winning-wallet-finder/HANDOFF.md

Session handoff — 2026-07-08 (rev 3: books aligned, Set E live, fresh $1k)

Self-contained pickup for a fresh session (human or AI). Read README.md gotchas 1–15, FINDINGS.md (especially the three 2026-07-08 sections at the end), and LIVE_ROLLOUT.md. The repo is authoritative; project memory has deeper history.

Where things stand (all green)

• Live paper bot: Fly.io wwf-copybot, region arn (Stockholm), push mode, ONE machine. Following Set E (7 volume wallets): LSB1, imwalkinghere, Kruto2027, 0xbadaf319, gkmgkldfmg, AIcAIc, 1kto1m. flyctl logs --app wwf-copybot.
• The book is a fresh $1,000, reset 2026-07-08 ~20:45 UTC — this is the CALIBRATION EXPERIMENT: measure the live-to-backtest ratio for Set E with every fix live from day one; that ratio (not the replay %) sizes real money. Old book (Jun 25 → Jul 8, +$229.79) is in git history + archive/copybot_fills.pre-reset-2026-07-08.jsonl.
• The three books agree by construction (2026-07-08 alignment work): same W/L/R/S taxonomy (win% = held-to-resolution only), same sold-vs-redeem price discriminator (redeem prints exactly the payout), same pinned floors, same follow set. Row-level cross-check bot↔backtest: 7/9 exact, 0 absent, 2 legit execution diffs. Sharps table reproduces PM's per-position books to the dollar; PM's LEADERBOARD is the biased number.
• Dashboards: jaxperro.com/trading = paper book + backtest + sharps; jaxperro.com/live = REAL MONEY page (NOT STARTED until Phase 2). Shared renderer: jaxperro/trading/copybot-section.js. ?botFeed=/?rmFeed= query params override feeds for local testing.
• Watchdog live (2026-07-08): Fly /health check auto-restarts a dark machine; GH Actions watchdog.yml probes from GitHub's infra and pings Discord after 3 failed probes. copytrade.py now lives at repo root (archive/ is 100% retired code again).
• Daily pipeline (Mac launchd 08:00, fires on wake — typically ~10:00, caffeinate-wrapped) appends one live-vs-model row/day to live/history/calibration.csv and re-derives everything under the new rules: entries cursors now invalidate daily, _open_split folds by the redeemable flag, rtally splits SOLD.
• Backtest: portfolio.py takes --bank (backtest.json "bank"). Published feed stays at $1,000 to match the paper book. $500 run: +2728% (small banks compound faster — their-bet ceilings bind later).

LIVE PLACEMENT — root cause SOLVED, one user action + one port remain

Why the real-money bot placed nothing (2026-07-09/10 investigation, every link proven): (1) the first two qualifying signals CRASHED the bot inside the unguarded live order path (machine events: exit_code=1 at 21:21:55 and 21:53:35, seconds after each signal) and boot-baseline then marked the fresh trades seen — both bugs FIXED (executor never raises; poll guard; baseline exempts trades younger than the stale window). (2) With crashes fixed, the probe exposed the real wall: py-clob-client is ARCHIVED and the CLOB rejects its orders ('invalid order version') — reads/auth still work, placement is dead for everyone on the old client. (3) The successor is Polymarket/py-sdk (pip install --pre polymarket-client, SecureClient). (4) Under it, the account's legacy proxy (0xbb2aC6b8…) is refused as maker ('use the deposit wallet flow'); bootstrap (host/wallet_bootstrap.py) minted a Builder API key with existing auth and DEPLOYED the signer's Deposit Wallet: 0x455e252e45Ee46d6C4cc1c8fAdD3899d68f245a1.

Remaining — WRAP STEP ONLY (status 2026-07-10 early AM): the $24.73 sits SAFE as native USDC at the deposit wallet (user's key controls it; recoverable/transferable at any time). Proven working tonight via probes: auth ✓, gasless execution ✓ (two real relayer txs: USDC approval 0x1044d3c7…, earlier 0x935be727…), setup_trading_approvals ✓ (max-uint on all spenders), order FORMAT accepted ✓. The one remaining gap: the exchange's balance view counts only the COLLATERAL token (0xC011a7… wraps native USDC; UI flows wrap on deposit; our raw transfer bypassed it). Attempted: public wrapper periphery 0x93070a… wrap(asset,to,amount) via gasless batch → 'batch would revert' (likely caller-role-gated); neg-risk adapter ruled out (its mints are redeem change). NEXT SESSION (do NOT guess with live funds): read the ts-sdk's deposit/wrap implementation for deposit wallets (Polymarket/ts-sdk — same unified family, will contain the exact supported call), replicate via execute_transaction, THEN port LedgerLiveExecutor to the new SDK (place_market_order: BUY amount=$USD / SELL shares=, FAK; construct SecureClient(private_key) — wallet resolves, no api_key needed at runtime once approvals exist), add polymarket-client to fly.Dockerfile, re-probe (host/order_probe_v2.py — contains the full working onboarding sequence), re-arm. CLEANUP: several builder API keys were minted by probe runs — fetch_builder_api_keys/revoke the extras. The armed live bot remains harmless (ok:False + honest missed rows on every attempt).

Queued next-session work (in value order)

0. MORNING FORENSICS — paper-book drift −11.32 (alarm intentionally red). The clobber bug is FIXED (settled records archive to token#ts keys on re-entry; two vanished records recovered from state git history and restored). What remains: with truth restored, drift reads −11.32 — a separate cash-outflow-without-ledger-counterpart accumulated during 2026-07-09's 27-copy day. Suspect classes: sell fills drained against already-settled records (sold_proceeds mutated post-pnl counts nowhere), settle math on re-entered tokens, trim flows. Method: replay copybot_fills.jsonl + state-history settles against the invariant term-by-term until the residual attributes. The fills ledger now has per-token BUY lines + book snapshots; state history is committed every few minutes. LIVE BOOK UNAFFECTED (0 trades, drift 0.00) and protected by the clobber fix before its first re-entry.

1. Exact-res_t migration: plan agreed 2026-07-08 — execute ETHERSCAN_MIGRATION.md phase by phase (0: chain sweep backfill → 1: shadow audit → 2: validate_timing → 3: portfolio → 4: selection+floors → 5: trust simplification LAST). One phase per session, before/after diffs, never coupled with follow-set changes.

2. Empirical fill model: every copy now logs a book snapshot (spread + 5c depth) to the fills ledger; once a few weeks accumulate, fit slippage vs stake/depth and replace portfolio.py's flat SLIP haircut; add a depth gate before real-money sizing.

3. Funding-cluster tracer port to Etherscan logs (Alchemy free tier now caps getLogs at 10 blocks — insider.py's ring detection is degraded until then).

Watch list (the standing question: does Set E hold up?)

• 0xbadaf319 — 2026-07-08 evening revealed his structure: PAIRED YES+NO clips ($6–37 each, sums ≈ $1.00) = two-sided arb/merge flow, not directional conviction. Copying one leg of a hedge is a bet HE never took; his replay P&L rode the larger leg of pairs. Moves to the TOP of the demotion watch list — judge on forward copies, not his account P&L.
• AIcAIc — held-win only 42%; his edge is sell-timing (most lag-fragile class). Second demotion candidate.
• 1kto1m (z=2.4) and gkmgkldfmg (z=2.05) — near the gate floor; strong month, shallow statistical edge.
• Re-ranking is cheap: the per-wallet replay harness pattern is in FINDINGS ("Aligning the three books"); python3 live/portfolio.py --wallets 0x… --days 30 --out /tmp/w.json per candidate, then combined-set runs.
• Rejected-with-evidence (don't re-add without new facts): oliman2 (true lifetime ~$19k), leegunner (negative to copy), 0xb0E43B/ArbTraderRookie (refund harvesters — real edge, uncopyable margins), lma0o0o0o (negative in the shared book), Winnertraders (6.6-day capital locks).

Real-money worker — ARMED AND LIVE (2026-07-09 ~04:37 UTC)

wwf-copybot-live is trading REAL MONEY: $24.73 book, Set E, $5 stakes, rule-0.6 caps ($5/trade · $25/day · $30 exposure), poll mode 60s, own state copybot_state.live.json, feed live/copybot_live_real.json → jaxperro.com/live. Armed via LIVE_CONFIRM env (user-typed phrase; unset to disarm at next boot; flyctl apps stop wwf-copybot-live = instant stop). auto_redeem OFF (no POL gas) — after a WIN, redeem manually in the UI; CASH≠CHAIN nags until done. Discord pings per placement/exit/settle via DISCORD_WEBHOOK secret. Phase 4 (supervised first fill checklist in LIVE_ROLLOUT) runs when the first conviction signal lands — zero live trades so far; fill-split merging (see below) shipped 2026-07-09 evening after 13h of sub-floor-clip skips. No watchdog on the live app yet (poll mode, no HTTP) — liveness = flyctl logs + feed freshness.

Venue fork — RESOLVED 2026-07-09 (user funded polymarket.com)

The first deposit landed on polymarket.us by mistake (off-chain venue, no keys — incompatible with the stack; $32 still sits there, user may withdraw). User then funded a polymarket.com account ($24.73 after fees) and the live worker armed against it. Residual security to-dos: revoke the polymarket.us API key f03d9eb5… (secret appeared in a chat screenshot) and rotate the Discord webhook committed to this repo's history pre-2026-07-09.

Next: Phase 2 — funding (USER ONLY)

Phase 1 is 8/8 complete (LIVE_ROLLOUT.md). Before funding, the user should decide the stranding tolerance explicitly: real money turns the primary risk from P&L into the ACCOUNT (US person routing orders via Stockholm through the geoblock — a frozen wallet is a different loss category than a losing bet). Caps stay $5/trade · $25/day · $30 exposure until Phase 6. Then Phase 3 preflight (preflight_live.py) → Phase 4 supervised first fill → Phase 5 edge-case matrix. The typed-phrase interlock is the human checkpoint — never automated.

Standing to-dos (user-only or external)

• None open. Alchemy key ROTATED 2026-07-08: old app "Jax's First App" (key …OdWgOi, the one leaked to public git history) deleted; new app wwf-rpc-2026-07 (key …w0BxV5) wired into gitignored config.json + the ALCHEMY_RPC_URL Fly secret. Verified: old key → HTTP 401 (the key in git history is now dead, so no history rewrite needed), new key serves Polygon, bot rebooted with settle-fallback ON and made a live copy, Notify webhook wh_blf4qjjvfdbqs9mc survived (app-delete doesn't touch Notify resources) — 7 addresses in sync. Railway project deleted (purges 07-10).

Operational quick-reference

• Change the follow set: edit live/copybot.paper.json wallets → ./live/deploy_bot.sh (validates → pins floors → syncs Alchemy webhook → restarts Fly → confirms banner). Keep backtest.json mirrored.
• State surgery (reset/repair the book): flyctl machine stop FIRST, push the new state, flyctl apps restart, then VERIFY the first heartbeat shows the book you wrote (gotcha 15: the bot's memory wins conflicts by design, and a boot clone can read a stale GitHub replica for a few seconds).
• Never 2 Fly machines (flyctl scale count 1) — one book, one writer.
• Cache is single-writer; don't run cache-touching scripts during the daily pipeline or a regen. Three cursors per wallet (gotcha 12).
• Bot commits its own state/feed — git pull --rebase --autostash before pushing. Image changes (fly.Dockerfile, host/) need flyctl deploy --remote-only; code/config changes just push + restart.