Terexitarius2/commoncryptotokens

# Top 10 Wallets Trading Common Crypto

This dataset contains per-day action labels for the top 10 wallets per asset-network pair, built from real on-chain wallet activity and aligned with OHLCV market context.

Primary data file in this repo:
- `top 10 wallets trading common crypto.parquet`

## What this dataset is

- Scope: top-ranked wallets per asset (up to 10 wallets per asset-network pair).
- Granularity: daily rows per wallet.
- Intended use: training/finetuning models that learn wallet behavior, conviction, and market-timing patterns.
- Source construction: wallet ranking from observed trading behavior, then daily action matrix aligned to market prices/signals.

## Current dataset snapshot

- Rows: 521,716
- Distinct symbols: 39
- Distinct networks: 7
- Distinct asset-network pairs: 39
- Distinct wallets: 270
- Date range: 2010-07-17 to 2026-04-18 (UTC dates)

## Symbols included

AAVE, ARB, AVAX, BNB, BTC, BTT, CRV, CVX, ETH, EURA, EURC, EURCV, EURQ, EURS, EURe, GLDx, JPYC, JST, LDO, LINK, META, MKR, NFT, OP, OPENX, SNX, SOL, SUN, SUNDOG, TUSD, UNI, USD1, USDC, USDD, USDT, VTX, WBTC, WIN, ZCHF

## Core columns

- `date`: UTC calendar date for the wallet decision row.
- `symbol`: asset ticker.
- `network`: chain/network id.
- `rank_in_asset`: wallet rank within the asset-network set.
- `wallet_address`: on-chain wallet address.
- `wallet_score`: wallet quality/strength score.
- `wallet_bias`: historical directional bias classification.
- `daily_wallet_action`: daily action label (e.g., BUY/SELL/HOLD).
- `action_confidence`: confidence score for daily action.
- `market_daily_move_pct`: same-day market move (%).
- `wallet_directional_move_pct`: wallet-directional market move (%).
- `trade_signal`: market signal label used in action generation.
- `signal_confidence`: confidence for the signal.
- `open`, `high`, `low`, `close`: daily OHLC values.
- `next_close`: next day close.
- `next_day_return_pct`: next-day return (%).

## Notes

- This is the training-focused curated file only (single parquet artifact).
- The repo intentionally excludes auxiliary intermediate artifacts.