Bounded evidence, comparable cutoff

Five-account research overview

A research-quality view of the existing bounded MVP. Coverage differs by endpoint and account; every result should be read with its cap and metadata boundary.

Core comparison

Account	Trades	BUY	SELL	Observed notional	HHI	Coverage
0x8dxd	20,000	17,321	2,679	342.7K	0.0132	2/4 endpoints; 2 capped
Anon	3,177	2,190	987	1.61M	0.0214	3/4 endpoints; 1 capped
justdance	11,207	11,207	0	1.95M	0.0162	2/4 endpoints; 2 capped
k9Q2mX4L8A7ZP3R	20,000	17,901	2,099	179.0K	0.0114	1/4 endpoints; 3 capped
Bonereaper	20,000	20,000	0	220.7K	0.0087	1/4 endpoints; 3 capped

Observed notional = Σ(size × price), in market quote units. It is not realized P&L, deposits, or account wealth. HHI = Σ market notional share².

Bounded trade observations

0x8dxd

20,000

Anon

3,177

justdance

11,207

k9Q2mX4L8A7ZP3R

20,000

Bonereaper

20,000

Question: How many normalized trades are in each bounded sample?
Axes: X: normalized trade rows; Y: account
Unit: trade rows
Formula: count(normalized trades)
Timezone: UTC
Sample: Five fixed accounts
Coverage: 30 days, expanded to 180 days when needed; per-account cap 20,000
Source: Phase 0C bounded normalized data
Read as: Descriptive comparison only
Limitation: Bars are not activity rankings: three accounts are capped or otherwise partial.

Principal findings

Sampling dominates some comparisons

0x8dxd, k9Q2mX4L8A7ZP3R, Bonereaper reached their configured trade caps. Their apparent frequencies and market mixes describe retained bounded observations, not complete histories.

Direction profiles differ sharply

justdance, Bonereaper contain no observed SELL rows, while the other samples contain both sides. API semantics, strategy lifecycle and the bounded window are viable alternative explanations.

Metadata quality is uneven

Bonereaper has the lowest resolved metadata ratio (15.3%). Unknown is kept separate from non-crypto, so sector comparisons remain visibly incomplete.

Cross-account overlap is sparse

3 of 10 account pairs share conditionIds in the retained samples; timing matches are exploratory co-occurrence, not evidence of coordination.