EDJEPredict Responsibly®
◆ FULL TRANSPARENCY · PUBLIC TRUST LEDGER

Every call. Every miss. Every source.

Trust is the whole product. We publish our accuracy, our data sources, our model weights, our commercial relationships, and the forecasts we got wrong. If we hide something, assume it's because we're embarrassed by it — so we don't hide anything.

73.2%
◆ Overall hit rate
148
◆ Resolved markets
+8.2
◆ Avg edge vs market (bps)
2.41
◆ Sharpe ratio

Live accuracy ledger

Every resolved market since EDJE went live. We leave every call published — wins and misses — and annotate the resolution. No market gets quietly deleted.

◆ ALL · 148 ◆ HITS · 108 ◆ MISSES · 40 ◆ ELECTIONS ◆ SPORTS ◆ ECONOMICS ◆ CULTURE ◆ CRYPTO
DateMarketEDJEActualEdgeResult
Oct 18◆ SPORTSDodgers win NLCS Game 768%YES+6.1HIT
Oct 16◆ ECONOMICSFed October rate hold89%HOLD+4.3HIT
Oct 14◆ ELECTIONSCA Prop 7 passes52%FAIL−3.2MISS
Oct 12◆ CULTUREOppenheimer wins Best Pic Globes73%YES+8.7HIT
Oct 10◆ CRYPTOBTC > $75K by Oct 1041%NO+5.4HIT
Oct 8◆ ELECTIONSMamdani wins NYC primary61%YES+9.2HIT
Oct 5◆ SPORTSChiefs cover vs Jaguars66%NO−5.8MISS
Oct 3◆ ECONOMICSNonfarm payrolls > 150K78%YES+7.3HIT
Oct 1◆ CULTURETaylor Swift tour hits $2.5B91%YES+2.1HIT
Sep 29◆ CRYPTOSOL ETF approval before Oct38%NO+4.6HIT
Sep 27◆ ELECTIONSShutdown resolved before Oct 172%NO−7.9MISS
Sep 24◆ SPORTSHaaland EPL top scorer round 769%YES+3.4HIT

◆ SHOWING LAST 12 · 148 TOTAL · LOAD MORE →

Per-model performance.

We disclose each model's standalone hit rate before reconciliation. Shows you which models carry which categories — and why we weight them the way we do.

Claude
◆ ANTHROPIC · 0.32 WEIGHT
Standalone76.4%
ChatGPT
◆ OPENAI · 0.26 WEIGHT
Standalone69.1%
Grok
◆ xAI · 0.19 WEIGHT
Standalone62.3%
Perplexity
◆ PERPLEXITY · 0.23 WEIGHT
Standalone71.0%

◆ RECONCILED ENSEMBLE = 73.2% · higher than any single model · METHODOLOGY →

Category breakdown.

Where we win and where we lose. If we're bad at a category, we tell you — don't trade blind.

◆ ELECTIONS
78.4%
37 resolved
◆ ECONOMICS
76.9%
26 resolved
◆ SPORTS
71.2%
52 resolved
◆ CULTURE
69.0%
13 resolved
◆ CRYPTO
65.0%
14 resolved
◆ MENTIONS
52.0%
6 resolved

◆ MENTIONS MARKETS (what will Trump/Vance/CEO say) are barely coin-flip. We publish them for context but don't recommend trading them without large edges.

Data sources.

Every number in every forecast traces back to a primary source. Here's the full stack.

◆ MARKET ODDS
Kalshi Public API
Live odds on all CFTC-regulated contracts. Refreshed every 5 minutes.
◆ MARKET ODDS
Polymarket Public API
Crypto-settled markets. Refreshed every 5 minutes.
◆ POLITICAL
FEC + State Election Boards
Candidate filings, fundraising, early vote data.
◆ POLITICAL
270toWin · Race to the WH
Battleground state ratings, electoral map.
◆ MACRO
FRED (St. Louis Fed)
Rates, employment, inflation, GDP.
◆ MACRO
BLS + Treasury Direct
Jobs report, debt stack, auction data.
◆ SPORTS
ESPN + Official League Feeds
NFL, NBA, MLB, NHL, EPL stats.
◆ SPORTS
PFF + Opta
Advanced metrics: EPA, xG, DVOA.
◆ FINANCE
SEC EDGAR
10-K, 10-Q, 8-K filings, S-1s for IPO calls.
◆ CRYPTO
CoinGecko + Chain Explorers
BTC/ETH/SOL prices, on-chain volume.
◆ CULTURE
Box Office Mojo + Spotify Charts
Film grosses, streaming numbers.
◆ VERIFICATION
Perplexity Citation Layer
Every quantitative claim cross-checked before publish.

What we got wrong.

If we hide our misses, we're not a research publisher — we're a hype machine. Here are the biggest calls we blew in the last 60 days, with annotations on why.

Shutdown resolved before Oct 1 · EDJE 72% YES → actual NOSEP 27
Why we missed: Over-weighted historical base rate (shutdowns usually resolve in <3 days). Under-weighted the appropriations-fight-as-performance signal that Grok was flagging from X. In the reconciliation layer we dampened Grok's realtime signal too aggressively on political process markets. We adjusted the weighting after this miss.
Chiefs cover vs Jaguars · EDJE 66% YES → actual NOOCT 5
Why we missed: Standard sharp-money bias. Public line was Chiefs –3, sharp money moved it to –4.5, we read the move as justified by injury news that turned out to be noise. Lesson: cross-check sharp money against actual injury reports from the Thursday practice sheet, not aggregate Twitter sentiment.
CA Prop 7 passes · EDJE 52% YES → actual NOOCT 14
Why we missed: California ballot measures are structurally harder for the ensemble because statewide polling is sparse. We're at 58% hit rate on ballot measures vs 78% on candidate elections. We now mark ballot measure forecasts with a lower-confidence chip and recommend smaller position sizing.

Commercial disclosures.

Every dollar flowing in or out of EDJE, disclosed.

◆ REVENUE SOURCES
  • Subscription fees ($19 Pro, $99 Desk)
  • Kalshi affiliate referrals
  • Polymarket affiliate referrals
  • Enterprise API licensing (future)
◆ WE DO NOT ACCEPT
  • Campaign or candidate money
  • Sponsored forecast placements
  • Sportsbook sponsorship (non-prediction-market)
  • Payment to boost or suppress a market
EDJE staff trading policy: No team member may trade on a market EDJE has published research on for 48 hours before or after that research goes live. Violations result in immediate termination and public disclosure.
◆ THE EDJE TRUST PLEDGE

What we will always do.