About Kodo Hockey
An independent fantasy hockey analytics site, built and run by one person. Every page is anchored to data we ingested and can point at. No vibes-based rankings, and no unattributed hot takes.
What this site does
Kodo Hockey projects the coming season, prices what a player is worth in your league, and shows the underlying signals that move those numbers: ice time and role, line combinations, power-play deployment, chance quality, goaltending, schedule and cap.
The draft tools are built for the 2026-27 season, which is an 84-game year rather than the usual 82. That is not a cosmetic detail: every counting projection on the site is stated on the 84-game basis, so comparing them against an 82-game source will understate them by a couple of percent.
Where the data comes from
- The NHL’s public feeds for schedules, box scores, play-by-play and shift charts. Shifts are what make the line-combination pages possible: real units measured from who was actually on the ice together, not a depth chart someone typed.
- CapWages for contract and salary-cap figures.
- Your league, if you connect one. Yahoo, ESPN and Fantrax are supported, and connecting one turns generic rankings into your rankings: who is rostered in your league, who is genuinely free, and what your categories reward.
The models behind the signals
The projections start from role rather than from last year’s point total, because role is what actually persists. Ice time, power-play unit and linemates are projected first; rates are shrunk toward position-and-role baselines by sample size; and the counting stats fall out of the two combined.
- Expected goals (xG v2) and on-ice models, used for chance quality and for separating a hot shooting percentage from a real change in the shots being taken.
- Per-60 rates with shrinkage, so a player with 200 minutes is not ranked as confidently as one with 1,200.
- A rookie model that anchors on projected role, and flags players whose projection is still mostly prior rather than evidence.
- Over 700 skaters projected across every category, all of them editable: you can override a projection and the draft board, values and downloads follow your edit rather than ours.
One thing the models deliberately do not include: a within-season recency term. It was built, measured, and rejected, because late-season points scoring did not predict the next season once role was accounted for. Ice time carries; points do not.
How the articles work
Articles are a collaboration between Claude (Anthropic) and the site operator. The model drafts from data we already hold; the operator reviews and edits every piece before it publishes. Nothing posts automatically.
The rule the whole pipeline is built around is that every name, stat and percentage in an article has to trace back to a real row in our database. If the data for a section is missing, the section is dropped rather than filled in, because a plausible invented stat line is worse than a shorter article.
Who runs this
Kodo is a one-person project, built and maintained by Brandon Landry, an economics-trained analyst, who also runs the Kodo baseball site. There is no team, no outside funding, and no paid placements.
Independence and fair use
Kodo Hockey is not affiliated with, endorsed by, or sponsored by the National Hockey League, any NHL club, or any fantasy platform. NHL and club marks belong to their respective owners. Statistical data is sourced from public NHL feeds; fantasy league data is accessed only through the official interfaces of the platform you connect, and only with your permission.
The site currently carries no advertising. How your data is handled is set out in the Privacy Policy.
Get in touch
Email b.lands16@gmail.com, or find the site on X at @kodohockey. Corrections are genuinely welcome: if a number here looks wrong, it might be, and I would rather hear it.