"Hashwar": A Battle Royale for individual GOL configurations

[hash0]

Hi everyone,

I’ve spent some time developing an "Adversarial Conway" framework called Hashwar, and I just launched the first tournament series, hoping to build it into a "computational sports" of sorts.

The goal was to move beyond passive observation and create a competitive environment for Conway’s Game of Life configurations. I’ve developed an adversarial meta-rule set where two glyphs are dropped into a shared "Arena" grid. Unlike standard Life, these patterns compete for territory—invading neutral cells and stealing space directly from their opponent.

The Tournament Format:

• Contenders: 32 unique configurations ("Glyphs") using S23/B3 on the 16x16 torus
  
  Environment: A shared arena of cells.
  
  Fairness: Each match consists of even rounds where contestants swap starting positions to account for any spatial bias.
  
  Scoring: Based on cumulative point totals (territory control) over the final round.

I’m categorizing the contenders by two metrics: Iterations (how long they can roam/evolve) and Peak Count (their highest density/resilience).

I’d love to get your thoughts on the adversarial logic and which types of configurations you think will dominate this specific 16x16 ecosystem (which is a topic of active research on my part - I thought I knew the answer to that when I started but it turns out after a couple of matches, that I don't yet understand what makes certain Glyphs more successful than others).

Check out the Tournament Kickoff here:
https://youtu.be/WTwoMNvMx24

FAQ

■ What are the game pieces?

Called "Conway Glyphs", these are individual GOL configurations on the 16x16 torus following ruleset S23/B3. Additionally, they follow two more requirements.

■ Which additional requirements?

Firstly, we truncate their evolution at the first detected cycle, i.e. we do not allow repeating patterns.

Secondly, we calculate the SHA-256 hash value for a Glyph's initial configuration and subject it to requirements, e.g. they have to begin with a specific substring.

■ What is this hashing requirement about?

Hashing allows to uniquely identify individual Glyphs. Moreover, we use it as a "proof of work" method to level the playing field. Since we want Glyphs to compete against each other, anyone constructing particularly long lived patterns (or skimming them from the Internet) would have an advantage.

Moreover, hashing allows us to determine a Glyph's "intrinsic colour" when we interpret a particular substring of the hash value as hexadecimal RGB codes. This makes recognition of individual Glyphs much easier in a tournament.

Lastly, hashing allows us to effectively and quickly identify loops in the trajectory of a Life.

■ Is this a thinly-veiled crypto-project / scam?!

No, definitely not. Guarding myself against the "Cryptobros" has been close to my heart, so I set out to explain my opinion of them succinctly in this video: https://youtu.be/sl7AuIoElnA

In short, I HODL cyptoscammers in contempt.

■ Can anyone submit "Glyphs" to compete?

Yes! In fact, I would be delighted to find like-minded geeks who fret over finding the "ultimate" tournament champion "Glyph"!

I plan to publish the exact rules on how to generate a valid "Glyph" for a tournament soon. The computations needed require a modicum of technical Knowledge, but can be performed with, say, Excel, Python, Javascript etc.

■ How do we know your "Adversarial Conway" algorithm is fair?

I plan to publish the source code of my custom-built computational engine to GitHub soon and will gladly subject myself to public scrutiny. I'm just one dev (not a very good one) with a passion for CA, so I'm looking forward to the community poking holes in my 'Adversarial Conway' engine once the repo is live.

[hash0]

Here's a challenge to anyone willing to take up the gauntlet:

As I was specifically looking for long-running, non-repeating patterns to feed into my HASHWAR tournament, I came across this configuration:

Code: Select all

x = 16, y = 16, rule = B3/S23:T16,16
4obob2o2b2obo$o3b4ob2o2b2o$3o3bo4b2ob2o$b5obo4b3o$3bob2o2bo4b2o$3bo3b
2obobobo$obobo2b2o3bobo$3obo6bo$2o2bo4bo2bo$3ob5ob3o$o2b2o3b3o$bo3bo2b
o2b2o$2o2bob2o2bobo$b3o5b4obo$o2bo2bo2b6o$2o2bo2b2ob5o!

In binary string format:

1111010110011010100011110110011011100010000110110111110100001110000101100100001100010001101010101010100110001010111010000001000011001000010010001110111110111000100110001110000001000100100110001100101100101000011100000111101010010010011111101100100110111110

This one runs for 1.273 non-repeating steps on the 16x16 torus according to S23/B3 (my "Adversarial Conway" algorithm class stipulates that only non-repeating iterations are taken into account). It's called "Ignis Draconis" for anyone who cares.

Does anyone have a longer running pattern adhering to these specifications?

[hash0]

One thought just occured to me:

The hash requirement is rather strict and essentially precludes anyone from "constructing" a long-running glyph by populating the initial state by hand. Everyone is forced to find a "soup origin" that just happens to adhere to the hash requirement while still giving rise to a long running pattern.

However, translational, reflectional, and rotational symmetries give rise to interesting possibilities: The "Conway Glyphs" my HASHWAR runs on all live exclusively on the 16x16 torus, so for every pattern there exist exactly 2.048 "doppelgänger patterns" with the same trajectory and stats.

Hence, it might indeed be possible to check all 2.048 "doppelgängers" of a constructed pattern whether one of them fullfils a hash requirement. While this is (almost) guaranteed to work for a hash requirement of only 1 or 2 hash characters, it will only work 50% of the time for hash requirements of 3 characters and even less with any additional character.

So, the "path of symmetries" now opens up a distinction in play styles for participants in the contest: The "Architects" (who construct patterns and try to find a variant that adheres to hash requirements) and the "Miners" who look for patterns in the soup.

Here is a "proof by example" - take the "Ignis Draconis" binary string from above and shift it by one row. The resulting pattern evolves for exactly the same number of steps as the original (1.273) and finishes with the same terminal step:
Ignis Draconis "Doppelganger" (1.273 steps, vertical shift)
1100100110111110111101011001101010001111011001101110001000011011011111010000111000010110010000110001000110101010101010011000101011101000000100001100100001001000111011111011100010011000111000000100010010011000110010110010100001110000011110101001001001111110

[hash0]

The discovery of "Ignis Draconis" piqued my interest in the distribution of GOLs population dynamics on the 16x16 Torus under standard B3/S23 rules, similar to Nathaniel's 2009 analysis on the 20x20 torus.

My findings are summarized in the following table where I looked at about 4000 random "soup patterns" and explored their runtimes:

Iterations Count
0 – 99, 1,754
100 – 199 1,195
200 – 299 490
300 – 399 185
400 – 499 65
500 – 599 30
600 – 699 11
700 – 799 1
800 – 899 1

The numbers here suggest a power law distribution.

Over the past three years, I have been hashing and exploring about 250 million such patterns on the 16x16 torus and found only 15 (fifteen) patterns with iterative count of 1,000 or higher (see below list).

This makes the "Ignis Draconis" pattern with 1,273 non-repeating patterns a roughly 1-in-10,000 find.

Point of note: Nathaniel's record holder pattern on the 20x20 runs for 2,169 steps - roughly 5 times the grid size of 400 cells. The same "law" holds true for "Ignis Draconis" which also runs for roughly 5 times the grid size of 256 cells. I am wondering if these two examples are representative of patterns that form the outer edge of the "envelope of longevity".

SUPPLEMENTARY DATA
My list of "Titan" patterns with (non-repeating) iteration count and terminal state:

Ignis Draconis (1.273 steps, 2-Flicker)

1111010110011010100011110110011011100010000110110111110100001110000101100100001100010001101010101010100110001010111010000001000011001000010010001110111110111000100110001110000001000100100110001100101100101000011100000111101010010010011111101100100110111110

4b2obob2o$ob3ob2ob2o$3o4b4o$2b2ob3obob2o$o4bo2b2o$4bob2ob3o$bobo4bo$4b
o2bobo$2b2obobobobo$o2b2obob3o$4bobo4b2o$o4bo2b2o$4b4ob4o$3ob4obob2o$
4b3obobobo$4b6ob2o!

22:53 05.02.2026 Ignis Draconis "Doppelganger (vertical shift)" (1.273 steps, 2-Flicker)
1100100110111110111101011001101010001111011001101110001000011011011111010000111000010110010000110001000110101010101010011000101011101000000100001100100001001000111011111011100010011000111000000100010010011000110010110010100001110000011110101001001001111110

4b6ob2o$4b2obob2o$ob3ob2ob2o$3o4b4o$2b2ob3obob2o$o4bo2b2o$4bob2ob3o$b
obo4bo$4bo2bobo$2b2obobobobo$o2b2obob3o$4bobo4b2o$o4bo2b2o$4b4ob4o$3o
b4obob2o$4b3obobobo!

Lost Monarch (1.226 steps, 2-Flicker)

0110001100111100010110111000100000110011111000100001011000010101111111100101111001110001011111101110011000100011000101011000011011111001110010101111110110000100110011101010011110001100111001110101001100100000000101001010101011000100110001101101101010101000

b2ob2o4b2o$o2b2ob3ob3o$5b2ob2ob3o$4b2obob3o$b2ob2ob2obob3o$o2b3ob2o2b
2o$o2b6ob2o$3ob2ob2o2b3o$b2obobob2ob2o$4b2ob2ob4o$2obob4ob2o$4b2ob2ob
2o$2ob2ob3ob3o$b2ob2ob2ob2o$8bo2bo$2b2o2b2o2b2o!

Copper Solstice (1.221 steps, 2-Flicker)

0110000001100010100010010010100011011101101001101000111001001000100100100100100000010111010100100000110000011110001010111100100011011101111000011000010001111010110001100000001100100010011100000100000001001000001110000111111100011000000011110000000000110010

b2o4b2ob2o$o3bo2bo2bo$b2ob3ob2ob2o$bo3b3ob2o$bo2bo2bo2bo$bo2bo4b2o$ob
o2bo2bo2b2o$5bo4b2o$o2bo4b4o$bo2b4ob2o$o2b2ob4o$b2o4b2ob2o$b2o7b2o$o
b2o2bo2b3o$5b4ob4o$o2b2ob2o4b2o!

Twilight Entropy (1.184 steps, 2-Flicker)

1100110010011000101001100100010010101000110101001101001000000001000010011001100000101011111111011000100000111100111000101110110111111000010100011010111110111101001011001000101010101100101111010110110000011111111001000110110010010110000100000010011101110000

2b2o2b2ob2o$bo2b2ob2o2bo$bo2bo2bo2bo$bo2bobo2bobo$bo3bobobobo$6bo2b2o
$8bo$b2o2b2o4b2o$8bo2b2o$ob4ob4o$2bo5b4o$o2b2o2b3ob2o$ob6o2b2o$o3bo3b
ob4o$b2ob2o2b2ob2o$obobo2b2ob2o!

[ROBIN] Teal Titan (1.177 steps, Static)

1110101111010011111010001110000101100001111101000111101101111101010111011101111011101001101010100001010000010100001011101110000110011010110011110000011111001010101001011000001010111111011110011000111010100011110010011111001100100010110110110000001110100111

3bobob4ob2o$3ob3o4b2o$b2ob4ob2o$b2ob4ob2o$2b2obobob5o$ob4ob4o$b3ob3ob
2o$3o2bobobob2o$5bobo2bo$4bob2ob4o$b2ob2ob2ob4o$5b6ob2o$2b2o4b2ob2o$o
b3obob2ob2o$b2ob4ob2o$b2ob2obob2ob2o!

El Viejo del Norte (1.164 steps, 2-Flicker)

1001010101010001101001111101111010111000000100101000000100110111000010110100110111100011100010110101011111010001010000001101010100101010000011010011011000110011111000110000010110101010110011111111100110011100000101010010110001100101011100111100100111101111

o2bobobobob3o$bo2bo5b4o$b2ob3o5b2o$bo4bobob2ob3o$4bobo2b2ob4o$o3b4ob
3o$4b3obob2ob3o$o3bo5b5o$4bobo2bobo$4bo5b2o$o2bo2bo2bo2bo$bo3bob2ob2o
$2b2ob2ob4o$3b2o5bo$b2obobobob2o$b4ob2ob2ob2o!

Viridian Mist (1.161 steps, Static)

1001110000000000011001001100001101000110001001111001100011100001100110011000000110011000100010010000110000100010100000100110100000000000111100000100111100110000000110100001101000010010001000011100101010110011010000000000100000010110111010011000000111100000

o2b4o$2b2ob2o4b2o$b2o3bo2b2ob3o$4b2ob2o2b3o$b2ob2ob2ob2o$b2ob2o4b2o$b
2ob2ob2ob2o$4b2o2bo2bo$4bobo2bo$2b2o10bo$4b4o5bo$b3ob2ob2o$6bo4bo$bo
4bo4bo$bo4b3o2bo$bob2ob2ob2o!

Sterling Moss (1.107, 2-Flicker)

0011000000011000010110100001100000001000001000010000000000001101010000000101011001100101101001010000100000000010001000100110100000010010000010000010011011000100010100000000110000000010100100001000011000000100010101000100000010001110001000000101000101011100

2b2o7b2o$bo2b2ob2o3b2o$8bo$b2o5bo$12bobo$bo8bo$b2ob2ob2ob2ob2o$4bo11b
o$4bo2bo2b2o$6bo5bo$4bo4bo2bo$b2ob2o4b2o$8bo$b2o4bo2bo$bo4bo2b2o$bo4b
3o2bo!

Proxima Jade (1.057, 2-Flicker)

0101000010001100011001100000000001000000010010111011000000101011101000100000101000001010001010101000001010001000000000000000000000010100111001011011010000011000110001010000000010000000010101100010010000100010000000111100000000100000000011010100101000110000

bo2bo4bo2b2o$2b2ob2o$8bo$8bo2bob3o$b2o8bo$2b2ob4ob2o$8bo2b2o$bobo2bob
obobo$8bo2b2o$16b$16b$5bo2b3ob2o$b2ob2ob2o4bo$b2ob2o5bo$10b2o$b2o4bo
2bo2bo!

Viridian Wake (1.047, 2-Flicker)

0011010011111001111000010111101001001011011010100011000001101111111111000101110010010111111000101111111101110110001110011101000001011010000011111011001010111110001001110100111010000100111111111110110111100010000111110101110010110011110111011001010100100111

2b2obo4b2ob2o$4b4ob2ob2o$bo2b2ob2ob2obo$4b2o2b2o2bobo$4b2o2b2ob2ob2o$
b7ob2ob2o$4b2ob2o5b2o$b7ob4ob2o$b2ob2ob2ob2o$5bo2b2o5bo$b4ob2ob5o$b2o
b2ob2ob2o$bo2b4ob7o$4b4ob2o4b2o$b7ob2ob2o$b2ob2ob2ob2o!

Verdant Fossil (1.030, Static)

1000000000000000000001101010001000000011110000101110000100001011010010110010000010010100101100001000000000000000000110000000100001000000000000000010000100111010000000001010000001000101010001010110000001100000100100101010100010001000000001010000010100011100

o$6b2obobo2bo$8b4o2bo$b3o4bobo$o2b2ob2o2bo2b2o$bo2b2o2b2ob2o$16b$3b2o
8bo$4bo11bo$8bo4bo$2b3ob2o10b$2bobo5bo$b2obob2ob2o$4b2o5b2o$b2o2b2ob
2ob2o$8bo!

Cobalt Nexus (1.023, Static)

1100100110000011000100000001010100100001101000110010101101010001001101111010000101000000000100000100011110000011111100101010001010001101001100101001110100000100000010100111000011010101111110100011010101100001000010001101111110111000000000010001101100010011

2b2ob2ob2o3b2o$4bo7bobo$2bo4b2o3b2o$2b2ob2ob2ob2o$2b2ob4ob2o$4bo8b2o$
4bo7b4o$6b2obob2o$2b2o2b2o2b2o$2bo2b2ob2ob4o$8bo$2bo4b2ob4o$b2obob2ob
2ob2o$o2bo4bo2bo$4bo2b4ob2o$b2ob2ob2o!

Bluegrass Kelso (1.013, Static)

0010100100100010000110010000001100110101000000100000000100010111000000000000100000000000000000000010101000000100010010111011000100100010110001000001100010001010010001000010000000000000100000100000011000000000100001011000101100100010010000010111000001000000

2bobo2bo4bo$4b2ob2o5b2o$2b2ob2o$6b2o$4bobo2b2o$16b$16b$2bobobobobo$4b
o2b2ob3ob2o$4b2o2b2ob2o$5b2ob2o$4bo2b2ob2o$8bo$4b2ob2o$2b2ob2ob2ob2o$
2b2ob2ob2ob2o!

Sulfur Sunday Silence (1.009)

1001011101101101110011001001001001001010011101011000011101111000011011111010001100000100111001100101000100010001111000101011110011000010111101000000010111000111101010100111000101011101011010001001001110110000111100011110111010111110101001000000001011010011

o2bob3ob2ob2o$3b2ob2ob2ob2o$2b2ob2ob2ob2o$2b2ob2ob2ob2o$4b2obob2ob2o$
4b2ob4ob2o$4b2ob2ob2ob2o$2b2o2b2ob4o$2b2ob2ob4ob2o$2b2ob2ob2ob2o$4b2o
b2ob3ob2o$b2ob2ob2ob2o$4b2ob2ob4o$4b4ob2ob4o$b2ob4ob2o$4b2ob4o!

SHAhrazad-256 (1.001)

1000101000001000000000000010010100100001100000001001000010010000100000011101010001000000000110010110100100001101011010000000001001000101010000100000001010001000001000110000000010000011010011010100000011010010001011010010000010100110010101000100010010000000

o3bo5bo$6b2o2bobo$2bo4bo8bo$2bo4bo4bo$bo7b3obobo$4bo8b2o$b2ob2ob2o2b
2o$2b2o2b2o2bo$8bo2bo$2bo4bo4bo$b2o6b2o$b2ob2o4b2o$b2o2b2ob2o$2b2o2b
2ob2o$2b2o2b2ob2o$8bo!

[I6_I6]

Here's a challenge to anyone willing to take up the gauntlet:

As I was specifically looking for long-running, non-repeating patterns to feed into my HASHWAR tournament, I came across this configuration:

x = 16, y = 16, rule = S23/B3
4o1bo1bo2b2o1bo1bo$3b4o1bo2b2o1bo2b$3o3bo4b2o1bo2bo$5o1bo4b3o4b$bo2b2o4b2o3bo3b$2bo3b7o2bo$2b3o1bo1bo3b6o$bo3bo3bo1bo4bo$2b2o4bo2bo3b3o4b$6bo4b2o4bo2b$2bo3b3o5o1bo3bo$2bo2b3o5b4o6b$bo3bo2bo2b3o3b2o1b$2b2o1bo2b2o1bo4b3o$4b6o1bo1bo2bo$2bo2b2o6o1bo2b2o!

In binary string format:

1111010110011010100011110110011011100010000110110111110100001110000101100100001100010001101010101010100110001010111010000001000011001000010010001110111110111000100110001110000001000100100110001100101100101000011100000111101010010010011111101100100110111110

This one runs for 1.273 non-repeating steps on the 16x16 torus according to S23/B3 (my "Adversarial Conway" algorithm class stipulates that only non-repeating iterations are taken into account). It's called "Ignis Draconis" for anyone who cares.

Does anyone have a longer running pattern adhering to these specifications?

Welcome to the forums!
If you put RLE in pattern or code tags, other people can run the pattern themselves in LifeViewer:

Code: Select all

x = 16, y = 16, rule = S23/B3
4o1bo1bo2b2o1bo1bo$3b4o1bo2b2o1bo2b$3o3bo4b2o1bo2bo$5o1bo4b3o4b$bo2b2o4b2o3bo3b$2bo3b7o2bo$2b3o1bo1bo3b6o$bo3bo3bo1bo4bo$2b2o4bo2bo3b3o4b$6bo4b2o4bo2b$2bo3b3o5o1bo3bo$2bo2b3o5b4o6b$bo3bo2bo2b3o3b2o1b$2b2o1bo2b2o1bo4b3o$4b6o1bo1bo2bo$2bo2b2o6o1bo2b2o!

You can use code or pattern tags by clicking the buttons above the editor that say '</>' or 'pattern'.
Unless the RLE itself was wrong, both LifeViewer and Golly interpret it as a 20x16 pattern, not 16x16.
Also note that rulestrings are usually represented with birth before survival, so it would be B3/S23.

According to your binary string (assuming new row every 16 digits), the pattern is this:

Code: Select all

x = 16, y = 16, rule = B3/S23:T16,16
4obob2o2b2obo$o3b4ob2o2b2o$3o3bo4b2ob2o$b5obo4b3o$3bob2o2bo4b2o$3bo3b
2obobobo$obobo2b2o3bobo$3obo6bo$2o2bo4bo2bo$3ob5ob3o$o2b2o3b3o$bo3bo2b
o2b2o$2o2bob2o2bobo$b3o5b4obo$o2bo2bo2b6o$2o2bo2b2ob5o!

Which seems to be correct. The :T16,16 at the end of the rulestring means that the pattern will be simulated in a 16x16 torus universe.
It settles into an oscillating pattern in 1271 generations, but it seems like you included a full extra cycle of the blinker (2 extra gens). Conventionally, you'd define the number of generations for a soup to stabilize as the generation when it starts its first oscillating phase.

This whole 'Adversarial Conway' thing is pretty interesting, but I still don't understand exactly how the scoring works. I get a rough idea of what it's like and I've watched your videos.
Also, how does the hashing work? These two little example patterns evolve into the same exact same thing in 3 generations. would they be counted differently?

Code: Select all

x = 15, y = 5, rule = B3/S23
o2bo9bo$b2o9b2o$o10bo2bo$bo10b2o$bo!
[[ AUTOSTART GPS 4 STOP 3 ]]

Or this variant of Ignis Draconis that differs by 1 cell?

Code: Select all

x = 16, y = 16, rule = B3/S23:T16,16
4obob2o2b2obo$o3b4ob2o2b2o$3o3bo4b2ob2o$b5obo4b3o$3bob2o2bo4b2o$3bo3b
2obobobo$obobo2b2o3bobo$3obo6bo$2o2bo4bo2bo$3ob5ob3o$o2b2o3b3o$bo3bo2b
o2b2o$2o2bob2o2bobo$b3o5b4obo$o2bo2bo2b2ob3o$2o2bo2b2ob5o!

Would such variants also count as 'doppelgängers'?

[hash0]

Welcome to the forums!

Thank you so much for the warm welcome - also for the very useful tipps & tricks regarding RLE in pattern tags. I'll heed your advice henceforth.

Also note that rulestrings are usually represented with birth before survival, so it would be B3/S23.

This is PRECISELY why I finally summoned the courage to join this forum instead of just lurking - to learn the conventions of the trade!

Unless the RLE itself was wrong, both LifeViewer and Golly interpret it as a 20x16 pattern, not 16x16.

Serves me right for trusting an AI with the conversion... *grumbles*. That's on me, I apologize for this gaffee. This whole RLE shortform was entirely new to me - I love it but since I am almost entirely devoted to the 16x16 torus pocket universe, it has so far not had much utility for me over the binary character string. The more you know.

It settles into an oscillating pattern in 1271 generations, but it seems like you included a full extra cycle of the blinker (2 extra gens). Conventionally, you'd define the number of generations for a soup to stabilize as the generation when it starts its first oscillating phase.

Huh, that's actually an index convention and an off-by-one error piled on top: My convention starts with iteration '0' (to give it that special 'origin' flavour - also: I wrote my own computational engines in Javascript, so that bled into my index choice) and, you are right, the algorithm counts the first repeating iterations towards the iteration count. I shall fix that.

This whole 'Adversarial Conway' thing is pretty interesting, but I still don't understand exactly how the scoring works. I get a rough idea of what it's like and I've watched your videos.

I shall produce a video that explains this in detail. Think of the 16x16 torus as a "rubber stamp", albeit one that changes its "face" with every iteration. The rubber stamp is applied to the shared canvas (the "Arena") - every cell on the shared canvas the rubber stamp encounters with a living cell has its value incremented by .1 up to a maximum of 1.0 (for the opposing torus it's -.1 and -1.0). The rubber stamp evolves and moves around the canvas in a random direction by one cell for every iteration.
Any shared canvas cell that has a value of more than |.5| is counted as 1 point towards its "owner" until the match is over. In "combative" mode, the algorithm allows the Glyps to "steal"/"conquer" cells from their opponent (a cell with value .2 can be turned to the other side if it is hit at least three times by the opponent, turning the value to -.1).
This also explains why a Glyph that has already reached a halting state can still earn points: The algorithm still counts the cells with |.5| value or higher towards its score for as long as at least one Glyph is still running. A Glyph that has reached a halting state just can't claim any new cells or conquer them from the opponent.

Also, how does the hashing work? These two little example patterns evolve into the same exact same thing in 3 generations. would they be counted differently?

Code: Select all

x = 15, y = 7, rule = B3/S23
bo2$o2bo9bo$b2o9b2o$o10bo2bo$bo10b2o$bo!
[[ AUTOSTART GPS 4 STOP 3 ]]

Hashing as proof of work is only relevant for the hash of the original configuration. Under this paradigm, it does not matter if two differing configurations eventually converge in their trajectory.

Or this variant of Ignis Draconis that differs by 1 cell?

Code: Select all

x = 16, y = 16, rule = B3/S23:T16,16
4obob2o2b2obo$o3b4ob2o2b2o$3o3bo4b2ob2o$b5obo4b3o$3bob2o2bo4b2o$3bo3b
2obobobo$obobo2b2o3bobo$3obo6bo$2o2bo4bo2bo$3ob5ob3o$o2b2o3b3o$bo3bo2b
o2b2o$2o2bob2o2bobo$b3o5b4obo$o2bo2bo2b2ob3o$2o2bo2b2ob5o!

Would such variants also count as 'doppelgängers'?

That, good Sir or Madam, is a very clever exploit! Short answer: No, under my own definition, this does not constitute a "doppelgänger" (which should have identical origin configuration). This leaves me sleepless, though, because I plan to implement a central register to let tournament participants truly "own" their Glyph. How can I prevent anyone from tweaking a prized, winning Glyph and submitting it in the tournament? The answer to this still eludes me... However, the central register will contain the timestamps when a pattern was first discovered, so any "derivative" pattern could be traced to its original and banned from the tournament, if necessary.

[unname4798]

Manual LifeViewer soup search:
≈500 gens only:

Code: Select all

x = 16, y = 16, rule = B3/S23:T16,16
6obo3b5o$o2b5o2b2obo$bobob4obob4o$o2b3o2b2ob2ob2o$6obob2ob3o$o2b3o2b2o
b3o$2o4b2ob3o3bo$4bo3b3obo2bo$3b7ob2obo$7bobo3b3o$2b2obob6obo$2o2bobo
bo4b3o$obo3bobo$2b2ob7ob2o$5bob2o2bobo$bob3o2b3ob2o!

[hash0]

Respect!

Let me demonstrate the computational efforts involved here. For my ongoing tournament "HASHWAR{512}" the contestants have to fullfil two requirements:

• Hash Requirement: The SHA-256 hash value of the binary character string of the origin configuration must start with '512'.

• Iteration Requirement: The pattern must run for at least 512 non-repeating steps when played on the 16x16 torus according to B3/S23 rules.

It took me 15,127,329 hash attempts to produce 3,744 candidate patterns which meet the hash requirement, of which only 37 meet the iteration requirement. I produce the roster with 8 "seeker" and "explorer" nodes (web-based Javascript engines, nothing fancy, not even GPU-optimized coding) which takes me about one entire business day to complete.

[unname4798]

How many hashes meet the iteration requirement, but not the hash requirement?
Also, you can use SHA-512 for a 16x32 torus.

Code: Select all

x = 16, y = 32, rule = B3/S23:T16,32
o2bo4bo2b3o$b2obob2obobo3bo$2bo3bobob3o$4o2bo2b2o3bo$2obo3b2obo4bo$ob
o2bob2o3b2obo$3o6b3obobo$4o2bo4bobo$3o2b3ob5o$2o4b5o3b2o$2ob4ob2o2b2o
bo$2obo2bo3bo3b2o$bob2ob2ob2ob4o$o8bo5bo$2obob2ob2o$6bobo3bo2bo$o5b2o
2bo2bo$bo3bobob6o$2o3bobob5o$b3o4b2obo2bo$o6bob4o2bo$2o4bobo5bo$3o2b2o
b2o2bobo$ob3o2b2o2bob2o$3bo3b3ob2obo$6b3o5b2o$b3obob4o3bo$6b3ob6o$3o5b
4ob3o$2bo2b4obo2b2o$bo2bobo3b2o2b2o$2obob3ob2ob3o!

Also ≈500 gens.

[hash0]

Good question!

I always hash first (what I call "seeking"), then explore the candidates that meet the hash req for iterative count.

The reason for this is that hashing is the much faster computation (at least for my seeker nodes) while exploring requires computing every individual step of the iteration (I am not using "HashLife", for example, despite my name).

HOWEVER, my histogram for iterative count frequencies hold true, independent from the hash requirement (because the SHA-256 algo as a cryptographic hash function is guaranteed to produce random hash strings, so whatever subset of GOLs I am looking at that fullfills a given hash req will not be special in any way from all possible GOLs):

Iterations Count
0 – 99, 1,754
100 – 199 1,195
200 – 299 490
300 – 399 185
400 – 499 65
500 – 599 30
600 – 699 11
700 – 799 1
800 – 899 1

[I6_I6]

...

• Hash Requirement: The SHA-256 hash value of the binary character string of the origin configuration must start with '512'.

...

Ignis Draconis doesn't meet that requirement. This is the SHA-256 hash value:

Code: Select all

099fc6f5d4ef906c06346c8d6feebc0c0d63c1cb37988d82f7c7756749f9f869

I just used this online encoder.

[hash0]

Correct!

"Ignis Draconis" was found as part of "Series 099", an initial feasibility study in 2024, if memory serves. It is not part of the contestants for "Hashwar{512}" (those can be found here: https://www.youtube.com/watch?v=WTwoMNvMx24&t=65s ).

And yes, it's hash value is "099fc6f5d4ef906c06346c8d6feebc0c0d63c1cb37988d82f7c7756749f9f869". Fun fact: While the first three characters of the hash determine the series it belongs to (099), the next six characters set it's intrinsic colour, here #fc6f5d which sits between bright orange and reddish pink, giving the pattern its name ("Ignis Draconis" is latin for "Fire Dragon").

[hash0]

Manual LifeViewer soup search:
≈500 gens only:

Code: Select all

x = 16, y = 16, rule = B3/S23:T16,16
6obo3b5o$o2b5o2b2obo$bobob4obob4o$o2b3o2b2ob2ob2o$6obob2ob3o$o2b3o2b2o
b3o$2o4b2ob3o3bo$4bo3b3obo2bo$3b7ob2obo$7bobo3b3o$2b2obob6obo$2o2bobo
bo4b3o$obo3bobo$2b2ob7ob2o$5bob2o2bobo$bob3o2b3ob2o!

Hey, wanna battle? Let's do a friendly match, so no requirement for our Glyphs to have the same hash class (like "512"). Give your pattern a funny/fancy name and hop over here:

https://lifehashes.net/Y175/
This is my (prototype-stage) battle engine. You input the two binary strings on the left and right side respectively. If you give the match itself a name (bottom middle input field), like "TESTMATCH123" you can look at the results here:
https://lifehashes.net/Y175/results.php ... STMATCH123

Here is my pattern, called "Admiral Piett Mondrian", running for 533 iterations):
0010010101001101000010001011110101101001110101100111101000011010010101011101110111010010101101011100010011101110100001111010010110000010011101100110100011001001010010001001101000010001100100111111000101010011010001011101110011001111010100111001010100000000

[unname4798]

559 gens:

Code: Select all

x = 16, y = 16, rule = B3/S23:T16,16
3obo2bo2b2o2bo$5bo2b2obo3bo$2bo2bo3b2o4bo$o2b3obo2b6o$obo2bobob3obobo
$2bob2o2bobo2b2o$2bobobobobo2b3o$2bob2o2b5o$3b3o2b7o$2bo2bob2o$o4b2o4b
2obo$o3b5obo2bo$4o4bob3ob2o$o4b3ob2ob2obo$5bobobo3bo$bo3bobo2b2obobo!

This is "Toroidal Pie".
Add RLE support, please.

[hash0]

Alright, here is the binary for "Toroidal Pie" (I will implement an RLE parser later) for you to double-check:
1110100100110010000001001101000100100100011000011001110100111111101001010111010100101100101001100010101010100111001011001111100000011100111111100010010110000000100001100001101010001111101001001111000010111011100001110110110100000101010001000100010100110101

The results for the match I played against Admiral Piett Mondrian are here:
https://lifehashes.net/Y175/results.php ... M260208-02

The Admiral won 8 out of 8 rounds with a total score of 2,382,699 against the Pie with 1,850,218.

This is something we keep seeing in our tournament, too: The Glyph with the lower iterative count (the Admiral with 533 steps) dominates. When I started this entire project, I had assumed that it's the longer running Glyph that dominates simply because it has more time to move around and conquer cells. This is certainly true when the iterative counts of the two competing Glyphs have a high difference between them (a Glyph with 800 steps will always outperform one with 400 steps), but exactly how high this difference has to be to guarantee a win is anyone's guess at the moment.

My current hypothesis is that it's the Glyphs with high density in the earlier iterations who dominate the duel simply because they will quickly conquer territory that they hold on for how long the match lasts which cumulatively let's them win over longer-running but lower-density Glyphs.

[unname4798]

652 gens:

Code: Select all

x = 16, y = 16, rule = B3/S23:T16,16
2bo3b10o$o3b2obobobobobo$bob3o2bo2bo$bob3ob2o2bo$2bo3b3o4bo$4bobo2b2o
2bobo$3b2obobob5o$bo4bob4obobo$5o4b4obo$2o2bobobo2b2o2bo$3obobob3ob3o
$2b2obobo4bobo$2obobobobobo$3o4b3o3b3o$o2bob3obobobobo$3b2o2b4o2b2o!

Blinker Haven (the ash is only blinkers)

[hash0]

That's the spirit! You, Sir, are hereby invited to my next Hashwar tournament as an honoured guest!

Here is "Blinker Haven" in binary:
0010001111111111100011010101010101011100100100000101110110010000001000111000010000001010011001010001101010111110010000101111010111111000011110101100101010011001111010101110111000110101000010101101010101010000111000011100011110010111010101010001100111100110

And the results for the 8-rounds match against "Admiral Piett Mondrian":
https://lifehashes.net/Y175/results.php ... M260208-03

WOW! That's an UPSET WIN - the Admiral won only 3 out of 8 rounds but has the higher score of 3,198,907 vs 2,962,185 for Blinker Haven.

Seems like the delta in iterative count between 533 and 652 is not yet big enough - but this is a well known pattern, e.g. here:
https://www.youtube.com/watch?v=vAYfaNe1qY0
In this video, "Electric Company Blues" (559 iterations) utterly dominates "Whitewing Variation" (679 iterations).

[hash0]

Any one interested in doing a "friendly" tournament?

No hash requirement, you just submit your GOL patterns that live for, say, at least 400 generations (B3/S23 ruleset on the 16x16 torus, no repeat patterns allowed) and then we let them compete on my "Adversarial Conway" arena, round-robin style.

Analysing the match results will further my goal to understand what makes these "Conway Glyphs" perform (is it longevity? is it density? is it a combination of both?). If you want to, I could even record the matches as they unfold and upload them to my yt channel.

[hash0]

Hi everyone,

Just a quick update on the HASHWAR{512} tournament progress. I’ve decided to accelerate the release schedule for the final matches of the first stage.
This will showcase how these matches are played according to the "Adversarial Conway" meta-rule set and also build some intuition as to what qualities and stats make a pattern a winner:

Group F (2 machetes): Live now.
Group G (2 matches): Dropping today at 20:00 CET.
Group H (2 matches): Dropping tomorrow at 04:00 CET.

By tomorrow morning, the entire 32-match stage One will be complete and available for binge-watching. If you want to catch up on the standings or see the upsets so far, you can find the full archive here:

https://www.youtube.com/playlist?list=P ... bqreoZQSUq

The Knockout Stage begins shortly after—thanks for following along!

[hash0]

Here is the current tournament bracket after Stage 1 is completed:

https://lifehashes.net/Y175/bracket.php

Stage 2 will determine the group winners, after which we'll see the Quarter Finals in Stage 3.

Anyone want to hazard a guess as to which Conway Glyph will win the tournament?

[unname4798]

I tried to guess that p8 oscillator in Joway's Gift of Lives, but b-engine guessed it correctly and betrayed me. So, no, I won't take the risk.

[b-engine]

I think you should read LifeWiki to know about the terminologies.

[hash0]

I tried to guess that p8 oscillator in Joway's Gift of Lives, but b-engine guessed it correctly and betrayed me. So, no, I won't take the risk.

I'm not following you... care to explain?

[unname4798]

oh, i am sorry. that is a period 8 oscillator. i hope not too much ang enters your mind once youve found this out.

its actually a period 8 oscillator called "stupid griefer who destroys my things to gain fake luv". (i told a griefer that i would name the next failed 12 cell still life attempt after them)

True solution of the 12 cell SL problem::

Code: Select all

x = 6, y = 6, rule = B3/S23
4bob$2bob2o$bo$4bo$2obo$bo!

CONGRATS! jonway owes you 10 coffees. i hope your mind gets filled with luv, and if it doesnt, go touch a catcus (that isnt a threat. if you think it is you need to go read more coolpedia)

[hash0]

Ah, nice! I saw you guys instigate some matches on my battle engine as logged in my database.

I promise to add RLE support.