§1.1 — Trading

PnL Dashboard

Live PnL reporting dashboard for the Kalshi weather algo trading

5/16

Trades Won

+26.3%

Realized, Live Since Jul 2026

-3.0%

Max Drawdown

Trading

live trade ledger

2026-09-2472–73°actual 66° — lost · fv 19.9% · size 1.66%-0.42%

2026-09-2371–72°actual 67° — lost-1.10%

2026-09-2267–68°actual 68° — won+5.93%

2026-09-2172–73°actual 72° — won+17.02%

2026-09-2073–74°actual 67° — lost-0.57%

›show 11 earlier trades

2026-09-1973–74°actual 69° — lost-0.07%

2026-09-18<80°actual 80° — lost-0.52%

2026-09-1786–87°actual 82° — lost-0.07%

2026-09-1574–75°actual 72° — lost-0.22%

2026-09-14<74°actual 75° — lost-0.93%

2026-09-12<75°actual 78° — lost-0.07%

2026-09-10<85°actual 84° — won+1.14%

2026-08-0483–84°actual 84° — won+2.41%

2026-07-2783–84°actual 83° — won+4.50%

2026-07-2582–83°actual 81° — lost-0.47%

2026-07-2481–82°actual 83° — lost-0.23%

5/16 won+26.34% realized (gross, pre-fee)

›notes

Max drawdown -3.02%, 2026-09-10 → 2026-09-20 — measured against bankroll at the time of that peak. Rows shown as % of current bankroll, not dollars, to keep account size private.

One more trade (Sept 26) is still open and not yet in this tally.

fv/size (fair value and applied Kelly stake, at decision time) only show from Sept 24 onward — that's when this got a durable DB column instead of living only in a rotating text log; earlier trades predate it and aren't backfilled.

›backtest comparison (bootstrapped)

2021-08-06 to 2026-05-16 (567 trades), bootstrapped 2000x — red is live, against the p05–p95 range.

Sharpe (annualized)live: 3.34

p05 0.42median 1.12p95 1.66

Max drawdownlive: -3.0%

p05 -49.4%median -31.3%p95 -20.5%

CAGRlive: +300.8%

p05 +9.0%median +43.7%p95 +88.4%

win rate 31.7% · total return +461.5% · positive over 99% of resampled histories

›why bootstrap, and why live looks extreme

Why bootstrap instead of trusting the single backtest run: with only 567 trades, the one realized sequence of wins and losses is just one draw from a much wider range of outcomes. Each of the 2000 resamples keeps the real calendar pattern of which days had a trade, but reshuffles — with replacement — which historical outcome lands on each of those days. That randomizes the sequence Sharpe and max drawdown are sensitive to, while leaving how often the strategy actually traded untouched. The point estimate (accent line) is what really happened; the band is the honest range around it.

The live markers are only 65 calendar days / 16 trades — CAGR and Sharpe get exaggerated by annualizing that short a window. Reference only, not a claim these rates hold over a year.

the model behind these trades → Kalshi Weather Model

← back to Trading