The Online Life-Like CA Soup Search

[Macbi]

I'm not sure if cell-count is a good way of measuring methuselahs. After all, an 8-cell pattern with a glider and a blinker can last quite a while. Bounding box seems more sensible a more sensible measurement of size.

[Andrew]

Has anyone plotted initial cell count versus maximum lifespan for methuselahs?

Not quite what you want, but this entry in Nathaniel's blog is interesting:
http://www.nathanieljohnston.com/index. ... e-of-life/

[knightlife]

an 8-cell pattern with a glider and a blinker can last quite a while

Clearly the bounding box has to enter the equation along with cell count and lifespan (admittedly a tough if not impossible equation to come up with). I suppose there will always be trivial cases no matter what methuselah definition you come up with. Notice the final glider collision in the latest 20x20 soup with lifespan 28158 takes quite a while to happen, but that time is added to the lifespan and I think that is perfectly acceptable.

[Eylrid]

I don't know if anyone has mentioned adding HexLife (or whatever it ends up getting called). One of the difficulties with that one is that rotations are different.

[Macbi]

New methuselah; 28987, so close, and yet so far.

[Lewis]

I've had an idea - Could the next version of the search script be somehow combined with the script that cleans up the data on the site?

[Sokwe]

For some reason the soup search results for long-lived patterns does not include anything found by me. I first noticed this a while back when it failed to show any of my long-lived patterns in 2x2, despite the fact that I had made the greatest contribution to that rule at the time. Now with the dialogue on the side that tells me the lifespan of long-lived that it has found, it has become more obvious that it is failing to put these patterns properly in the database. For example, my current search in B35/S23 says that it has found a long-lived pattern of a lifespan between 1100 and 1149 generations. I know my census data has been uploaded at least twice since this was found, but it does not appear on the list of long-lived patterns. This is only for long-lived patterns, my normal census data is obviously making it to the database.

[iconmaster]

For some reason the soup search results for long-lived patterns does not include anything found by me...

This has happened to me too.

[Nathaniel]

For some reason the soup search results for long-lived patterns does not include anything found by me.

This is very strange... I haven't been able to find the source of the problem, and I can't imagine why it would be able to upload one type of data but not the other. Could you check your Golly data directory? On my computer it's located at:

C:\Users\Nathaniel\appdata\Roaming\Golly

Even if it was unable to upload your methuselahs, there should be files in there called "methuselah_XXXX.rle", where XXXX is the lifespan of that methuselah. Are those files in there? If so, check to make sure that nothing particularly notable has been missed (eg. a lifespan 30000 pattern) while I try to figure out what's going on.

[Axaj]

Ugh... whenever I try to get it to run, I get a error saying "no module named soup_search_exe". I have Python 2.6.3, Golly 2.0, and Mac OS 10.8, if that means anything.

[Eylrid]

an 8-cell pattern with a glider and a blinker can last quite a while

Clearly the bounding box has to enter the equation along with cell count and lifespan (admittedly a tough if not impossible equation to come up with). I suppose there will always be trivial cases no matter what methuselah definition you come up with. Notice the final glider collision in the latest 20x20 soup with lifespan 28158 takes quite a while to happen, but that time is added to the lifespan and I think that is perfectly acceptable.

Well, I suppose you could start with a simple equation: lifespan/(cellcount*boundingboxarea), and complicate it as needed.

[The Big H]

Possibly a silly question, but how come the two naturally occurring switch-engine variants don't show up on the spaceships & puffers list?

[Nathaniel]

Possibly a silly question, but how come the two naturally occurring switch-engine variants don't show up on the spaceships & puffers list?

Due to me being way behind in my cleaning of the data. They're in the database, just not in the publicly-displayed (ie. clean) part of it yet.

[Sokwe]

Even if it was unable to upload your methuselahs, there should be files in there called "methuselah_XXXX.rle", where XXXX is the lifespan of that methuselah. Are those files in there?

Yep, they're there. I sorted out the patterns that would make the top 50 if they were added. I had twelve for B36/S125 (2x2) and one for B35/S23.

[Sokwe]

Even if it was unable to upload your methuselahs, there should be files in there called "methuselah_XXXX.rle", where XXXX is the lifespan of that methuselah. Are those files in there?

Yep, they're there. I sorted out the patterns that would make the top 50 if they were added. I had twelve for B36/S125 (2x2) and one for B35/S23.

Edit: I'm pretty sure it was because I never put down an email address. I have tried it since with an email address and it has successfully uploaded a methuselah.

[Lewis]

Are the lists on the site still updated? No new patterns have been added at the bottom of the lists.

[Sokwe]

Are the lists on the site still updated?

Yes, they are; I have been checking the statistics semi-regularly just out of curiosity, and they are still being updated (the rate of new methuselah discoveries has greatly decreased, however). Amazingly, over 1,000,000,000 soups have been censused in Conway's Game of Life (1,088,744,565 currently)

No new patterns have been added at the bottom of the lists.

I believe Nathaniel has been too busy to add new patterns to the lists (a task that I believe he does manually).

[Sokwe]

We have a new record holding methuselah at 31192 generations!

Here's a 26-cell pattern that stabilizes after 31082 generations:

Code: Select all

x = 43, y = 26, rule = B3/S23
2o$o$o11bo$12b3o14b3o5$3b2o$3bo12$29bo11bo$29b3o10bo$42bo$25bo15b2o$
24b2o!

[Nathaniel]

Wow, congrats Erik!

[Macbi]

Woo! Record! Nice work!

P.S. Erik should name it, yes?

[ErikDeNeve]

>> We have a new record holding methuselah at 31192 generations! <<

Yesss!! I didn't even notice it until today !
Our computer room was very cold these past weeks, which gave me the perfect excuse to run multiple PCs with soup searches to help warm it up.

>> P.S. Erik should name it, yes?

I'll have to think about it

[ErikDeNeve]

So - let's officially name it Edna. Nathaniel Johnston suggested it, being the name of Methuselah's (apocryphal) wife and also ending in -a, and I also like to co-name it after Dame Edna Everage, the famous megastar-and-housewife ( http://www.dame-edna.com/ )


It will be interesting to see if anyone can find ancestor patterns, to drive up the generation count by a few, like what happened to Lidka...

[ErikDeNeve]

I added an entry for Edna to the life wiki: http://www.conwaylife.com/wiki/index.php?title=Edna
I think the resulting pattern looks a little like a sitting mouse...
Also, another Lidka-beating pattern was discovered yesterday ! At 29,947 generations it's running a little behind Edna; I'm calling it Madge.

[Richard]

If we search soups of 3 different sizes, say, 20x20, 20x40, and 40x40, the results could be combined to evaluate separate production rates for objects per unit area, per unit length of perimeter, and per corner. When we have large enough samples for good statistics, we can see whether spaceship production depends almost entirely on perimeter and corners, while the production rates of still lifes and oscillators per area should be a good match to those obtained by Achim Flammenkamp for toroidal soups. I suppose we might find negative rates of production per corner or per perimeter for objects that are only likely to be created deep in the interior of a soup.

This assumes that the rate of creation of each object is a linear function of area, perimeter, and number of corners. One nonlinear effect I would expect to see is that mutual collisions of gliders deplete the net production of gliders, to a degree roughly proportional to the sum of the squares of the lengths of the sides of the soup, along with a corresponding enhancement of the production of objects created by right-angle glider collisions. My choice would be to model net glider production for a side of length "s" as (a(s+b))/(1+c(s+b)^2) where "a" represents the gross rate of glider production per unit of perimeter, "b" represents the correction for corners, and "c" represents depletion by mutual collision; the depletion term would have to be replaced by an exponential function for "s" very large. To evaluate this effect would require a 4th soup size or shape, and additional soup geometries would help confirm this functional relationship and rule out other important nonlinear effects.

Statistics for the lifetimes of soups of different sizes and shapes would also be interesting.

For Conway's Life, at least, I think this would be a better use of our computer cycles than running additional 20x20 soups. A simple way to start this process would be to distribute additional versions of the script, one for each new soup geometry to be run, provided this doesn't compromise the tamper-proof quality of the script; and I promise to be patient while you figure out how to present all the additional results!

[Veronika]

Just to point out, the highscore for longlived patterns in 2x2 are messed up