Live Trading Dashboard

updated 2026-08-05 12:22 PM PDT · auto-refreshes every 30 min

Data integrity (persistence canary)

45.8d
history span
15,032
distinct markets
312,611
resolved rows
509,249
total rows

Earliest record: 2026-06-20 05:00 PM PDT. Should stay pinned for the whole run. If it jumps forward, the /data volume reset and history was lost.

Cumulative performance

$-6,965
total net gain/loss across all closed trades
2704
settled
27%
win rate
1
open

Daily performance (PST)

daytradeswin%P&L
Wed Aug 052931%$-7
Tue Aug 041638%+$1
Mon Aug 031833%$-58
Sun Aug 023441%+$30
Sat Aug 013132%+$27
Fri Jul 314639%$-1
Thu Jul 30617%$-1
Wed Jul 298132%$-61
Tue Jul 287548%+$75
Mon Jul 274238%$-70
Sun Jul 265340%+$67
Sat Jul 252733%+$53

Gate effectiveness (counterfactual, $100/mkt equal-weight)

25%
accepted win (n=997)
$-3
accepted $/mkt
21%
rejected win (n=8132)
$-60
rejected $/mkt

Rejected avg $-60/mkt vs accepted $-3/mkt.

gate blockedmktswin%$/mkttotal $
gate_3_probability_band476617%$-66$-314,740
gate_0_daily_limits287722%$-55$-158,614
gate_2_strike_selection43560%$-41$-17,885
gate_6_model_agreement3972%$+100$+3,919
gate_9c_per_market_notional1267%$+78$+940
gate_4_edge_threshold333%$+104$+311

$/mkt = equal-weight $100 stake per market at its ask, net of Kalshi fee. This is the edge test — win% alone hides whether rejected winners are fairly-priced favorites (no edge).

Recent transactions

time (PST)strikesideqtybuyresultP&L
08/05 12:05PM64899.99yes838¢open
08/05 11:39AM64699.99yes1030¢WON+$3
08/05 11:31AM64699.99yes342¢lost$-1
08/05 11:31AM64699.99yes342¢lost$-1
08/05 08:33AM64499.99yes1142¢WON+$2
08/05 08:29AM64599.99yes833¢lost$-1
08/05 08:16AM64699.99yes834¢lost$-1
08/05 07:30AM64499.99yes628¢lost$-1
08/05 07:30AM64499.99yes829¢lost$-1
08/05 07:22AM64599.99yes229¢lost$-0
08/05 07:22AM64599.99yes229¢lost$-0
08/05 07:22AM64599.99yes329¢lost$-0
08/05 07:08AM64399.99yes534¢WON+$1
08/05 07:08AM64399.99yes240¢WON+$0
08/05 07:08AM64399.99yes240¢WON+$0
08/05 05:49AM64399.99yes1042¢lost$-2
08/05 05:45AM64499.99yes327¢lost$-1
08/05 05:45AM64499.99yes329¢lost$-1
08/05 05:45AM64499.99yes429¢lost$-1
08/05 05:19AM64199.99yes243¢WON+$1
08/05 04:27AM64099.99yes535¢lost$-1
08/05 04:22AM64099.99yes338¢lost$-1
08/05 04:16AM64099.99yes440¢lost$-1
08/05 03:42AM64199.99yes732¢lost$-1
08/05 03:30AM64199.99yes730¢lost$-1
08/05 03:30AM64199.99yes734¢lost$-1
08/05 02:22AM64099.99yes447¢WON+$1
08/05 02:22AM64099.99yes646¢WON+$1
08/05 01:28AM64199.99yes345¢WON+$0
08/05 12:17AM64199.99yes843¢lost$-1

Buy = actual entry fill price. Result = outcome at close (WON if net P&L > 0); positions may be closed early by exit rules (model-flip, hard-time, session-salvage), not only at resolution.

Calibration optimizer

Live read (informational, recomputed now). Formal recommendation is weekly & change-capped; you apply it manually.
windownow κrec κstatus
0-10min2.01.9adopt
10-20min2.01.9adopt
20-30min2.01.9adopt
30-45min2.01.9adopt
45-60min2.01.9adopt

Last formal run: 2026-07-05T20:09

Notes

Live fills are REAL (subject to slippage and partial fills) — not the paper sim's optimistic full-fill-at-ask. Gate counterfactual and optimizer use one row per market (deduped) from candidates.db, labelled by resolve_due. Win% on closed trades is P&L-based (net > 0) since positions may close early. Calibration quality does not imply profit: a well-calibrated model on a near-efficient market trades rarely and roughly breaks even.