Linear Thresholds Are a Quiet Mistake
Most circuit breakers look like this: tighten at minus 5 percent, cut sizing at minus 10, restrict entries at minus 15, halt at minus 25. Even spacing, round numbers, easy to read in a config file. And calibrated to the wrong thing.
Crashes do not move in even steps. They compound. The transition from a controlled losing stretch to a genuine cascade is not a straight line, and when you space your thresholds linearly you are tuning your governance to ordinary days rather than to the geometry of the events that actually threaten the account.
The failure is in the spacing. When a cascade starts, the time between threshold crossings collapses. Minus 5 to minus 10 might take two hours under normal stress. Minus 15 to minus 25 might take eight minutes in a real cascade. Linear thresholds assume equal time between steps. They do not account for acceleration. By the time you hit the halt level you may already be in free fall, having taken full-size entries most of the way down.
What a Crash Actually Does
A crash moves through stages: a small surprise that tests the first threshold, a larger move that tests the second, then a sharp discontinuous drop through the third and fourth. Each stage arrives faster than the last, because forced liquidations, stop cascades, and other participants' margin calls feed each other and amplify every successive move.
Your threshold structure has to mirror that acceleration. The gap between your first and second levels should be narrower than the gap between your third and fourth, because by the time you sit between three and four, conditions are deteriorating faster and your exposure should already be way down. Linear spacing violates this requirement by construction.
Phi as a Spacing Principle
Phi, roughly 1.618, shows up across natural systems as the ratio that governs growth, branching, and packing under constraint. It is also the ratio at which geometric sequences tip from controlled expansion into runaway. Whether that is a deep property of complex systems or a useful structural rhyme is an open question, and the practical argument does not need it settled.
Applied to circuit breakers, the principle is simple: space each threshold at phi times the previous one, measured from a base drawdown. With a 5 percent baseline:
- Normal operations: within minus 5 percent
- Tighten sizing: minus 5.00 percent (baseline)
- Restrict entries: 5.00 times phi, about minus 8.09 percent
- Near-halt posture: 5.00 times phi squared, about minus 13.09 percent
- Full halt: 5.00 times phi cubed, about minus 21.18 percent
Now the thresholds widen geometrically. The first gap, 5 to 8.09, is tight. The second, 8.09 to 13.09, is wider. The third, 13.09 to 21.18, wider still. Your governance is most sensitive exactly where it should be: at the early stage of a cascade, when intervention is still cheap and reversible.
A linear 5-10-15-25 ladder does not escalate until you are well into the loss. A phi ladder escalates sooner, proportionally at each step, and leaves room for a measured near-halt before the full halt fires.
The Effect in Practice
The claim here is not that phi is a law of markets. It is that exponential spacing fits the geometry of cascades better than a number line does. The first transition fires earlier under phi, near minus 8 instead of minus 10, which means governance starts cutting size before you are deep enough that new positions compound the damage.
The later spacing matters more. You do not slam into full halt at minus 15. You hit near-halt at minus 13, which buys the system a chance to shed exposure partially and let the market breathe before the full halt at minus 21. With linear levels at 15 and 25, you tend to jump from partial reduction straight to halt with nothing between them, because a ten-point gap is too wide to navigate cleanly mid-cascade.
Implementation
The code is trivial. The reasoning behind it is the point:
```python
PHI = 1.618
BASE_THRESHOLD = 0.05 # 5% drawdown baseline
GOVERNANCE_THRESHOLDS = {
"normal": BASE_THRESHOLD,
"tighten": BASE_THRESHOLD,
"restrict": BASE_THRESHOLD * PHI,
"near_halt": BASE_THRESHOLD * (PHI ** 2),
"halt": BASE_THRESHOLD * (PHI ** 3),
}
```
Every threshold derives from a principle instead of a hunch. When a partner or reviewer asks why your halt sits at 21 percent, you can say: these levels are phi-spaced from a 5 percent baseline because exponential spacing matches the acceleration of cascade events. That is auditable. "They looked about right" is not.
The ratio also travels. A 30,000 dollar account and a 3,000,000 dollar account use the same structure with the same base percentage and the same spacing. You are not re-deriving magic numbers every time capital changes.
Where It Breaks
Phi spacing assumes your losses deteriorate in a roughly exponential pattern. If your strategy loses in sharp idiosyncratic spikes rather than accelerating cascades, the spacing may not fit. The honest response is to test both ladders against your own loss history and keep the one that would have escalated earlier and more proportionally through your worst drawdowns.
The ratio is not sacred. The discipline is: derive threshold spacing from a structural claim about how systems fail, not from whatever numbers read cleanly when you first wrote the config.
Principle Over Parameters
Linear thresholds are easy to read and easy to explain, and they are a quiet form of cargo-cult risk management. The structure looks like governance but was never designed to match the thing it governs. Circuit breakers exist to catch cascades. Cascades are not linear. Your thresholds should not be either.
Run your ladder against your worst historical drawdowns and ask one question: did governance escalate at the right moments, or did it hold at full activity until the cliff edge? If the answer is the cliff edge, you have a spacing problem, and phi gives you a principled way to fix it.
Architecture beats scale. Principle beats parameters. Build thresholds shaped like the failure they exist to prevent, not like a number line.