DailyBiasNew York AM
9:30 AM → noon · New York time
Waiting for manual session preparation.
Input agreement · probability not calibrated. Score —.
Explore all evidence ↗More research & your notes Macro context, market confirmation and your session thesis
Growth, inflation & policy
Source health
Evidence, in plain English.
Write your session thesis.
Follow the evidence.
Every pulled input, organized by its role in the macro picture.
Cross-asset gauges
Latest pulls · separate from saved callsWhat drives the current bias?
Related inputs are averaged within groups. Bullish rates support is halved when eligible VIX and credit-spread inputs both show deterioration on comparable observation dates. This is an experimental conflict rule. Headlines, old Fed language and unsourced monthly surprises receive no directional votes.
What the filter excluded
Leveraged funds positioning
Release ledger · expectations, actuals & revisions.
Official agencies publish actual values. Consensus needs a separate survey source. Lock the expectation before a release, then record the official actual. Both numbers and timestamps stay visible.
User-entered sourced values; the app checks timing, not the contents of the source. Trading Economics provides actual, consensus and revised fields through an optional licensed integration. Set TRADING_ECONOMICS_API_KEY on the server to enable scheduled provider collection. No subscription is required for this manual workflow. Values captured after publication are not automatically certified as first-release vintages.
Read the macro picture.
Source health
Full pulled data · JSON
Your session history.
Compare the original bias with what the market did during that exact window.
Automatic grades use delayed NQ minute bars from the first minute after saving to the session end. Moves within ±0.05% are mixed. Missing coverage stays ungraded. Remove duplicates to exclude them from the scoreboard; Restore is available.
Measure your bias.
Results for the selected research session.
By conviction
By category
By evidence engine
Keep old and revised scoring results separate before judging a change in accuracy.
Win rate = right ÷ (right + wrong). Mixed and pending are excluded. Category direction is compared with the direction implied by your overall outcome. These descriptive results are not a calibrated probability.
Data & storage
Checking private history…
Advanced connection settings
Already connected. Only change this address if you move your data service.
Settings.
Appearance and Home-page preferences save in this browser.
Appearance
Show on Home
How this desk builds a call
Fresh daily rates, risk appetite and dollar readings are grouped. Related indicators share one vote. Monthly levels and weekly positioning remain context. Release surprises require a sourced expectation saved before the event.
Direction, evidence and the ±0.05% flat band freeze at the first save. Human reviews can override an automatic grade. Removed records remain recoverable and cannot auto-save again in the same window.
Best next data upgrades
- Sourced release expectations locked before publication, compared with the first official actual.
- Separate growth, inflation and policy surprises instead of using one generic headline direction.
- Macro trend and financial-stress context, with dates and without duplicate votes.
- Separate forward testing by session and market regime before changing weights.
Published directions are Bullish lean, Bearish Lean, Bullish Bias or Bearish Bias. A full Bias needs a net score of at least 2.5 in magnitude, two supporting groups and no meaningful opposing group; small or conflicting edges remain leans. Missing, stale, single-group or exactly balanced evidence stays Not ready. No win-probability percentage is claimed without forward calibration.
Model development
Checking usable training records…
Training requires session features saved before the open and verified subsequent returns. We do not train a text model to sound more confident or tune weights to the same records used for evaluation.