# pool_eth_after backfill 1-5 Oct (#2487): partial, 17% of rows
DB (research-box): /opt/sw/dataeng/v3depth.db, table depth(id=v3hist.swaps.id, eth_after). Point-in-time pool WETH balance after each swap, rebuilt backwards from current WETH balanceOf(pool) +/- swap and liq deltas. head(pool,bal,n_ev,min_bal,ok) per pool.
Query: attach /opt/sw/v3hist.db as v; select s.*, d.eth_after from v.swaps s join depth d on d.id=s.id;
Coverage 1-5 Oct (swaps total / with eth_after): 1 Oct 627k/162k, 2 Oct 527k/85k, 3 Oct 503k/84k, 4 Oct 506k/66k, 5 Oct 399k/47k. Total 2.56M / 443k (17%), 1020 pools.
Gaps:
- 967k rows, 93 pools: busiest pools; reconstruction goes negative (min_bal -0.1 to -3.7 ETH; missed Mint/fee/collect events), so dropped as unreliable (ok=0).
- 479k rows, 504 pools with 42-char address: no WETH balance (not WETH-paired, or empty).
- 672k rows, 298 pools with 66-char id (v4 pool ids): no per-pool balanceOf, not reconstructable this way.
Exact point-in-time for the rest needs eth_call per swap at historical blocks (about 2M pool-blocks, infeasible on free RPC; paid cap too small). Reply if you want a best-effort version for the 93 busy pools (unverified, drift up to ~1 ETH).