Skip to content

Navigation Menu

Sign in
Appearance settings

Search code, repositories, users, issues, pull requests...

Provide feedback

We read every piece of feedback, and take your input very seriously.

Saved searches

Use saved searches to filter your results more quickly

Appearance settings

fix(data): define market covariates on cross-calendar holidays#26

Merged
ethancjackson merged 2 commits into
mainVectorInstitute/agentic-forecasting-live:mainfrom
covariate-holiday-fillVectorInstitute/agentic-forecasting-live:covariate-holiday-fillCopy head branch name to clipboard
Jul 18, 2026
Merged

fix(data): define market covariates on cross-calendar holidays#26
ethancjackson merged 2 commits into
mainVectorInstitute/agentic-forecasting-live:mainfrom
covariate-holiday-fillVectorInstitute/agentic-forecasting-live:covariate-holiday-fillCopy head branch name to clipboard

Conversation

@ethancjackson

Copy link
Copy Markdown
Collaborator

US-holiday gaps in the Yahoo market covariates made Darts silently skip those TSX origins for the covariate LightGBM rung only (6/72 protected-eval, 27/1134 daily), breaking origin-set comparability with the other rungs. Levels ffill, returns zero-fill (causal both ways; same rationale as the existing StatCan treatment). Adds regression tests incl. the exact 2025-07-04 shape.

Affected stores will be force-refreshed and re-scored after merge.

🤖 Generated with Claude Code

ethancjackson and others added 2 commits July 18, 2026 06:55
The Yahoo market covariates (VIX, WTI, gold, USDCAD, SPX, US10Y) kept their
US-session calendar, so on US market holidays the TSX trades (MLK, Memorial
Day, Juneteenth, July 4, Thanksgiving, the 2025-01-09 closure) the panel had
no row and Darts refused to predict — silently dropping 6/72 protected-eval
and 27/1134 daily origins for darts_lightgbm_cov only, which skews rung
comparability. Levels are now forward-filled onto the full business calendar
(last known value); returns are zero-filled (no session, no move — keeps the
running sum equal to the cumulative price change instead of double-counting).
Both fills are causal, mirroring the StatCan features' existing treatment.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
…ook)

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
@ethancjackson
ethancjackson merged commit 52ad4eb into main Jul 18, 2026
2 checks passed
@ethancjackson
ethancjackson deleted the covariate-holiday-fill branch July 18, 2026 11:01
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant

Morty Proxy This is a proxified and sanitized view of the page, visit original site.