├ method

This is the whole method. Every constant is listed, every step is written out, and nothing is left for anyone to guess at.

Every site, by its own random field and the field it would need alone. The filled points are bursts that actually happened, and they happen lower than this picture says, because sites help each other.

Bursts by size class. A run that is almost all ones is a wire below critical. A run with a single enormous entry is a wire above it.

The wire is 16 by 16 by 16 sites, 4,096 in total, each either down or up. Each site is given a random field once, at genesis, drawn from a Gaussian of width 2.16 by Box-Muller from a generator seeded with 0x52454D00, in fixed site order.

When a site flips it changes what its neighbours feel, which can flip them, which can flip theirs. That cascade is one burst. The size of a burst is how many sites flipped before it stopped.

The external field only ever rises. When the market cap gives a field lower than the highest already seen, nothing is done at all.

local = H + J * (up_neighbours - down_neighbours) + h[i]

if local > 0 and s[i] is down: flip it, push its six neighbours onto the queue

repeat until the queue is empty

burst size = number flipped

field step	0.002
poll	every 30 seconds
starting field	-2.00

A site flips up when the field it feels is positive. The field it feels is the external field, plus one for each of its six neighbours that is up and minus one for each that is down, plus its own fixed random field. When a site flips it changes what its neighbours feel, which can flip them, which can flip theirs. That cascade is one burst.

The relaxation is a queue, not a sweep. When a site flips, its six neighbours are pushed onto the queue and checked, and any that flip push theirs. The queue empties and the burst is over. Sweeping the whole lattice repeatedly would reach the same final state but would lose the burst boundary, and the burst boundary is the only thing this site actually measures.

The field advances in steps of two thousandths, and the lattice is relaxed at every single step. It is not relaxed once across a large jump. A large move in market cap should produce a run of many separate bursts of different sizes, because that is what the effect is. Collapsing it into one large burst would produce a correct final number and a false history.

The disorder is 2.16 because that is the critical value for this model in three dimensions, reported by Perkovic, Dahmen and Sethna in 1999. Below it the wire goes over in one enormous avalanche and there is nothing to watch. Above it the wire crumbles in small pieces and there is nothing to watch either. The whole site depends on that number being the published one rather than one chosen for the look of the result.

Nothing is stored except the state of the 4,096 sites and the highest field seen. Remanence, the layer counts, the histograms and every figure on this site are recomputed from those two things on load. There is no derived number kept anywhere that could drift away from the wire it claims to describe.

Everything else is bookkeeping. The market cap is fetched, converted to a field, compared against the highest field seen, and either ignored or walked up in steps. There is no other logic in the program and nowhere for a judgement call to hide.

├ what is real and what is ours

Real. The effect, found by Barkhausen in 1919. The model, the zero-temperature random-field Ising model of Sethna and colleagues, 1993. The critical disorder of 2.16 in three dimensions, from Perkovic, Dahmen and Sethna, 1999. The market cap, read from a public price API. The pool liquidity, read from the same public price API.

Ours. That a market cap should be a magnetic field at all. The particular mapping from one to the other. The seed the wire was built from, which is published so that anyone can build the same wire. Nothing else. There is no rule here that was chosen to make the picture look better.