Datasets:
Quick question about a few dataset fields
Hi, thanks for sharing the dataset. Could you clarify a few fields? For creator_past_tokens, when does the count start, and does it ever reset? For creator_past_rugs, what counts as a rug, and when is it added to the count? For holder_count, top1 and top10, when is the snapshot taken? Do you count wallets or token accounts, exclude any accounts, and what total are the percentages based on? If there is calculation code or a short explanation available, that would be really helpful. Thanks!
creator_past_tokens / creator_past_rugs
Calculated once, right when a new token is detected (before it has any trading history). It counts every other token by that same creator address across the full dataset. creator_past_tokens is just that total count. creator_past_rugs is defined loosely — it’s every prior token by that creator that never graduated, not a judgment call about intent or liquidity pulls. So a token that simply failed to gain traction counts the same as one that was deliberately rugged. It’s a lifetime count and never resets.
holder_count, top1, top10
Snapshotted once, about 30 seconds after the token’s creation (and again ~30s after graduation, which overwrites the earlier snapshot). It comes from a single Solana RPC call (getTokenLargestAccounts), which only returns the top 20 holder accounts — so holder_count isn’t a true total-holder count, it’s “how many accounts showed up in that top-20 response.” It’s accounts, not deduplicated wallets, so one wallet split across multiple accounts would show up more than once.
Before computing top1/top5/top10, the single largest account is dropped if it holds more than half the supply — pre-graduation that’s always the bonding curve itself, post-graduation it’s typically the DEX liquidity pool. Everything else is kept. The percentages are all calculated against a fixed total supply of 1,000,000,000 tokens (not the sum of fetched holder balances).
Happy to help if you need anything more