1. Wallet selection
The dataset follows a fixed, continuously maintained set of 20,000+ Ethereum wallets. A wallet is chosen for what it holds, not for how much it trades:
- Holdings threshold. About $250K or more in ETH and volatile tokens. Stablecoins and pegged assets (USD, EUR and gold-backed tokens) don’t count toward the threshold.
- Where candidates come from. Large on-chain transfers and swaps, exchange flows, and top-holder lists for the tracked tokens.
- What is removed. Known exchange, bridge, router and other infrastructure addresses, and wallets whose trading volume is out of all proportion to their holdings (a market-maker pattern).
- Re-checked every few hours. Wallets that fall well below the threshold or stop qualifying are removed, and newly qualifying wallets are added. The set therefore changes over time.
2. Every move, labeled for what it is
A move is recorded when a transfer or swap touches a tracked wallet. The counterparty decides the label: a known exchange address makes it an exchange deposit or withdrawal; a DEX pool or router makes it a swap.
| action | What happened | venue_type | A trade? |
|---|---|---|---|
| CEX_WITHDRAW | Tokens left a known exchange address for a tracked wallet. Read as the accumulation side of flow. | exchange | No |
| CEX_DEPOSIT | A tracked wallet sent tokens to a known exchange address. Read as the distribution side of flow. | exchange | No |
| SWAP_BUY | A tracked wallet received the token in a DEX swap. | dex | Yes |
| SWAP_SELL | A tracked wallet gave up the token in a DEX swap. | dex | Yes |
An exchange withdrawal is not a purchase, and a deposit is not a sale. What happens on the exchange’s own books is not visible on-chain. Most tracked whale flow is exchange transfers: in the current sample, 78% of rows are exchange moves and 22% are DEX swaps. That is why every row carries venue_type, and why the two are never added together as “buying” or “selling”.
3. Pricing
Each move carries its USD value at the time of the move. Prices pass plausibility checks before they are stored, and implausible per-token prices are rejected rather than recorded. Our aggregate queries also leave out any single row above $1 billion, so one mispriced row can’t dominate a total. Values are approximations and can be off for thinly traded tokens.
4. History depth
- Tracked whale wallets: moves are kept permanently from October 7, 2026, and history grows every day.
- Exchange flows for other wallets (the counterparty side that is not in the tracked set): kept for a rolling 30 days.
- Older exchange-flow rows from before October 7, 2026 were not retained. Earlier DEX history exists for some wallets but is not uniform, so it is not offered as a dataset.
5. The free sample file
The public sample is a small slice of real data that shows the format: moves by tracked whale wallets over the last 7 days, for the 10 coins with the most tracked-whale USD volume in that window, capped at the most recent 5,000 moves and regenerated weekly. Current file: generated 2026-10-16 UTC, 4,536 rows, 0.8 MB.
| Column | Meaning |
|---|---|
| time_utc | Block time of the move, ISO 8601 UTC. |
| tx_hash | Ethereum transaction hash. |
| wallet | The tracked whale wallet (public on-chain address). |
| token | Token symbol. |
| action | CEX_WITHDRAW, CEX_DEPOSIT, SWAP_BUY or SWAP_SELL (see the table above). |
| usd_value | USD value at the time of the move. |
| venue_type | exchange or dex. |
Line 1 is a comment starting with #; the header is line 2. There is one row per transaction, wallet and direction, so a transaction hash can appear more than once. The top 10 includes stablecoins when they rank by volume. A few very active wallets can account for many rows.
# Python
import pandas as pd
df = pd.read_csv("https://deepbluealpha.io/data/sample.csv", comment="#")
df.groupby(["token", "venue_type"])["usd_value"].sum()
6. Known limits
- Labels are heuristics. Exchange and DEX labels come from known address lists. A new or unlabeled exchange or venue can be missed or mislabeled, and wallet labels never identify a person.
- Coverage is the tracked set, not every wallet on Ethereum, and the set changes as wallets are added and removed. Totals from different weeks describe slightly different groups of wallets.
- Delays and revisions. Moves can arrive late, be revised, or be affected by chain reorganizations and by outages at data or infrastructure providers.
- Ethereum mainnet only. Moves on other chains and layer-2 networks are not included.
7. Not investment advice
This data describes what tracked wallets did in the past. It does not predict prices and is not a recommendation to buy, sell or hold any asset. Use of the data is governed by our Terms of Service, Disclaimer and Business & API Terms. Questions: team@deepbluealpha.io.
