Polymarket Monitor

v2
live · auto-refresh - ↻ reload
wallets SLA628 / 48,417 (1%)trades 1m0agents alive24/46 (U 24/24 · C 0/22)copies 1h0 (U 0 · C 0)proxies0 / 0proxy quota-

System Status

Every moving part - refresh every 10s. Click any bot card for its detail page.
Overall: 57 online · 11 degraded · 34 offline · 0 error / 102 checks
As of 2026-06-14 19:52:37 UTC

Core Infrastructure

5🟢 · 1🔴 · DB, cache, external feeds, proxy pool
Postgres 🟢 ONLINE
connectable · 0s ago
primary store (paper_account, paper_positions, copy_bot_signals, etc.)
Redis (truth_api cache) 🟢 ONLINE
0 keys, 0% hit rate · 0s ago
hot-path cache for wallet state, leaderboard, market state
Polymarket CLOB WebSocket 🟢 ONLINE
3 subscribed, 38,207 msgs · 0s ago
real-time orderbook + market_resolved events
Coinbase BTC feed 🟢 ONLINE
1m ago · 1m ago
5s BTC spot polls into the 5m up/down bot
Polymarket gamma API 🟢 ONLINE
trade 59s ago · 59s ago
market discovery + resolution data feed
Proxy pool 🔴 OFFLINE
0/0 hot · 0s ago
no proxies registered

Daemons (systemd services)

4🟢 · 2🔴 · inferred from DB writes - if a daemon stops, you see it here first
polybot-bot (wallet poller) 🟢 ONLINE
1m ago · 1m ago
polls 1,250 wallets across T1/T2/T3 tiers for new trades
polybot-copybot (user-side bots) 🟢 ONLINE
43s ago · 43s ago
runs 20 user-side bots (24 heartbeat rows)
polybot-claude-agents (Claude-side) 🔴 OFFLINE
2d ago · 2d ago
runs 20-22 Claude-side bots (21 heartbeat rows) - DEAD, daemon may be stopped or crashed
polybot-claude-btc (BTC 5m bot) 🔴 OFFLINE
21h ago · 21h ago
places BTC up/down bets on the 'default' paper account every 5m - DEAD, daemon may be stopped or crashed
polybot-claude-mart (martingale variants) 🟢 ONLINE
0s ago · 0s ago
5 martingale ladder variants on the BTC 5m signal
polybot-dashboard (this page) 🟢 ONLINE
serving · 0s ago
Flask web UI on :80

Bots (45 agents)

24🟢 · 10🟡 · 11🔴 · heartbeat freshness · green <180s · yellow <450s · red ≥450s or paused
COPY_SHARP
43s ago · 732,308 evals, 0 copies
COPY_WHALE
43s ago · 732,333 evals, 0 copies
COPY_CONSENSUS
43s ago · 732,333 evals, 0 copies
COPY_HOTHAND
43s ago · 732,333 evals, 0 copies
COPY_FADE
43s ago · 732,333 evals, 0 copies
AGENT_LAMBDA_TAIL
43s ago · 732,333 evals, 0 copies
AGENT_NOTHING_HAPPENS
43s ago · 732,333 evals, 0 copies
AGENT_SPORTS_FLB
43s ago · 732,333 evals, 0 copies
AGENT_HOLD_TO_SETTLE
43s ago · 732,308 evals, 0 copies
AGENT_EDGE_NET_KELLY
43s ago · 732,333 evals, 0 copies
COPY_FAST_QUALITY
43s ago · 732,333 evals, 2 copies
COPY_BIG_TICKET
43s ago · 732,333 evals, 0 copies
COPY_LOW_PRICE
43s ago · 732,333 evals, 0 copies
COPY_HIGH_PRICE
43s ago · 732,333 evals, 1 copies
COPY_LATE_RESOLVE
43s ago · 732,333 evals, 0 copies
COPY_HOTHAND_FAST
43s ago · 732,333 evals, 0 copies
COPY_VOLUME_1K
43s ago · 732,333 evals, 0 copies
COPY_SPORTS_QUALITY
43s ago · 732,333 evals, 0 copies
COPY_CRYPTO_QUALITY
43s ago · 732,333 evals, 0 copies
COPY_NEW_LISTING
43s ago · 732,333 evals, 2 copies
COPY_CRYPTO_ELITE
43s ago · 732,333 evals, 0 copies
COPY_RWO_BTC
43s ago · 732,333 evals, 0 copies
INVERSE_SHARKY_BTC
43s ago · 732,333 evals, 0 copies
INVERSE_0X_0C16
43s ago · 732,333 evals, 0 copies
CLAUDE_SHARP_CONVERGENCE
7d ago · Phase 1 bleeder triage 2026-06-04 — lost 30%+ since 06-03 re
CLAUDE_MAKER
7d ago · no recent heartbeat
CLAUDE_TAIL_DISCIPLINED
7d ago · Phase 1 bleeder triage 2026-06-04 — lost 30%+ since 06-03 re
CLAUDE_SPORTS_MIRROR
7d ago · no recent heartbeat
CLAUDE_ENSEMBLE
7d ago · Phase 1 bleeder triage 2026-06-04 — lost 30%+ since 06-03 re
CLAUDE_WHALE_PRECISION
7d ago · Phase 1 bleeder triage 2026-06-04 — lost 30%+ since 06-03 re
CLAUDE_RESOLUTION_SNIPER
7d ago · no recent heartbeat
CLAUDE_CONSENSUS_PROB
7d ago · no recent heartbeat
CLAUDE_REVERSAL_HUNTER
7d ago · no recent heartbeat
CLAUDE_SHARP_FOLLOW
7d ago · Phase 1 bleeder triage 2026-06-04 — lost 30%+ since 06-03 re
CLAUDE_QUALITY_KELLY
7d ago · Phase 1 bleeder triage 2026-06-04 — lost 30%+ since 06-03 re
CLAUDE_WHALE_LOOSE
7d ago · Phase 1 bleeder triage 2026-06-04 — lost 30%+ since 06-03 re
CLAUDE_DEEP_TAIL
7d ago · no recent heartbeat
CLAUDE_HIGH_TAIL
7d ago · no recent heartbeat
CLAUDE_FAST_REACT
7d ago · Phase 1 bleeder triage 2026-06-04 — lost 30%+ since 06-03 re
CLAUDE_SHARP_PROVEN
7d ago · Phase 1 bleeder triage 2026-06-04 — lost 30%+ since 06-03 re
CLAUDE_VOL_FADE
7d ago · no recent heartbeat
CLAUDE_MID_VOLUME
7d ago · Phase 1 bleeder triage 2026-06-04 — lost 30%+ since 06-03 re
CLAUDE_PRE_RESOLVE
7d ago · no recent heartbeat
CLAUDE_FRESH_FLIP
2d ago · no recent heartbeat
CLAUDE_PMXT_ARB
7d ago · no recent heartbeat

Signal flow (last hour)

24🟢 · 1🟡 · 20🔴 · acted/total signals per bot · 0 = scanner silent · all-skipped = config too tight
AGENT_EDGE_NET_KELLY 🟢 ONLINE
0/2067 1h · 0s ago
2067 signals 1h, all skipped - bot alive and filtering correctly (e.g. strict consensus / edge-net thresholds)
AGENT_HOLD_TO_SETTLE 🟢 ONLINE
0/85 1h · 11m ago
85 signals 1h, all skipped - bot alive and filtering correctly (e.g. strict consensus / edge-net thresholds)
AGENT_LAMBDA_TAIL 🟢 ONLINE
0/190 1h · 11m ago
190 signals 1h, all skipped - bot alive and filtering correctly (e.g. strict consensus / edge-net thresholds)
AGENT_NOTHING_HAPPENS 🟢 ONLINE
0/0 1h · -
0 signals 1h but bot heartbeat-fresh - idle (scanner-driven or strict filter); not an outage
AGENT_SPORTS_FLB 🟢 ONLINE
0/0 1h · -
0 signals 1h but bot heartbeat-fresh - idle (scanner-driven or strict filter); not an outage
CLAUDE_CONSENSUS_PROB 🔴 OFFLINE
0/0 1h · -
0 signals last hour AND heartbeat stale - real outage; check the daemon
CLAUDE_DEEP_TAIL 🔴 OFFLINE
0/0 1h · -
0 signals last hour AND heartbeat stale - real outage; check the daemon
CLAUDE_ENSEMBLE 🔴 OFFLINE
0/0 1h · -
0 signals last hour AND heartbeat stale - real outage; check the daemon
CLAUDE_FAST_REACT 🔴 OFFLINE
0/0 1h · -
0 signals last hour AND heartbeat stale - real outage; check the daemon
CLAUDE_FRESH_FLIP 🔴 OFFLINE
0/0 1h · -
0 signals last hour AND heartbeat stale - real outage; check the daemon
CLAUDE_HIGH_TAIL 🔴 OFFLINE
0/0 1h · -
0 signals last hour AND heartbeat stale - real outage; check the daemon
CLAUDE_MAKER 🔴 OFFLINE
0/0 1h · -
0 signals last hour AND heartbeat stale - real outage; check the daemon
CLAUDE_MID_VOLUME 🔴 OFFLINE
0/0 1h · -
0 signals last hour AND heartbeat stale - real outage; check the daemon
CLAUDE_PMXT_ARB 🔴 OFFLINE
0/0 1h · -
0 signals last hour AND heartbeat stale - real outage; check the daemon
CLAUDE_PRE_RESOLVE 🟡 DEGRADED
mothballed · -
MOTHBALLED (starting_balance=0) - by design, e.g. OPT-58 symmetry mothball; not a real outage
CLAUDE_QUALITY_KELLY 🔴 OFFLINE
0/0 1h · -
0 signals last hour AND heartbeat stale - real outage; check the daemon
CLAUDE_RESOLUTION_SNIPER 🔴 OFFLINE
0/0 1h · -
0 signals last hour AND heartbeat stale - real outage; check the daemon
CLAUDE_REVERSAL_HUNTER 🔴 OFFLINE
0/0 1h · -
0 signals last hour AND heartbeat stale - real outage; check the daemon
CLAUDE_SHARP_CONVERGENCE 🔴 OFFLINE
0/0 1h · -
0 signals last hour AND heartbeat stale - real outage; check the daemon
CLAUDE_SHARP_FOLLOW 🔴 OFFLINE
0/0 1h · -
0 signals last hour AND heartbeat stale - real outage; check the daemon
CLAUDE_SHARP_PROVEN 🔴 OFFLINE
0/0 1h · -
0 signals last hour AND heartbeat stale - real outage; check the daemon
CLAUDE_SPORTS_MIRROR 🔴 OFFLINE
0/0 1h · -
0 signals last hour AND heartbeat stale - real outage; check the daemon
CLAUDE_TAIL_DISCIPLINED 🔴 OFFLINE
0/0 1h · -
0 signals last hour AND heartbeat stale - real outage; check the daemon
CLAUDE_VOL_FADE 🔴 OFFLINE
0/0 1h · -
0 signals last hour AND heartbeat stale - real outage; check the daemon
CLAUDE_WHALE_LOOSE 🔴 OFFLINE
0/0 1h · -
0 signals last hour AND heartbeat stale - real outage; check the daemon
CLAUDE_WHALE_PRECISION 🔴 OFFLINE
0/0 1h · -
0 signals last hour AND heartbeat stale - real outage; check the daemon
COPY_BIG_TICKET 🟢 ONLINE
0/198 1h · 36m ago
198 signals 1h, all skipped - bot alive and filtering correctly (e.g. strict consensus / edge-net thresholds)
COPY_CONSENSUS 🟢 ONLINE
0/1905 1h · 0s ago
1905 signals 1h, all skipped - bot alive and filtering correctly (e.g. strict consensus / edge-net thresholds)
COPY_CRYPTO_ELITE 🟢 ONLINE
0/32 1h · 49m ago
32 signals 1h, all skipped - bot alive and filtering correctly (e.g. strict consensus / edge-net thresholds)
COPY_CRYPTO_QUALITY 🟢 ONLINE
0/2067 1h · 0s ago
2067 signals 1h, all skipped - bot alive and filtering correctly (e.g. strict consensus / edge-net thresholds)
COPY_FADE 🟢 ONLINE
0/1 1h · 53m ago
1 signals 1h, all skipped - bot alive and filtering correctly (e.g. strict consensus / edge-net thresholds)
COPY_FAST_QUALITY 🟢 ONLINE
0/2067 1h · 0s ago
2067 signals 1h, all skipped - bot alive and filtering correctly (e.g. strict consensus / edge-net thresholds)
COPY_HIGH_PRICE 🟢 ONLINE
0/2067 1h · 0s ago
2067 signals 1h, all skipped - bot alive and filtering correctly (e.g. strict consensus / edge-net thresholds)
COPY_HOTHAND 🟢 ONLINE
0/0 1h · -
0 signals 1h but bot heartbeat-fresh - idle (scanner-driven or strict filter); not an outage
COPY_HOTHAND_FAST 🟢 ONLINE
0/1 1h · 38m ago
1 signals 1h, all skipped - bot alive and filtering correctly (e.g. strict consensus / edge-net thresholds)
COPY_LATE_RESOLVE 🟢 ONLINE
0/2067 1h · 0s ago
2067 signals 1h, all skipped - bot alive and filtering correctly (e.g. strict consensus / edge-net thresholds)
COPY_LOW_PRICE 🟢 ONLINE
0/2067 1h · 0s ago
2067 signals 1h, all skipped - bot alive and filtering correctly (e.g. strict consensus / edge-net thresholds)
COPY_NEW_LISTING 🟢 ONLINE
0/19 1h · 51m ago
19 signals 1h, all skipped - bot alive and filtering correctly (e.g. strict consensus / edge-net thresholds)
COPY_RWO_BTC 🟢 ONLINE
0/0 1h · -
0 signals 1h but bot heartbeat-fresh - idle (scanner-driven or strict filter); not an outage
COPY_SHARP 🟢 ONLINE
0/0 1h · -
0 signals 1h but bot heartbeat-fresh - idle (scanner-driven or strict filter); not an outage
COPY_SPORTS_QUALITY 🟢 ONLINE
0/2067 1h · 0s ago
2067 signals 1h, all skipped - bot alive and filtering correctly (e.g. strict consensus / edge-net thresholds)
COPY_VOLUME_1K 🟢 ONLINE
0/116 1h · 36m ago
116 signals 1h, all skipped - bot alive and filtering correctly (e.g. strict consensus / edge-net thresholds)
COPY_WHALE 🟢 ONLINE
0/68 1h · 54m ago
68 signals 1h, all skipped - bot alive and filtering correctly (e.g. strict consensus / edge-net thresholds)
INVERSE_0X_0C16 🟢 ONLINE
0/0 1h · -
0 signals 1h but bot heartbeat-fresh - idle (scanner-driven or strict filter); not an outage
INVERSE_SHARKY_BTC 🟢 ONLINE
0/28 1h · 54m ago
28 signals 1h, all skipped - bot alive and filtering correctly (e.g. strict consensus / edge-net thresholds)

Recent auto-killer events

Rules A1-A6 reset, Rule C unpause, manual events - last 15
WhenBotActionReason
20m agoREVERSE_FADE_KELLY🟡 auto_resetwr=0.0% < 35% after 217 bets
20m agoSTRONGEST_SIGNAL_FOLLOWER🟡 auto_resetwr=2.5% < 35% after 510 bets
20m agoMART_BUST_TEST🟡 auto_resetwr=0.0% < 35% after 422 bets
20m agoREGIME_AWARE_BTC🟡 auto_resetwr=8.3% < 35% after 133 bets
20m agoSTREAK_FOLLOW_K4🟡 auto_resetwr=0.0% < 35% after 55 bets
20m agoSTREAK_REV_KELLY🟡 auto_resetwr=4.1% < 35% after 123 bets
20m agoINVERSE_WHALE_FOLLOW_1TO1🟡 auto_resetwr=1.6% < 35% after 448 bets
20m agoINVERSE_WHALE_FOLLOW_V2🟡 auto_resetwr=2.8% < 35% after 142 bets
20m agoMTF_RSI_BB_WIDE_DOWN🟡 auto_resetwr=0.0% < 35% after 30 bets
20m agoQUANTUM_BRAIN🟡 auto_resetwr=2.3% < 35% after 435 bets
20m agoSTREAK_FADE_K4🟡 auto_resetwr=8.9% < 35% after 56 bets
20m agoSTREAK_FOLLOW_K3🟡 auto_resetwr=0.0% < 35% after 121 bets
20m agoSTREAK_FOLLOW_K4_MART🟡 auto_resetwr=0.0% < 35% after 59 bets
20m agoSTREAK_REV_3PLUS🟡 auto_resetwr=4.0% < 35% after 124 bets
20m agoULT1_TIERED_PAROLI🟡 auto_resetwr=4.0% < 35% after 126 bets

Recent trades

Last 15 paper_trades rows - BUY/SELL/RESET activity across the fleet
WhenAccountSideUSDPnLNote
59s agoINVERSE_WHALE_FOLLOW_1TO1🔵 BUY$40.00$+0.00
59s agoINVERSE_WHALE_FOLLOW🔵 BUY$40.00$+0.00
1m agoWHALE_FOLLOW🔵 BUY$40.00$+0.00
1m agoINVERSE_BOOK_FAKE_FADE🔵 BUY$15.00$+0.00
1m agoBOOK_FAKE_FADE🔵 BUY$15.00$+0.00
1m agoWHALE_FOLLOW_CAP55🔵 BUY$40.00$+0.00
1m agoENSEMBLE_VOTE_4_V2🔵 BUY$30.00$+0.00
1m agoSTRONGEST_SIGNAL_FOLLOWER🔵 BUY$30.00$+0.00
2m agoQUANTUM_BRAIN🔵 BUY$112.14$+0.00
2m agoMART_BUST_TEST🔵 BUY$5.00$+0.00
2m agoMART_PURE🔵 BUY$10.00$+0.00
2m agoINVERSE_BINANCE_FRONT_RUN_CAP55🔵 BUY$40.00$+0.00
2m agoBINANCE_FRONT_RUN_CAP55🔵 BUY$40.00$+0.00
5m agoWHALE_FOLLOW🔵 BUY$40.00$+0.00
6m agoSTRONGEST_SIGNAL_FOLLOWER🔵 BUY$30.00$+0.00

Open positions - honest breakdown

Total 16 open paper_positions split by where they are in the resolution lifecycle. FUTURE = market still running (normal). AWAITING = past end_date, Polymarket finalizing. STUCK = past_due >24h - real attention item.
16 FUTURE - market still running
0 AWAITING - PM oracle pending
0 STUCK >24h - needs attention

Bug-fixes log (167 deployed · ★ = G0-readiness blocker · 51 ★)

audit trail of bug fixes applied across all 42 bots/agents · most recent first
IDRoundFixScopeG0?
BUG-21R3Closed positions LIMIT raised 50 → 200copy + claude detail pages
BUG-22R3Per-worker cache divergence - documented + auto-bust via last_reset_atagent detail caches
BUG-23R3Per-bot threading.Lock on cache rebuild - eliminates thundering herdagent detail rebuild path
BUG-24R3Removed wasted calendar_seed/breakdown_seed computesnapshot helper
BUG-25R3Cache key includes last_reset_at - auto-busts on OPT-129 cross-processcopy + claude detail★
BUG-26R3Heatmap n filtered to closed statuses (no open/refunded inflation)copy + claude detail
BUG-27R3Heatmap wins include exit-rule statuses (OPT-54b class miss)copy + claude detail
BUG-28R3agg query: ties_or_refunds surfaced explicitlycopy + claude detail
BUG-29R3Refunded vs won/lost: explicit ties_or_refunds fieldcopy + claude detail
BUG-30R3avg_position_usd uses closed signals only (matches avg_hold_time)fact_sheet
BUG-31R3Sharpe requires ≥5 daily samples - undersampled shows '- (Nd)'fact_sheet
BUG-32R3Profit factor labeled as daily-aggregate (was misleadingly per-trade-implied)fact_sheet
BUG-33R3Refund-like closed rows render '-' instead of fake '0.00c'pos table
BUG-34R3All timestamps now show 'UTC' suffixpos table
BUG-35R3Every numeric cell carries data-sort - sort works correctlypos table
BUG-36R3thesis + name + bot_id HTML-escaped in headerdetail header
BUG-37R3brain_mini telemetry tokens have hover tooltipspos table
BUG-38R3Calendar/breakdown JS has try/catch with in-UI error displaydetail JS
BUG-39R3Mothballed bot detection documented + tag surfacing intact_agent_known
BUG-40R3Heatmap title clarifies 'decides' (decided_at), not fill-timedetail render
BUG-11R2Breakdown WR uses wins/(wins+losses), not wins/eventscopy + claude breakdown★
BUG-12R2Calendar + Breakdown both key on resolved_at (consistent period filter)copy + claude detail★
BUG-13R2Breakdown unrealized fetched live via ws_book (was hardcoded $0)copy + claude breakdown★
BUG-14R2Source-wallet top/bottom lists no longer overlapcopy + claude detail
BUG-15R2Stale-price position cells render '-' (not entry-price)copy + claude detail★
BUG-16R2Alpha score uses honest daily-PnL series (not dishonest cached ROI)fact_sheet★
BUG-17R2Strategy tag derived from cfg.pool_builder (not blanket 'Sharp Follower')fact_sheet tags
BUG-18R2fmtUsd emits '-$100' (sign before dollar)detail JS
BUG-19R2HTML-escape + URL-encode position table datapos table
BUG-20R2'8s' stale hint replaced with accurate cadence infodetail render
BUG-1R1Bankroll + ROI subtract OPT-129 reset_credits (math invariant honest)leaderboard + detail★
BUG-2R1Detail page price fetch uses ws_book Redis fast pathdetail snapshot
BUG-3R1Failed price fetch shows '-' (not fabricated breakeven)detail snapshot★
BUG-4R1Peer rank uses honest ROI (paper_positions, not cached state)peer_rank★
BUG-5R1Stale '10 copy/brain + 1 BTC' subtitle replaced with dynamic countpeer_rank header
BUG-6R1Entry timing N/A for price-pinned bots (HIGH_PRICE, LOW_PRICE)behavioral
BUG-7R1Long/Short bias only counts binary markets (yes/no)behavioral
BUG-8R1Source-wallet W/L includes exit-rule wins (OPT-54b class)source_wallet_lb
BUG-9R1True median for even-N samples (peer rank + position size)behavioral + peer_rank
BUG-10R1Close-method table split: exit-rule vs natural-resolutionexit_rules render
BUG-LB-21leaderboardCOPYBOT_JS wrapped in <script> tags (was leaking as text)/copy-bots★
BUG-LB-22leaderboardRealized/Unrealized cells gain color via td-level class/copy-bots
BUG-C-1claude-parityClaude detail: Bankroll/ROI subtract reset_credits/claude-bots/<b>★
BUG-C-2claude-parityClaude detail: ws_book Redis fast path + stale-price tracking/claude-bots/<b>★
BUG-C-3claude-parityClaude detail: stale-price banner + zero-PnL guard/claude-bots/<b>★
BUG-C-4claude-parityClaude peer rank uses honest paper_positions ROI (not cached)/claude-bots/<b>★
BUG-C-6claude-parityClaude behavioral: entry_timing N/A for price-pinned bots/claude-bots/<b>
BUG-C-7claude-parityClaude behavioral: long_pct only over binary markets/claude-bots/<b>
BUG-C-8claude-parityClaude source-wallet W/L includes exit-rule wins/claude-bots/<b>
BUG-C-9claude-parityClaude peer rank + behavioral: true median for even-N/claude-bots/<b>
BUG-C-11claude-parityClaude breakdown WR = wins/(wins+losses)/claude-bots/<b>★
BUG-C-14claude-parityClaude source-wallet top/bottom lists no longer overlap/claude-bots/<b>
BUG-C-16claude-parityClaude tags use honest alpha_score (not cached ROI)/claude-bots/<b>★
BUG-C-21claude-parityClaude detail: closed positions LIMIT raised 50 → 200/claude-bots/<b>
BUG-C-26claude-parityClaude heatmap n filtered to closed statuses/claude-bots/<b>
BUG-C-27claude-parityClaude heatmap wins include exit-rule statuses/claude-bots/<b>
BUG-C-28claude-parityClaude agg query: ties_or_refunds surfaced explicitly/claude-bots/<b>
BUG-C-30claude-parityClaude fact-sheet avg_position_usd: closed signals only/claude-bots/<b>
BUG-REG-1regression-fixPostgres LIKE '%' literal: parameterize so psycopg shim parses cleanly/claude-bots/<b>/peer_rank★
BUG-BTC-1btc-updownBTC account WR uses won/(won+lost), not won/n_closed (ties broke the formula)/btc-updown scoreboard★
BUG-BTC-2btc-updownBTC account ties_or_refunds surfaced (separate from won/lost)/btc-updown scoreboard
BUG-BTC-3btc-updownBTC last-bets timestamps now have explicit 'UTC' label/btc-updown track
BUG-C-LB-1claude-leaderboarddata-sort on every numeric cell (was mis-sorting '$-50' as +50)/claude-bots★
BUG-C-LB-2claude-leaderboardBankroll/Cash/Open$ unified to .2f precision (was .0f, broke reconciliation)/claude-bots
BUG-C-LB-3claude-leaderboardDot tooltip 'never seen' for ls=0 (was '1.76e9s ago')/claude-bots
BUG-C-LB-4claude-leaderboardStubbed counter now uses claude_agents.is_stubbed() (heartbeat had no 'stubbed' column)/claude-bots health bar★
BUG-C-LB-5claude-leaderboardTable + <th> gain 'sortable' class for /copy-bots parity/claude-bots
BUG-C-LB-6claude-leaderboardstub_overlay env_var HTML-escaped (defense-in-depth)/claude-bots
BUG-C-LB-7claude-leaderboardPaused dot tooltip surfaces last-seen info (was just 'paused')/claude-bots
BUG-C-LB-8claude-leaderboardMothballed bots no longer inflate 'dead' counter in health bar/claude-bots health bar★
BUG-C-LB-9claude-leaderboardW/L column data-sort=n_resolved (was lexical '10/0' < '2/0')/claude-bots
BUG-C-LB-10claude-leaderboardMothballed detection threshold simplified - starting_bankroll only (was also requiring tv≤0.0001 which residual cash broke)/claude-bots★
BUG-POPOVER-CLIPcross-page(?) popover position:fixed + JS-positioned so it escapes .card overflow clipping (was: bottom half hidden behind table frame)/copy-bots + /claude-bots leaderboards
BUG-C-LB-11claude-leaderboard-r2Hero card monetary values unified to .2f precision (was .0f - broke reconciliation with .2f rows)/claude-bots hero
BUG-C-LB-12claude-leaderboard-r2cb_total_agents queries paper_account.starting_balance instead of nonexistent cfg field (was inflating denominator to 22)/claude-bots hero★
BUG-C-LB-13claude-leaderboard-r2stale_sec_c clamped at 0 with skew tooltip (was rendering -43s as green pos)/claude-bots
BUG-C-LB-14claude-leaderboard-r2Hero 'Active agents' pill annotated with what's excluded (paused/mothballed/stubbed)/claude-bots hero
BUG-C-LB-15claude-leaderboard-r2_live_mtm_for_bots returns live realized_pnl_usd from paper_positions - hero + leaderboard agree to the centleaderboard + hero★
BUG-C-LB-16claude-leaderboard-r2Agent name HTML-escaped on leaderboard rows + mothballed branch/claude-bots
BUG-C-LB-17claude-leaderboard-r2Dropped dead pointer_html construction (built but not returned)/claude-bots
BUG-C-LB-18claude-leaderboard-r2Dropped dead thesis_html='' placeholder/claude-bots
BUG-C-LB-20claude-leaderboard-r2Dropped dead feed_columns + skips_html builds (~330+ wasted string ops per snapshot)/claude-bots
BUG-C-LB-19claude-leaderboard-r2[partial] _strategy_meta_string is_arb check broken on Claude (cfg lacks bot_id) - documented; structural fix deferred since it requires API signature change/claude-bots popover
BUG-C-LB-21claude-leaderboard-r3Hero winner requires ≥0.05pp gap (was flipping on 0.001% noise)/claude-bots hero
BUG-C-LB-22claude-leaderboard-r3threading.Lock on _CLAUDE_SNAP_CACHE rebuild - eliminates thundering herdsnapshot cache
BUG-C-LB-23claude-leaderboard-r3Per-bot brain_stats consolidated to single closed-only query (~88 fewer queries per rebuild)snapshot rebuild
BUG-C-LB-24claude-leaderboard-r3Calibration threshold ±1.0pp named CAL_EDGE_SIGNIFICANT_PP + MIN_SAMPLES=10 documented/claude-bots
BUG-C-LB-25claude-leaderboard-r3bs.n_closed includes ties_or_refunds field consistent with detail page closed_totalbrain_stats
BUG-C-LB-26claude-leaderboard-r3avg_edge_net filters to CLOSED signals only (no open/refunded contamination)brain_stats★
BUG-C-LB-27claude-leaderboard-r3avg_paid=None defers cal_edge (was defaulting to 0.5 → wrong for price-stratified bots)brain_stats★
BUG-C-LB-28claude-leaderboard-r3CLAUDE_BTC_5M synthetic heartbeat surfaces real eval count from btcup_bets/claude-bots health bar
BUG-C-LB-29claude-leaderboard-r3CLAUDE_BTC_5M last_seen=0 when no activity (was 'now' fallback hiding dead state)/claude-bots health bar★
BUG-C-LB-30claude-leaderboard-r3Synthetic CLAUDE_BTC_5M heartbeat re-injects when real one is >2h stale/claude-bots
BUG-C-LB-31claude-leaderboard-r3Dropped unused bucket_dist + category_dist + skip_reasons per-bot queries (rolled into #C-LB-23)snapshot
BUG-C-LB-32claude-leaderboard-r3pool_sizes cached separately with 5-min TTL (was scanning every 60s rebuild)snapshot
BUG-C-LB-33claude-leaderboard-r3Hero default user_total / cb_total use registry len() not stale '10'/claude-bots hero
BUG-C-LB-34claude-leaderboard-r3Symmetric pause exclusion - cb_active now also filters paused Claude agents/claude-bots hero★
BUG-C-LB-35claude-leaderboard-r3Hero hint clarified - 'paused + mothballed agents excluded from BOTH sides'/claude-bots hero
BUG-C-LB-36claude-leaderboard-r3Hero 'Positions / W-L' (not 'Trades') - clarifies that count includes opens, W-L only closed/claude-bots hero
BUG-C-LB-37claude-leaderboard-r3Dropped unused agg.n COUNT(*) (rolled into #C-LB-23 consolidation)brain_stats
BUG-C-LB-38claude-leaderboard-r3is_stubbed() set up to evaluate once per row pass (memoization deferred to runtime)/claude-bots
BUG-C-LB-39claude-leaderboard-r3Defensive b.get('bot_id') (KeyError-skip instead of 500 on malformed row)/claude-bots
BUG-C-LB-40claude-leaderboard-r3now_ts staleness during long render documented in code (compounds with 60s cache)/claude-bots
BUG-BTC-4btc-updown-r2Defensive recommendation lookup - unknown rec falls back to HOLD styling instead of KeyError 500/btc-updown hero★
BUG-BTC-5btc-updown-r2Defensive market dict .get() - partial/malformed gamma-api responses no longer 500 the page/btc-updown hero★
BUG-BTC-6btc-updown-r2Market question HTML-escaped - quote/'<' in title no longer breaks layout (defense-in-depth)/btc-updown hero
BUG-BTC-7btc-updown-r2Resolution timestamp includes date when >1h out (was 'HH:MM:SS' only, ambiguous across days)/btc-updown hero
BUG-BTC-8btc-updown-r2Holders table: name HTML-escaped; full name in title attr for hover/btc-updown holders
BUG-BTC-9btc-updown-r2Holders table: data-sort attrs on Shares/Avg/Value/PnL - numeric sort works/btc-updown holders
BUG-BTC-10btc-updown-r2Holders PnL precision .0f → .2f - cent-level resolution restored/btc-updown holders
BUG-BTC-11btc-updown-r2Holders long names truncate with ellipsis (was silently cut at 18 chars)/btc-updown holders
BUG-BTC-12btc-updown-r2Track-record row: defensive .get() on side/entry_price/bet_usd - legacy rows no longer 500/btc-updown track★
BUG-BTC-13btc-updown-r2Track-record numeric cells gain data-sort attrs - When/Stake/Score/Conf/PnL all sort correctly/btc-updown track
BUG-BTC-14btc-updown-r3_btcup_snapshot wraps every gamma-api call in try/except - upstream timeouts serve stale cache instead of 500-ing the page/btc-updown snapshot★
BUG-BTC-15btc-updown-r3threading.Lock on _BTCUP_SNAP_CACHE rebuild - eliminates duplicate gamma-api fan-out under gunicorn fork/btc-updown snapshot
BUG-BTC-16btc-updown-r3btc_updown_page route handler wraps each sub-render in try/except - one section's failure degrades to error card instead of 500-ing the entire page/btc-updown route★
BUG-BTC-17btc-updown-r3Scoreboard bare except now logs to stderr - silent account-vanish bug surfaces in journald/btc-updown scoreboard
BUG-BTC-18btc-updown-r3Per-account bankroll precision unified (.2f vs .0f mismatch) - cent-level reconciliation with leaderboard/btc-updown scoreboard
BUG-BTC-19btc-updown-r3Per-account rPnL pill .0f → .2f - drops <$1 rounding error vs leaderboard sum/btc-updown scoreboard
BUG-BTC-20btc-updown-r3Strip title attr escapes side+status (DB-sourced) - quote in either no longer breaks the attribute/btc-updown scoreboard
BUG-BTC-21btc-updown-r3Brain recent-bets `when` column adds 'UTC' suffix - matches track-record table above/btc-updown brain
BUG-BTC-22btc-updown-r3Brain recent-bets data-sort attrs on When/Size/Signal/BTC Δ/PnL - numeric sort no longer lexical/btc-updown brain
BUG-BTC-23btc-updown-r3Brain recent-bets HTML-escape side/acct_name/outcome - defense-in-depth + unknown-status badge renders literally instead of falsely tagged 'open'/btc-updown brain
BUG-SORT-1fleet-sortUniversal table sort: data-sort attribute is now SOURCE OF TRUTH - was being ignored because _detectNumericColumn's regex didn't match '$+123.45' (sign between $ and digits from :+,.2f), so PnL/ROI/Realized columns fell back to text sort and ended up sorting alphabetically by name across every pageBASE_JS / all sortable tables★
BUG-BTC-24btc-updown-r4_render_btcup_patterns: defensive .get() on n_up/n_down/pct_up/current_streak/max_up_streak/max_down_streak - partial pattern dicts no longer 500 the section/btc-updown patterns★
BUG-BTC-25btc-updown-r4_render_btcup_patterns: p['seq'][-100:] now defensively defaults to [] when missing or None/btc-updown patterns★
BUG-BTC-26btc-updown-r4_render_btcup_patterns: run_dist/conditional/hour_dist .get() guard - AttributeError on missing parent keys eliminated/btc-updown patterns★
BUG-BTC-27btc-updown-r4cur_outcome else-branch no longer defaults to 🔴 - neutral '•' for unknown direction (was misleadingly red)/btc-updown patterns
BUG-BTC-28btc-updown-r4Sequence dot title attribute HTML-escapes the outcome value (defense-in-depth)/btc-updown patterns
BUG-BTC-29btc-updown-r4Run-length distribution table capped at 20 rows - last row aggregates all longer streaks (was unbounded; outlier N=50 streak rendered a 50-row table)/btc-updown patterns
BUG-BTC-30btc-updown-r4_bias_cls hoisted out of the K loop - was being redefined 6× per render/btc-updown patterns
BUG-BTC-31btc-updown-r4_render_btcup_brain: defensive .get() on won/lost/total_pnl/roi_pct/avg_pnl_per_bet/stake/open - partial brain dicts (e.g. mid-warm-up) no longer 500/btc-updown brain★
BUG-BTC-32btc-updown-r4Brain breakdown row helpers (_bk_row callers) - every bucket dict accessed via .get() with safe defaults; single bad row no longer 500s the whole brain section/btc-updown brain★
BUG-BTC-33btc-updown-r4Hourly heatmap cells: defensive .get() on wr/n/pnl/hour + isinstance guard - missing keys/wrong types no longer 500/btc-updown brain★
BUG-BTC-34btc-updown-r4streak_str streak_n==1 branch - was emitting trailing ' ·' with no value ('win ·'); now reads '1W (last bet)' / '1L (last bet)' / '-'/btc-updown brain
BUG-BTC-35btc-updown-r4Brain recent-bets Outcome col data-sort attribute - Up/Down groups correctly under our sort fix/btc-updown brain
BUG-BTC-36btc-updown-r4Bet strip: rz==0 (tie/refund/breakeven) now renders as 'bt tie' neutral state - was being shown red as 'bt lost' (false-loss visual)/btc-updown scoreboard
BUG-BTC-37btc-updown-r4Bet strip size precision: ${int(size)} dropped cents - $5.50 displayed as $5; now uses $5 for whole-dollar, $5.50 otherwise/btc-updown scoreboard
BUG-BTC-38btc-updown-r4_fetch_btc_account_summary LEFT JOIN dup-row fix: pp <-> btcup_bets relation no longer multiplies when multiple bets reference one paper_position (correlated MAX(bb.id) subquery)/btc-updown scoreboard★
BUG-BTC-39btc-updown-r4_render_mart_ladder: label now annotates 'CAPPED' when state.step exceeds the ladder length - operator can see terminal-exposure state/btc-updown scoreboard
BUG-BTC-40btc-updown-r4Hero 'Bet eligible?' sub-line: 'in_time_window=None · in_price_window=None' replaced with ✓/✗/- glyphs (was printing 'None' literal when keys missing)/btc-updown hero
BUG-BTC-41btc-updown-r4Track-record Side col data-sort attribute - Up/Down groups correctly/btc-updown track
BUG-BTC-42btc-updown-r4Track-record Status col data-sort uses rank (won=2/open=1/lost=0/refund=-1) - sort by outcome rank, not alphabetical 'L→O→R→W'/btc-updown track
BUG-BTC-43btc-updown-r4BTCUP_JS auto-reload: column-sort state now persists across 10s refresh via sessionStorage - was being wiped every reload, making sustained sort impossible/btc-updown JS★
BUG-LB-MATH-1fleet-mathBankroll != starting + Realized + Unrealized: paper_trading.buy() debits cash by (cost_basis+fee) but cost_basis stored is price-only, so the buy fee leaks from cash but never appears in realized_pnl_usd. Fix: derive Realized from the bookkeeping identity (honest_total - starting - unrealized) so the visible math is self-consistent on EVERY row. Tooltip surfaces the trade-ideal vs true realized diff per bot./copy-bots + /claude-bots leaderboards + hero★
BUG-LB-MATH-2fleet-mathHero card was still summing naive realized: live-MTM refresh in _claude_hero_comparison only overwrote total_value+unrealized on the user side (and didn't run at all on the Claude fallback path), so hero Realized/ROI used stale copy_bot_state/claude_bot_state cached values. Now both paths overwrite realized_pnl_usd with the bookkeeping-honest value - head-to-head math reconciles to the cent./claude-bots hero card★
BUG-LB-MATH-3fleet-mathPer-agent detail page (/copy-bots/<id> + /claude-bots/<id>): Realized PnL KPI cell was reading naive totals.total_realized_pnl from paper_positions sum. Switched to bookkeeping-derived Realized = total_value - starting - unrealized so the header KPIs add up exactly. Fee-drag tooltip on the Realized cell shows trade-ideal vs true realized diff./copy-bots/<id> + /claude-bots/<id> detail headers★
BUG-THEME-1fleet-theme/claude-bots/<id> detail page was missing the entire AGENT_DETAIL_CSS block AND the <div class='agent-detail'> wrapper that /copy-bots/<id> used - so all 22 Claude detail pages rendered with default browser styling (no fact-cell grid, no peer-card, no hour×day heatmap layout, no uniform vertical rhythm). Hoisted the CSS to a module-level constant; both renderers now reference it. All 42 bot detail pages now share an identical theme./copy-bots/<id> + /claude-bots/<id>
BUG-STATUS-1status-falsepospolybot-claude-mart was being flagged DEAD whenever no MART variant had placed a bet in 30 minutes (the daemon only writes martingale_state on bet placement, not every cycle, so quiet BTC signals = false-positive offline). Now ORs three liveness signals: martingale_state.updated_at, MAX(MART_* paper_positions.opened_at), and the parent BTC daemon's freshness (mart is a passive consumer - if BTC is alive, mart is too)./status daemons
BUG-STATUS-2status-falseposMothballed bots (starting_balance=0, e.g. CLAUDE_PRE_RESOLVE post-OPT-58) and stubbed bots (waiting for env var) were being counted as OFFLINE in the signal-flow check, inflating the offline rollup with false positives. Now they're surfaced as DEGRADED with a clear 'MOTHBALLED' / 'STUBBED' reason so the operator can distinguish intentional idle from real outages./status signals
OPT-225btc-arenaBTC strategy arena reset to 6 strategies: MART_PURE ($10k, $5→$5120 11-step doubling, busts → reset) + 5 fighters on $5k each (ANTI_MART paroli, STREAK_FADE/RIDE, HOLDER_PRO/CON). Three natural opposing pairs. Outcome history extended 1k→10k. Daemon now supports per-strategy should_bet/side/size hooks. Self-improvement framework: rolling stats logged for manual nudges (no auto-changes yet)./btc-updown + martingale_strats + claude_btc_martingale_bot
BUG-DET-41detail-leaderboard-r5_agent_peer_rank used `SUM(pp.shares * pp.avg_entry_price - pp.cost_basis)` for the unrealized contribution - but cost_basis IS shares×avg_entry by definition (paper_trading.buy() stores it that way), so the term was ALWAYS 0. Peer ranking silently ignored every bot's unrealized PnL, biasing the order. Now reads via _live_mtm_for_bots (same source as leaderboard + hero)./copy-bots/<id> + /claude-bots/<id> peer rank★
BUG-DET-42detail-leaderboard-r5_agent_exit_rules was running 7 separate COUNT/SUM queries (one per status), now collapsed to single GROUP BY - ~10× faster on the index. Initializes zeros for missing statuses so the renderer iterates the full set without KeyError./copy-bots/<id> + /claude-bots/<id> exit-rule card
BUG-DET-43detail-leaderboard-r5_enrich_btc returned a position dict missing the `price_stale` field that _enrich (copy side) populates - renderer worked via .get default but the shape inconsistency was a latent bug for any future code that does `is False` check./copy-bots/BTC_5M_BOT detail
BUG-DET-44detail-leaderboard-r5extra_js bot_id was interpolated as `const BOT_ID = "{bot_id}";` - quote/newline in bot_id would break script parsing and silently disable calendar/breakdown loads. Now JSON-encoded via json.dumps so the value is always a valid JS string literal. Defense-in-depth on both copy + claude detail pages./copy-bots/<id> + /claude-bots/<id> JS
BUG-CLB-41detail-leaderboard-r5_live_mtm_for_bots open-MTM loop used bare `except Exception: pass` on float parse - a single malformed cost_basis/shares row understated bankroll silently. Now logs parse failures to stderr (journald-visible) with the row id + raw values for forensics./copy-bots + /claude-bots leaderboards + hero★
BUG-CLB-42detail-leaderboard-r5_live_mtm_for_bots accepted mid=0.0 as a valid live price - prediction-market mids of exactly 0.0 are typically resolved-side-empty books, not real prices, and using them as MTM would compute shares*0=0 open_value, falsely inflating unrealized losses. Now requires mid > 0 to use as live price; 0.0 falls through to HTTP fallback._live_mtm_for_bots★
BUG-ARENA-1btc-arena-fixResolution loop's `WHERE a.name LIKE 'MART_%'` pattern missed the 5 new arena strategies (ANTI_MART, STREAK_FADE, STREAK_RIDE, HOLDER_PRO, HOLDER_CON) - their bets stayed `open` forever, state never advanced, strategies looked frozen. Now uses explicit IN-list from ms.ACCOUNT_NAMES so every arena bot resolves on the 60s sweep.claude_btc_martingale_bot._resolve_mart_bets★
BUG-ARENA-2btc-arena-fixMART_PURE side selection was static (always followed base signal recommendation). Per user spec: 'Up/Down should be based on the last users who bought live, don't stick to one side'. Now dynamic per-bet: flip after 2 consecutive same-side losses, else follow holder skew (|skew|>0.2), else follow BTC delta (|Δ|>0.05%), else fall back to base signal. State tracks last_side + last_losses_same_side for the flip rule.martingale_strats.MART_PURE.side()
BUG-DEEP-1deep-audit-r6Calendar cell title attribute interpolated `d.date`/`d.n_trades` raw into HTML - defense-in-depth XSS-adjacent if API ever returns special chars. Added _esc() helper, escape every value bound into innerHTML/title across both calendar JS implementations (copy + claude)./copy-bots/<id> + /claude-bots/<id> calendar JS
BUG-DEEP-2deep-audit-r6Race condition on rapid period/category clicks - without abort, multiple in-flight fetches race and the LAST-completed (which may be the FIRST issued) wins, leaving stale data. Added AbortController per loader (loadAgCal / loadAgBreakdown) that cancels prior in-flight request before issuing new one. AbortError silently ignored./copy-bots/<id> + /claude-bots/<id> calendar/breakdown JS
BUG-DEEP-3deep-audit-r6/claude-bots/<id> calendar+breakdown JS had NO try/catch on fetch (only copy side had it after BUG-38). Failed loads on Claude detail pages silently froze the widgets. Now both pages share the same error-handling pattern: fetch failures show an in-UI error card with the HTTP status / error message./claude-bots/<id> calendar/breakdown JS★
BUG-DEEP-BTC-1deep-audit-r6btc_updown.aggregate_holders_by_side used `or` fallback chain: `h.get('totalPnl') or h.get('realizedPnl') or 0`. A holder with exactly 0.0 totalPnl (legit breakeven) silently fell through to realizedPnl - wrong value used in the holder_skew signal. Same shape on value/size/avgPrice. Now uses explicit `is not None` coalesce so 0.0 stays 0.0.btc_updown.aggregate_holders_by_side★
BUG-DEEP-BTC-2deep-audit-r6Holders with unknown outcomeIndex (not 0 or 1) were silently dropped - could indicate API schema drift or malformed response. Now logged to stderr with a count + returned as n_unknown so the operator can spot when upstream payload shape changes.btc_updown.aggregate_holders_by_side
BUG-DEEP-FACT-1deep-audit-r6Trader fact-sheet `biggest_win_usd`/`biggest_loss_usd` defaulted to 0.0 - for a bot with 0 wins/losses, the cell rendered '$+0.00' which falsely implies the bot's best ever trade was a breakeven. Now None when no data, renders '-' instead. Both detail pages updated./copy-bots/<id> + /claude-bots/<id> fact sheet
BUG-DEEP-SW-1deep-audit-r6Source-wallet leaderboard WR shown for any sample size - a wallet with 1W 0L rendered '100%' which mis-ranked it as a top performer. Now applies WR_MIN_SAMPLE=5 (same threshold used elsewhere): under-sampled wallets get wr=None → renderer shows '-'. wr_n exposed for tooltip context./copy-bots/<id> + /claude-bots/<id> source-wallet LB
BUG-DEEP-POS-1deep-audit-r6Position table price-bucket badge interpolated raw bucket string into HTML - defense-in-depth XSS vector if copy_bot_signals.price_bucket ever takes user-controlled input. Now html.escape(bucket) before render. Visual unchanged./copy-bots/<id> + /claude-bots/<id> position table