H-032: Gate.io EVER 'smart wallets': deposits of wallets picked on 2022-23 foretell 7-day drops better than random wallets
Fail Family: flows · Status: rejected
Do 'smart money' EVER deposits on Gate.io predict drops? The idea did not beat simply buying the coin and holding it. It is rejected.
Result
17 smart wallets of 85; from 2024-01-01 their 33 deposits were followed by a mean 7-day log return of -0.0152 vs -0.0378 for random matched wallets (p = 0.931, need p < 0.05); exit rule Sharpe -0.19 vs hold -0.08
| Selection 2022-01-01 – 2023-12-24 | 85 eligible wallets, 17 smart, 889 events |
|---|---|
| Mean 7-day move after deposits (selection): smart / all | -8.70% / -3.08% |
| Test from 2024-01-01 | 33 smart-wallet events from 6 wallets |
| Mean 7-day move after deposits (test): smart / random wallets | -1.51% / -3.70% |
| p (random wallets, matched) | 0.931 |
| Exit rule vs holding: Sharpe | -0.19 / -0.08 |
| Exit rule vs holding: CAGR | -72.9% / -72.1% |
| Exit rule vs holding: max drawdown | -98.4% / -99.0% |
| Exit rule: time in the coin | 81% |
Why it might work
Idea by F1DevBin: wallets that move EVER to an exchange before it falls may know something, and if they do, the same wallets should keep doing it. Data: Gate.io's Everscale hot wallet (found 2026-10-28 from a withdrawal to the owner's wallet; ~70 M EVER, active since 2021-10). Users send EVER to personal deposit addresses that Gate sweeps into the hot wallet; a deposit address is a sender to the hot wallet that forwarded at least 90 % of its outflow there, and a 'wallet' is whoever sent EVER to such an address (crl data everscale, public GraphQL, addresses stored only as short SHA-256 ids; the owner's addresses are excluded when the data is built; transfers from Gate's own addresses are dropped). An event is a wallet's transfer of at least 10 EVER into a deposit address that had already been swept before it (so the address was knowable as Gate's at that moment); transfers within 7 days of the wallet's previous event join it; wallets with more than 100 transfers in a calendar month of the selection period are services and are dropped. Outcome: log return of EVER (Gate.io daily closes) from the close of the event's UTC day to the close 7 days later. Selection 2022-01-01 to 2023-12-24 (the last 7-day window ends before 2024, so no later price is used): wallets with 5+ events; the 20 % with the lowest mean outcome are 'smart' (need 25+ eligible wallets). Test from 2024-01-01: the mean outcome of the smart wallets' events (need 30+) against 1000 random sets of eligible wallets matched on their number of selection events; passes if p < 0.05 (one-sided: as low or lower). Also reported: holding EVER but staying out for 7 days after any smart-wallet event, decided at that day's close, with Gate's 0.2 % fee plus 0.2 % slippage per side, against holding. EVER only: this wallet-level data exists only on Everscale (an exception to the all-coins rule); GRAM on TON is a separate future hypothesis. Expectation written before any result: the 2022-23 'smart' wallets were mostly lucky and will not beat random wallets from 2024.
Verdict history
- 2026-10-28: Fail (this engine, data to 2026-10-27)
Pre-registered specification
kind: wallet_deposits
asset: EVERUSDT
universe:
- EVERUSDT
hot_wallet: 0:3addd84bf73267312a477049fd9b8db761bf39c585c150f8e6f9451347af2b6c
dataset: ever_gate_deposits.csv.gz
filters:
min_value: 10
swept_before: true
merge_days: 7
max_monthly: 100
selection:
start: '2022-01-01'
end: '2023-12-24'
min_events: 5
horizon_days: 7
smart_share: 0.2
min_wallets: 25
test:
start: '2024-01-01'
min_events: 30
'null':
method: random_wallets_matched
n: 1000
buckets:
- 5
- 6
- 8
- 11
- 16
- 26
significance:
alpha: 0.05
rule:
exit_days: 7
fee: 0.002
slippage: 0.002