disappointed that I didn't notice this two-cell reduction earlier:

Code: Select all

#C 19-cell 28128-generation methuselah
#C based on a pattern found by Nathaniel Johnston (Oct 2, 2009)
x = 23, y = 16, rule = B3/S23
10bo$9bo$10bo$11b2o2$13bo3$b2o$obo18b2o$o20bo$o2$4bo2$3b3o!

Also, this 26701-generation pattern found by Schneelocke might be worth a look, as it has a few small phases:

Code: Select all

x = 20, y = 20, rule = B3/S23
bobobo2b6obobo$5bo2b2ob3o2bobo$bobo2bo3b3o3bo2bo$3bo5bo8bo$b3ob2o3bo2b
o2bo$2b2obobo3b3o3bo$obo10bo$3o5bo4bo2bo2bo$5b2o3b2o2bo$4o2bo4bo3bo$ob
2obo3bo3b4obo$2b2o2b2o2b5obo2bo$2obobob2o2b3o2bo$3bo3b3ob2obo3b2o$bo2b
ob3o5b3o$6bo2bobob2o2bo$obo2bo3bobo4bo2bo$3b3o4b2obobob2o$o8bo2bo3bo$o
bo3b3o8bo!

[schneelocke]

Also, this 26701-generation pattern found by Schneelocke might be worth a look, as it has a few small phases:

Code: Select all

x = 20, y = 20, rule = B3/S23
bobobo2b6obobo$5bo2b2ob3o2bobo$bobo2bo3b3o3bo2bo$3bo5bo8bo$b3ob2o3bo2b
o2bo$2b2obobo3b3o3bo$obo10bo$3o5bo4bo2bo2bo$5b2o3b2o2bo$4o2bo4bo3bo$ob
2obo3bo3b4obo$2b2o2b2o2b5obo2bo$2obobob2o2b3o2bo$3bo3b3ob2obo3b2o$bo2b
ob3o5b3o$6bo2bobob2o2bo$obo2bo3bobo4bo2bo$3b3o4b2obobob2o$o8bo2bo3bo$o
bo3b3o8bo!

This pattern has 28 live cells and a 10x20 bounding box and lasts for 26524 generations:

Code: Select all

x = 10, y = 20, rule = B3/S23
4b2o$4bo9$4bo3bo$b5o3bo$2o3bo$bo2b2o2$7bo$5b2ob2o$9bo$5bo3bo$6b3o!

This has 26 live cells but a slightly larger bounding box of 14x22 and lasts for 26514 generations:

Code: Select all

x = 14, y = 22, rule = B3/S23
3b2o$3bo8$bo$2o5$11b2o$11bobo$5bo7bo$5b2o3bob2o$5b3o2b2o$7b2o$6b2o!

[schneelocke]

One of my soup searchers found a new methusaleh with a lifetime of more than 30,000 generations (30872, to be precise, if I'm not mistaken):

Code: Select all

x = 20, y = 20, rule = B3/S23
3bobobobo3bo3b3o$5b2o4bo2b3o2bo$bo2bobobo4bobo$2bobo2bobob2ob5o$3bobo6bob4o$2obob
o7bob3o$o3bobo9bo$4b2obobo2b2obob2o$o7bobobo3bobo$4obo2bob3o2bo2bo$3bo3bobo3b2o3b
o$2bobo5bo4b3o$b3o2b2obo3b3o2bo$2o5bo3bob3ob3o$4bo2b2ob2o7bo$o2bo2bo2b3o$4b5o2b2o
b3obo$bo2bobo2bo2bo$3b2o3bo4bo$3bob3o3bobob4o!

This one actually emits a HWSS, too, in addition to the usual host of gliders, which I thought was rather neat.

Ed, the son of Fred?

[Keiji]

30,872 generations confirmed by my Golly. I liked the HWSS emission too!

[patma]

I ran the soup search for a few days without problems. Now, after a reboot and starting up again, one or two instances crash some 10-30 minutes after started. I run 4 instances, one for each core.
Why is this? It obviously worked perfectly before, what's up now?

EDIT: Figured out what's happening. For some reason this time it doesn't cut off the gliders making a line properly at 29340 or so like it did before. It continues to somewhere over 100000 and then crashes. Why? Fix please?

EDIT2: This being rule B36/S125, btw. The glider leaving the line really messes with the script.

[Lewis]

When will the tables for each rule next be updated/extended?

I would also like to see a census of the rule B3/s24. It's siilar to 23/35 in that both rules settle fast, but 3/24 has some interesting rare oscillators (p5, p8 etc.) and a p6 c/3 ship.

[Lewis]

I've just realised that the b3/s23 census on this site has counted more objects in total than Achim Flammenkamp's census (at http://wwwhomes.uni-bielefeld.de/achim/ ... _life.html).

I imagine the navbar should say "Home", not "TOLLCASS Home" Other than that, extracted and running okay so far.

It's taken me this long to realise that TOLLCASS actually stands for something ('The Online Life-Like Cellular Automata Soup Search'), I think....

[Extrementhusiast]

Speaking of Achim Flammenkamp, he made a hilarious typo (I've bolded the typo, and commented in brackets):

A total of 5 billion random objects was [should be "were"] censored [instead of censused] to get these data

PS: I am requesting rules B3/S013 and B34/S267, because they both settle down very quickly (within 500 gens).

[HartmutHolzwart]

Speaking of Achim Flammenkamp, he made a hilarious typo (I've bolded the typo, and commented in brackets):

How about your German?