🌡️ Highest-Temperature Betting Helper

What's changed — and, more to the point, what each change buys you. → marks the payoff.

The idea in one line

Seven independent weather models forecast tomorrow's high; we compare that to what Polymarket charges for each temperature bucket, and surface the bets where the models and the market disagree — with an honest read on when that gap is opportunity and when it's a trap.

Today · 13 Jul 2026

🎯 Sharper bets

• Accuracy by region NEW — a new table scores each model's real average miss across US / Europe / Asia. → see who to trust where: GFS wins the US, ICON wins Europe. yr is now logged so it joins this over time.
• Real ensemble probabilities NEW — the win% now comes from 100+ actual forecast runs (how many land in each bucket), not a bell-curve around the median. → honest odds on the lopsided days a smooth curve gets wrong — exactly the days the market misprices.
• Bias correction per city — the raw forecast is nudged by how far the models have actually missed at that exact airport. → you stop walking into the same trap twice: cold-biased cities like Seoul & Shanghai get pulled toward reality instead of quietly losing.
• Honest win-odds — the win% now uses each city's real spread of past misses, not the models' over-confident agreement. → a "94%" you can actually trust, not a number that just looks safe.
• Net edge — every edge is shown after a fee/slippage cost. → the number that survives the real market; no more chasing a "+5%" that costs 6% to get into.
• ¼-Kelly stake — a bankroll-aware bet size. → a concrete "how much" that grows the pot without one bad day wiping it out.
• Min. profit filter — hide everything below your return threshold. → a one-click shortlist of only the bets worth the risk.

👓 Easier to read

• Wunderground resolution source NEW — the recorded true-high now comes from Wunderground, the exact source Polymarket names in its resolution, with IEM METAR as cross-check. → the track record is scored against precisely what settles your bets.
• History as tabs NEW — Track record & True history are one tabbed panel now, with a colour-coded bias-trend bar per day (red = ran warm, blue = cold, green = stable) and a "settled days X/10" guard on the running P&L. → less scrolling; you see at a glance whether a place ran hot or cold. Numbers in the picks table are mono/aligned.
• Theme switcher NEW — ☾ Dark · ◐ System · ☀ Light in the top bar (remembers your choice; System follows your OS). → read it however suits your screen and the time of day.
• Full visual redesign NEW — new dark palette, Space Grotesk + IBM Plex fonts, mono tabular numbers, a sticky live-status header, one control panel with segmented Day (now up to 5 days ahead) & margin, city toggle-chips, and the Top 5 as cards. → faster to scan, works on mobile, and you can look further ahead.
• ? help icons + this Guide — every term and number in plain language. → anyone you share it with can read it, no meteorology or trading background needed.
• Visual refresh — one font system, aligned (tabular) numbers, calmer tables. → faster scanning and fewer misreads when the numbers matter.
• Risk text now grows instead of cutting off, and Select-all / None are real buttons. → the full warning every time, and quicker city selection.

🛡️ More trustworthy

• Calibration scorecard NEW — a new section scores every settled bet: for each price you paid, how often it actually won, plus a Brier score. → the honest answer to "is the edge real?" — realised win-rate vs price, not theory.
• "Too early to conclude" guard — the daily logger refuses to declare profit until ≥10 settled days. → guards against the most common way these systems fool you: a lucky day or two that looks like skill.
• Cleaner codebase — one shared source for the city list & parsing. → fewer silent bugs when something changes.

Where this is going — and why it matters

• ✅ Live — Calibration (bias + real-spread σ): the win% is grounded in this city's actual track record.
• ✅ Live — Net edge (#6): only edges that survive costs count.
• ✅ Live — ¼-Kelly sizing (#7): turns a signal into a bet size.
• ✅ Live — Calibration scorecard (#4): realised win-rate vs price paid, per band, plus a Brier score. → the quality control that tells you if the edge is real. (Model-probability Brier accrues as new bets are logged with it.)
• ✅ Live — Ensemble distribution (#1): real probabilities from 100+ forecast members instead of a bell-curve guess. → catches the odd days (fronts, marine layer) a smooth curve can't.
• ◑ Partial — Skill-weighted blend (#5): accuracy-by-region is now measured (who's best where); weighting the blend by it is the next step.
• ⏳ Planned — Exact station-max (#3): predict the number Polymarket resolves on, rounding and all — not a grid approximation.
• ⏳ Planned — Correlation view (#8): ten same-day bets in one heat wave are really one bet — size them together, not separately.
• ⏳ Planned — Walk-forward backtest (#9): P&L broken down by city, edge and flag, so you see which signals actually pay and which are noise.
• ⏳ Planned — MOS/NBM feeds (#2) & automated capture + order-book snapshots + resolution alerts (#10) — the server-side pieces (need hosting, per the sharing note).