poidh v1 — per-wallet payouts, earnings and NFT holdings

496

bounties

278

completed

169

cancelled

still open

722

claims / NFTs

415

wallets

3.7269

ETH paid out

6.5365

ETH ever funded

Download the CSV → 415 rows, one per wallet, 16 columns. The three the bounty asked for are eth_paid_out, eth_earned and nfts_held; the rest are the working behind them.

"Paid out" and "earned" are not the same number, so both are given. The contract takes a 2.5% fee on acceptance — fee = amount * 25 / 1000, which holds exactly on all 278 completed bounties — and sends the remaining 97.5% to the claimant. So an issuer's outflow is the full bounty amount while the claimant's inflow is 97.5% of it. Across all of v1: 3.726923002 ETH left issuers, 3.633749927 ETH reached claimants, and 0.093173075 ETH went to the treasury at 0x1a706a29…3fef9. The CSV carries all three as separate columns, so you can add up whichever definition you meant.

444 of the 722 NFTs are held by the contract, not by people. v1 mints a claim's NFT to itself and only releases it when that claim is accepted — and it releases it to the bounty issuer, not the claimant. Every rejected or never-judged claim therefore leaves its NFT locked in the contract forever. That is why the contract address is the largest holder in the table below, and why only 80 real wallets hold anything. It is listed rather than dropped so the column still sums to totalSupply() = 722.

Every wallet

click a header to sort

How it was built

Five small scripts, no dependencies beyond an ABI decoder, each linked below. There is no database and no indexer: the whole history is one eth_getLogs call.

pull_logs.js — eth_getLogs over the contract's full range in a single call. Arbitrum's public endpoint is not an archive node, but logs come from the receipt index, so no historical state is needed. Output: logs.json (3,466 logs, blocks 117,178,213 → 443,301,520).
decode.js — decodes them into events.json: 496 BountyCreated, 722 ClaimCreated, 278 ClaimAccepted, 171 BountyCancelled, 1,072 Transfer.
verify.js — 22 invariants, all of which must pass before a CSV is written (output).
build_csv.js — writes the CSV and summary.json. All arithmetic is in wei as BigInt; ETH strings are exact decimal expansions, never floats.
verify_chain.js — asks the contract itself (output).

The claim NFT's metadata and its image are generated by a sixth script, gen_card.py, which reads every figure it prints out of summary.json, the CSV and verify.txt at render time and refuses to draw at all if the verification output is not clean — an image of a data pull that disagrees with the data pull is worse than no image.

Why you can trust the numbers

This contract has no verified source — it is not on Sourcify and Arbiscan wants an API key — so the event signatures came from openchain's topic database and the ABI was reconstructed. That is exactly the situation where a plausible-looking CSV can be quietly wrong, so every assumption is pinned by something that would break loudly if it were false:

The ETH ledger closes. Funded (6.536511134) = paid out (3.726923002) + refunded (2.591392421) + still escrowed (0.218195711), and that last figure equals the contract's live balance to the wei. Nothing is unaccounted for, and no bounty was double-counted.
Cross-event identity. Every ClaimAccepted names a bounty issuer and a claimant that match the corresponding BountyCreated and ClaimCreated records; every accepted claim really belongs to the bounty that accepted it.
The partition is exact. 278 completed + 169 cancelled + 49 open = 496, with no bounty in two states. (There are 171 cancel events for 169 bounties — two were cancelled twice.)
NFT holdings were checked against the chain, not just derived. Replaying 1,072 transfers gives a balance for each wallet; balanceOf() was then called for all 415, including the 334 the replay says hold nothing. Zero mismatches, and the column sums to totalSupply().
The indexed parameters were established, not assumed. topic1 holds a 20-byte address in every BountyCreated and ClaimCreated log, which fixes issuer as the indexed field in both.

Two invariants failed on the first run and both were real: the claim NFT is minted to the contract rather than the claimant, and the event's 7th word repeats the claim id rather than being the timestamp. Both are fixed above; the failures are why the checks exist.