Track record
What the forecasts would have made or lost, recomputed from the hash-chained ledger on every visit. Wins and losses both stay on this page.
The pipeline's targets
| Stage | Input → output | Target | Now |
|---|---|---|---|
| Signals in | public news, searches, filings and research → evidence matched to each market question share of questions with fresh evidence. Loop: a question without evidence triggers a targeted news fetch (news lane). | 80% | 100% |
| Forecasts out | the crowd's price and the matched evidence → one probability per question crowd Brier minus forecast Brier on settled questions. Loop: the engine halts its adjustments when the 95% interval is below zero. | 0 | – (no data yet) |
| Edge vs the market | the forecast and the live price → a side and a price gap points of disagreement on open forecasts; a gap under the spread cannot be captured. Loop: positions open only at 2 points or more. | 2 | 0.93 |
| Value captured | simulated positions at the market price → profit and loss return on settled positions, %. Loop: the track record recomputes every position from the ledger. | 0 | – (no data yet) |
Forecast book
One $100.00 position per forecast that moved at least 2 points from the market, bought at the market price at that moment, held to resolution. 15 forecasts agree with the crowd and hold nothing. Expected P&L at entry, if the forecasts are right: +$41.93.
| Question | Bought | Now | P&L |
|---|---|---|---|
| US x Iran Effective Ceasefire by November 22? 2026-11-12 · forecast 66.9% vs crowd 63.5% | YES @ 64¢ | 52¢ | −$18.11 |
| Jensen Huang announced resignation or removal by December 31? 2026-11-12 · forecast 59.4% vs crowd 48.9% | YES @ 49¢ | 47¢ | −$3.92 |
| US-Iran Final Nuclear Deal by December 31, 2026? 2026-11-12 · forecast 3.7% vs crowd 7.5% | NO @ 93¢ | 93¢ | $0.00 |
| Israel x Iran ceasefire continues through December 31? 2026-11-12 · forecast 71.6% vs crowd 64.5% | YES @ 65¢ | 65¢ | $0.00 |
Paper desk
The earlier pilot: $10,000.00 of simulated money traded by a price-blind estimator. Equity now $8,500.69 (-14.99%), realized −$1,313.21 on 9 closed trades (2 won, 7 lost), 3 open.
| Question | Bought | Now | P&L |
|---|---|---|---|
| Will Anthropic IPO by November 18, 2026? 2026-11-10 to 2026-11-12 · resolved | YES @ 15¢ | 0¢ | −$390.01 |
| Jensen Huang announced resignation or removal by November 13, 2026? 2026-11-11 to 2026-11-12 · resolved | NO @ 56¢ | 0¢ | −$465.63 |
| Jensen Huang announced resignation or removal by November 12, 2026? 2026-11-11 to 2026-11-12 · resolved | NO @ 64¢ | 0¢ | −$459.79 |
| Will Anthropic IPO by November 16, 2026? 2026-11-10 to 2026-11-10 · resolved | YES @ 9¢ | 0¢ | −$526.82 |
| Will Bilibili Gaming win Worlds 2026? 2026-11-09 to 2026-11-09 · retired | YES @ 38¢ | 37¢ | −$13.25 |
| Pete Hegseth impeached by December 31, 2026? 2026-11-09 to 2026-11-09 · retired | YES @ 72¢ | 69¢ | −$47.79 |
| Next US-Iran senior diplomatic meeting by November 30, 2026? 2026-11-09 to 2026-11-09 · retired | NO @ 85¢ | 83¢ | −$20.99 |
| Heat vs. Kings: O/U 235.5 2026-11-09 to 2026-11-09 · resolved | NO @ 73¢ | 100¢ | +$365.65 |
| Spread: Heat (-9.5) 2026-11-09 to 2026-11-09 · resolved | NO @ 80¢ | 100¢ | +$245.41 |
Accuracy against the crowd
The pilot's estimator, scored on 9 resolved questions: Brier 0.2184 against 0.1067 for the market price (lower is better). The market was more accurate, so the estimator was switched off and the current engine starts from the market price and moves only on cited signals. The current engine has no settled forecast yet; its score appears here when the first market resolves.
Pipeline clock: 6 of 6 stages on time · forecast tick on time 100% of hours this week
| Stage | Every | Last | State |
|---|---|---|---|
| Market prices the 5-minute cron refetches; the synthetic feed is never scored | 5 min | 4 min ago | on time |
| Signal stream q08-signal-engine scrapes; the news lane fetches by question name when the stream is thin | 1 h | 8 min ago | on time |
| Forecast tick a tick older than 80 minutes re-runs on the next 5-minute fire | 1 h | 54 min ago | on time |
| Settlement checks every question past its deadline is fetched by name each hour until it settles | 1 h | 0 min ago | on time |
| Financial context FRED is refetched every 6 hours per series | 6 h | 2 h ago | on time |
| Product manager runs inside the forecast tick once a day | 23 h | 9 h ago | on time |
Clock data (JSON). A late stage repairs itself on the next fire; one that stays late is filed for the code loop automatically.
Product loop
A product manager runs inside the pipeline once a day: it reads its own numbers (12 questions, 0 settled, signal coverage 100%, 1 broadcasts this week, 656 page views), turns its own knobs (updates an hour: 5, posts a day: 3) and files the next growth or repair change for the fleet's code loop. Growth targets: Reader growth – against 1.1 (no data yet); Distribution 0 against 10 (missed); Cross-promotion 38 against 10 (met); Questions covered 12 against 12 (met); Calls with an edge 1 against 3 (missed); Signal coverage 1 against 0.8 (met). Experiment running: broadcast_format, arm registered-first. Last run 2026-11-12 15:15 UTC: 1 knob change (max_updates 4→5), 0 change filed. Its data.