GuardPay research · TRON
Can you tell in advance that USDT will be frozen
We trained a model on Tether freezes on the TRON network up to October 25, 2026. Then we tested it on 208 wallets that Tether froze from October 26 to November 13: the model had never seen them. We scored each wallet on data from a week before its freeze. For comparison we took 2,000 random active wallets.
wallets frozen after the model was trained
of them (17.8%) got a high risk a week before the freeze
of 2,000 random wallets (1.6%) got a high risk by mistake
more often a high risk for future frozen wallets than for ordinary ones
Check any TRON address: GuardPay shows the freeze forecast, whether it is already frozen and whether it had transfers with frozen addresses.
High risk a week before the freeze
Share of wallets the model gave a high risk. Scored on data from 7 days before the freeze; everything later was hidden.
All scores
| Score a week before the freeze | 208 frozen | 2,000 random |
|---|---|---|
| High risk | 37 | 31 |
| Elevated | 12 | 13 |
| Moderate | 10 | 13 |
| Low | 146 | 1919 |
| Cannot score | 3 | 24 |
“Cannot score” means there was no data to score the wallet.
Key findings
Why it matters
Method
What to know about the data
What we do with this
The freeze forecast appears in the GuardPay check for every TRON address. Since November 20 we keep a forecast journal: the list of high-risk addresses is fixed in advance and revealed only after a freeze.