Proof it works · a real DOJ case, run blind

Found all 3 wallets the DOJ seized,
without being told where to look

From the victim's report alone.

8 of 8Intermediate wallets the DOJ named, found

3 of 3Final wallets the DOJ seized, found

+0.9%A-7 amount vs the DOJ's

+1.0%A-8 amount vs the DOJ's

+10.8%A-9 amount vs the DOJ's

Scoreboard: Hound vs DOJ ¶ 74

Blind run

Wallet	Address	DOJ	Hound	Difference
A-7	0xa177…0021	314,657 USDT	317,403.01 USDT	+0.9%
A-8	0x4e5c…abb5	11,954 USDT	12,077.78 USDT	+1.0%
A-9	0xafd7…78a5	43,071 USDT	47,702.92 USDT	+10.8%

How this was scored

Amounts: victim A.H.'s money in each final wallet, against DOJ ¶ 74
Bands: within 5% 5 to 20% over 20%
DOJ printed the start in ¶69, truncated; we resolved it; the engine got nothing else.
DOJ also traced 37 other victims and accounts with mixed methods and subpoena data: context only, never scored
Wallets reached: A-2, A-3, A-4, A-5, A-6, B-2, B-3, B-4, A-7, A-8, A-9
Never given to the engine (11), plus the 4 where her sends start and land, which it had to be given

Backtest · Tether freezes in 2025

192 frozen wallets measured

45 daysMedian time the money sat there

44%Already gone by the freeze (median)

399 daysSlowest tenth took longer than this

Method

detection_lag_hours = first USDT inflow -> AddedBlackList event; escape_share = USDT out before freeze / USDT in before freeze
Population: 657 Ethereum USDT addresses Tether blacklisted in 2025 (eth_getLogs)
random.seed(13), sample 200
Wallets with >5000 token transfers skipped
OFAC hits tagged 'sanctions' and excluded from the headline; the rest is LE/scam/hack, not separated
Survivorship: measures only addresses Tether did freeze
BlockSec's 54% burned figure is theirs

This weekend's scan

Reported, not proven

91,218 USDTReported and not yet frozen

14 of 963Reported addresses holding it, of those checked this weekend

Show the 14 addresses

Address	Counted	Category
0x6f72…22a1	55,883.15 USDT	fake returns