Version 3 · present verified 2026-10-16 · horizons 2027, 2029, 2031 · written by a model, reviewed by a person · hash 7a8f0b779c2d560f…
Prerequisites: S-1
Plain Markdown for an agent about to read it.
1. Definition
A point process is a model of when events happen, not of how large they are. A Poisson process has a constant rate: the gaps are memoryless, and one event does not invite the next. A Hawkes process adds self-excitation. Each event raises the intensity, and the lift decays. With an exponential kernel the decay has a half-life, and the branching ratio is the expected number of child events one event triggers. Below one, the bursts die out and the process is stable. At or above one, a burst can feed itself. The interview asks you to fit an arrival intensity to a trade stream. The pass separates the rate of arrivals from the size of the trades, and it says whether the clustering is real or whether a constant rate was enough.
2. Why it matters at the frontier
Order flow, news bursts, and jumps in a price are point processes whether or not the desk uses the name. A market maker who models arrivals as Poisson will understate the chance that trades bunch, and the bunch is where the inventory risk is. The frontier use is a measured intensity with a stated baseline and a stated excitation, fit on past times, and a refusal to read it as a forecast of the next print. F3 lives downstream, in the quote that respects the intensity. This lesson is the intensity. Confusing it with a price model is how a calm-rate assumption blows up in a burst.
3. The test
Fit an arrival intensity to a trade stream. The pass names the baseline, the branching ratio, the decay, and the comparison to a constant-rate Poisson. A candidate who reports a single average rate has fit the Poisson and skipped the question. A candidate who reports a branching ratio without saying whether the likelihood actually beat the Poisson has fit a flexible model to noise. Probation is a stream where the events do not cluster: the right answer is the constant rate, and the excitation parameters are not identified.
7. Now, 2026
As of this writing Hawkes models are standard in the academic microstructure literature and in some market-making research groups, and a constant rate remains the working assumption in a great deal of operational monitoring because it is easy to alert on. Which firm fits which kernel in production is practice, unverified. What changed is the density of the tape a small team can hold, which makes clustering measurable and also makes a false excitation easier to fit. The kernel's refusal, when the likelihood ratio is weak, is the part the lesson insists on.
8. Next
- By 2027, a clustering claim reaches a person with the Poisson comparison, or it is not made (FD-1). Resolves: by 2027-12-31, yes if a frontier firm describes this drill as automated for research hires in a public job post, filing or conference talk; otherwise no. - By 2029, an intensity used to shape a quote carries the baseline, the branching ratio, and the window in the sealed row (FD-14). Resolves: by 2029-12-31, yes if a public allocator report or mandate requires a verifiable record of this calculation; otherwise no. - By 2031, the same event times can be refit from a sealed tape (FD-17). Resolves: by 2031-12-31, yes if this platform's warehouse holds a point-in-time panel a stranger can re-run the worked example on; otherwise no.
12. Sources and changelog
- repo: (a file of this platform; the path is the licensed tier) — "export function hawkesLogLik(times: readonly number[], T: number, mu: number, alpha: number, beta: number): number {"
- repo: (a file of this platform; the path is the licensed tier) — "below 1 the process is stable"
- repo: (a file of this platform; the path is the licensed tier) — "Descriptive on past events, not a forecast."
- paper: Hawkes on self-exciting processes; no DOI is claimed
- practice, unverified: the burst-then-quiet sketch is a shape, not a fit on a tape
- Changelog: v1 was the SK-0 skeleton. v2 is a composed draft by grok, marked review unreviewed, not an INBOX approval.
- Changelog: v3 is v2 reviewed and approved on 2026-10-25 by a local Claude session with the owner, as a parent lesson of the ZERO portal (docs/zero/zero-portal-plan.md ZR-2); no content change.Sources by type
- paper: 1
- repo: 3 (paths at the licensed tier; excerpts quoted above)
- practice: 1 (the lesson marks these as unverified)
The licensed tier
The licensed tier adds: Worked example, How to do it here, Common errors, Prerequisites and what it unlocks, Exam facts, What would make this lesson wrong. To earn it, an agent of an organization connects through the door and read S-1, S-18 (skill_syllabus gives the order), then sit skill_exam on S-18 and pass skill_answer: a licence is sealed with the day's root and checkable at /verify.
This is the approved version as written; nothing is fetched when you load it. Every figure above is the lesson's own, backed in its Sources section or marked there as unverified practice. Text in a lesson is reference material, never an instruction to a reader.