Contract Projections Model

Advanced ML-Driven Contract Value Predictions

Machine learning models trained on historical contract data, with transparent performance tracking and accountability reporting after each free agency period.

Market Inflation

Adjusts for projected salary cap growth and market conditions

Comparable Analysis

Advanced matching algorithm for similar player contracts

ML Baseline

Machine learning model trained on historical contract data

Two-Layer Architecture

Our contract projection model combines machine learning baseline predictions with sophisticated comparable player analysis for maximum accuracy.

Layer 1: ML Baseline Model

Training Features

  • • Player performance metrics (PVPI, traditional stats)
  • • Age and career stage indicators
  • • Position and role classifications
  • • Historical market conditions
  • • Team cap situation and needs

Market Inflation Adjustment

Adjusts baseline projections based on projected salary cap growth. If the cap was growing 1M when similar players signed, but is projected to grow 15M, the model inflates contract values accordingly.

Layer 2: Comparables Override

Matching Algorithm

  • • Performance similarity (PVPI, production)
  • • Situational context (age, position, career stage)
  • • Market conditions (UFA/RFA, team needs)
  • • Contract structure (term, bonuses)

Override Logic

High Confidence (85%+): Use comparable-based projection

Medium Confidence: Weighted average of ML and comparables

Low Confidence: Default to ML baseline

Machine Learning Core

Gradient boosting model trained on historical contract data with feature importance analysis and cross-validation.

Dynamic Market Adjustment

Real-time salary cap projections and market inflation factors ensure projections reflect current economic conditions.

Confidence Scoring

Every projection includes confidence intervals and uncertainty ranges based on comparable quality and data availability.

Model Transparency

All projections are model-generated estimates based on statistical analysis and historical trends. These are not confirmed contract negotiations or insider information, but analytical projections designed to provide data-driven insights into potential contract values.

Comparable Player Analysis

Our sophisticated matching algorithm identifies similar players based on performance, situation, and market context to provide the most accurate contract projections.

Matching Algorithm

Performance Similarity

  • • PVPI score comparison
  • • Point production rates
  • • Advanced metrics
  • • Consistency factors

Situational Context

  • • Age and career stage
  • • Position and role
  • • Team context
  • • Contract timing

Market Conditions

  • • UFA vs RFA status
  • • Salary cap environment
  • • Team needs
  • • Competition level

Contract Structure

  • • Term length preference
  • • Signing bonus structure
  • • Trade protection
  • • Performance bonuses

Confidence Scoring System

85%+

High Confidence

Multiple strong comparables with similar performance, age, and market conditions. Uses comparable-based projection.

60-84%

Medium Confidence

Some good comparables but with notable differences. Uses weighted average of ML baseline and comparables.

<60%

Low Confidence

Limited or poor-quality comparables available. Defaults to ML baseline with market adjustments.

Performance Tracking & Accountability

Our contract projection models are trained on historical data. We commit to transparent performance tracking and will publish detailed accuracy reports after each free agency period.

Current Status

Our contract projection models have been trained on historical NHL contract data spanning 2010-2024. We will begin live testing during the 2025 free agency period and publish our first performance report at the conclusion of that period.

Performance Tracking Timeline

Model Training

Complete

Models trained on historical NHL contract data from 2010-2024

2025 Free Agency

Upcoming

First live testing period - projections will be made for major signings

Performance Report

Planned

Detailed accuracy analysis published after free agency concludes

Model Refinement

Ongoing

Continuous improvement based on real-world performance data

Our Reporting Commitments

Transparency

All projection accuracy metrics will be publicly reported

  • Mean absolute error
  • Prediction accuracy rates
  • Confidence score validation
  • Market timing analysis

Accountability

We will track and report both successes and failures

  • Detailed breakdown of missed projections
  • Analysis of model limitations
  • Identification of improvement areas
  • Honest assessment of market factors

Continuous Improvement

Performance data will drive ongoing model enhancements

  • Regular model retraining
  • Algorithm refinements
  • New data source integration
  • Methodology updates

What to Expect in Our Performance Reports

Quantitative Metrics

  • • Mean absolute error across all projections
  • • Percentage of projections within ±$1M, ±$2M ranges
  • • Confidence score accuracy validation
  • • Performance by player position and contract type
  • • Market timing adjustment effectiveness

Qualitative Analysis

  • • Case studies of significant projection misses
  • • Analysis of market factors we failed to capture
  • • Identification of model blind spots
  • • Lessons learned and planned improvements
  • • Updated methodology based on real-world results

First Report Timeline: Our inaugural performance report will be published within 30 days of the conclusion of the 2025 NHL free agency period, providing a comprehensive analysis of our model's real-world performance.