Does offseason spending predict team improvement?

Executive summary

This study analyzes nine NHL seasons and 1,648 unrestricted free-agent signings to test whether offseason UFA spending predicts team performance. The headline result: after controlling for prior performance, teams show mean reversion — the fitted relationship between prior finish and next-season change is negative (slope ≈ −0.42) and statistically meaningful, so extreme prior results tend to move toward the mean rather than amplify. In practice, this means large UFA spending spikes rarely produce reliably better team outcomes once you account for where a team started; front offices should treat big UFA moves as higher-variance bets, not guaranteed performance upgrades. See the Methodology and Audit pages for data sources, inclusion rules, and robustness checks.

Methodology at a glance

• What we analyzed. Nine NHL offseasons of unrestricted free-agent signings (2017–2025), totaling 1,648 UFA movement events and 284 team-season outcomes (2018–2026).
• Data sources. Contract and signing data from Spotrac and transaction scrapes; game results and team outcomes from the NHL API via nhlscraper. See the project source on GitHub for raw files and scripts.
• Scope and exclusions. This study covers unrestricted free-agent signings only. Restricted free agents, trade-deadline acquisitions, AHL call-ups, and entry-level signings are out of scope for Version 1. The 2026 team outcome season is included; 2026 offseason UFA signing data is pending until the July 2026 window opens.
• Key variables.
  • Outcome: season-over-season team performance change (team points % / playoff outcome delta).
  • Primary predictors: prior season finish and offseason UFA investment (total dollars and a position-weighted Movement Impact Score).
  • Reported headline: the fitted prior-finish coefficient (mean-reversion slope) is approximately −0.42 and is statistically significant across specifications.
• Modeling approach (brief). We estimate the relationship between prior season performance and next-season change while adding controls for offseason UFA spending (dollars and MIS), and run robustness checks excluding COVID-affected seasons. Results are reported with standard errors and p-values; full model specs are in the appendix.
• Robustness and audit. Results hold in the full 2018–2026 sample and when the COVID seasons are excluded; see the Audit page for data-quality checks and sensitivity tests.
• Limitations (short). Correlation is not causation; UFA spending may be endogenous to front-office expectations and injuries. The analysis does not capture in-season trades or prospect development. See the full methods appendix for additional caveats.

Full methods appendix (GitHub source)  ·  Audit reproducibility checks

What this research shows

The strongest predictor of whether an NHL team improves or declines next season is how well they performed this season. Teams that finished near the top tend to fall back. Teams that finished near the bottom tend to improve. This mean reversion effect has a slope of −0.42 and is significant at p < 0.001 across all model specifications.

Offseason UFA spending — measured in total dollars or through a position-weighted Movement Impact Score — does not significantly predict season-over-season performance change after controlling for prior season performance. This holds in both the full 2018–2026 sample and when the COVID-affected 2020 and 2021 seasons are excluded.

This study covers unrestricted free agent signings only. Restricted free agents, trade deadline acquisitions, AHL call-ups, and entry-level signings are out of scope for Version 1 and are documented in the methodology. The 2026 team outcome season is already included; 2026 offseason UFA signing data will be added after the offseason window opens in July 2026.

The research behind it

This project is grounded in behavioral economics research on how disruption and roster change affect performance. The primary inspiration is Hengchen Dai's research on the reset effect in Major League Baseball, featured in the Freakonomics Radio episode Are You Ready for a Fresh Start?

Dai found that when a struggling player is traded — triggering a statistical reset — their performance improves significantly. This project applies that framing to the team level: does offseason roster movement produce a measurable reset effect in NHL team performance?

The answer from the full 2018-2026 outcome window: mean reversion dominates. Where a team starts predicts where they end up far more reliably than what they spent in the offseason.

How to use this site

📊 Key Findings

Two research questions answered with data tables and plain-language interpretations. Start here for the headline results.

🏒 Overview

All 32 teams ranked by Movement Impact Score for each season. See which teams invested most heavily and how that correlated with performance change.

🏟 Team Detail

Jump to any team and season to see their specific UFA signings, MIS breakdown by position tier, and season-over-season performance delta.

🔍 All Signings

Browse and filter all 1,648 UFA movement events by year, position, team, geography, and contract value.