* Include participant EV/futures data in data sync export/import
The export now includes the full participantExpectedValues records
(placement probabilities, source, sourceOdds) alongside participants.
On import, EV records are upserted in merge mode and created fresh in
replace mode (cascade handles cleanup). Backward-compatible with older
exports that lack the participantExpectedValues field.
https://claude.ai/code/session_01TVzuuDR7jPtmihaXTrsX7B
* Address all data-sync code review findings
- Wrap importSportsDataFromJSON in a db.transaction() so any mid-import
failure rolls back cleanly instead of leaving partial data
- Add zod validation of imported JSON before any DB writes, giving a
clear error on malformed/incompatible files
- Fix N+1 queries: build participantIdMap during participant import so
the EV and results sections resolve IDs from memory, not extra queries
- Fix merge mode silently skipping existing participants — now updates
shortName, externalId, and expectedValue
- Add participantResults to export/import (was deleted in replace mode
but never exported, so results were permanently lost)
- Restore calculatedAt timestamp on EV import instead of defaulting to
now(); stored as ISO string in the export for portability
- Document that onDelete:cascade covers participantExpectedValues and
participantResults when participants is deleted; remove the now-
redundant explicit participantResults delete from replace mode
- Bump export version to 1.1; old v1.0 files still import cleanly since
participantExpectedValues, participantResults, and calculatedAt are
all optional in the zod schema
- Collapse all six DB fetches in exportSportsDataToJSON into one
Promise.all for parallel execution
- Add countAllParticipants() and countAllParticipantEVs() model funcs
- Show Participants and EV Records counts on the dashboard stat cards
- Replace raw DOM form creation/submit in handleImport with useFetcher,
giving proper loading state, no full-page reload, and type-safe data
- Remove dead export intent handler from the action function
https://claude.ai/code/session_01TVzuuDR7jPtmihaXTrsX7B
---------
Co-authored-by: Claude <noreply@anthropic.com>
* fix: remove probability sum validation and add batch save with total EV
- Remove the requirement that manual probabilities sum to 1.0, allowing
partial distributions (e.g. when a participant has <100% chance of top 8)
- Replace per-row save buttons with a single "Save All" button that submits
all participants at once
- Add a Total EV row at the bottom of the table summing all participant EVs
- Update action to batch process all participants from a single form submission
https://claude.ai/code/session_01WHRynBCcugSK7HHEmN6Yuy
* fix: validate participantIds server-side instead of trusting form input
Fetch participants from the DB in the action using the season ID from
the URL params, eliminating the untrusted participantIds hidden field.
Remove the now-unused hidden input from the frontend form.
https://claude.ai/code/session_01WHRynBCcugSK7HHEmN6Yuy
---------
Co-authored-by: Claude <noreply@anthropic.com>
* feat: sync odds-based EV to participants table for draft room ranking
When admin enters futures odds or manual probabilities, the calculated EV
is now synced from participantExpectedValues to participants.expectedValue.
This closes the gap where EVs were calculated but never reflected in the
draft room ranking. Also changes expectedValue from integer to decimal(10,2)
for better precision, and displays "—" for participants with no odds data.
https://claude.ai/code/session_01JWG2zg2EMzCn6RgEufPC1T
* fix: correct expectedValue type to string in participants route
Missed instance of `expectedValue: 0` (number) that failed typecheck
after schema change from integer to decimal.
https://claude.ai/code/session_01JWG2zg2EMzCn6RgEufPC1T
---------
Co-authored-by: Claude <noreply@anthropic.com>
- Added loader and action functions for managing expected values in sports seasons.
- Implemented UI for recalculating probabilities based on participant results.
- Created a service to update probabilities after results are finalized, including handling finished and unfinished participants.
- Developed tests for the probability updater service to ensure correct functionality.
- Introduced a preview feature to show potential changes before applying updates.
Implements Phase 5.2 of the EV system with Harville-Malmuth Independent Chip Model
for calculating participant placement probabilities from futures odds.
## Key Features
### ICM Probability Calculator
- Implements Harville-Malmuth method for distributing probabilities
- Converts American odds to championship probabilities
- Generates P(1st) through P(8th) for all participants
- Column-normalized: each placement sums to 100% across all teams
- Works with any number of participants (not limited to 8)
### Admin UI - Futures Odds Entry
- Enter American odds (e.g., +550, -200) for championship futures
- Live preview of ICM-calculated probability distributions
- Displays all 8 placement probabilities
- Persists odds for editing on subsequent visits
- Automatic probability normalization (removes bookmaker vig)
### Database Schema Updates
- Renamed participant_expected_values.season_id → sports_season_id
- Updated foreign key to reference sports_seasons instead of seasons
- Added source_odds field to store original futures odds
- Migration 0025: Column rename and FK update
- Migration 0026: Add source_odds field
### Model Layer
- participant-expected-value: CRUD operations for probability distributions
- Supports multiple probability sources (manual, futures_odds, elo_simulation)
- Automatic EV calculation based on league scoring rules
- Probability validation and normalization
### Service Layer
- icm-calculator: Harville-Malmuth probability distribution
- probability-engine: Odds conversion and Elo utilities (for future use)
- bracket-simulator: Monte Carlo simulation (for future hybrid approach)
- ev-calculator: Expected value computation from probabilities
## Technical Details
- Uses exponential decay favoring top positions for strong teams
- Preserves championship probability ordering in final distributions
- Row sums vary (strong teams ~100%, weak teams lower)
- All probabilities between 0-1, mathematically valid
- Comprehensive test suite: 97 tests passing
## Future Enhancements
- Hybrid approach: ICM pre-playoffs, bracket simulation during playoffs
- Integration with league-specific scoring rules
- Historical probability tracking for accuracy analysis
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
