About DraftResearch.com

We built the dataset.
Then we built
the platform.

DraftResearch.com is a hockey analytics platform built on 61 draft classes of NHL outcomes across 47 feeder leagues. Interactive tools for analysts, broadcast-ready intelligence for media, and data-driven consulting for front offices – all from one dataset.

Methodology The Dataset Contact

The Work

One dataset, four tools.

Every NHL draft pick across 61 draft classes has a CarProd tier assignment. The Career Production model classifies players using career games played, percentile rankings within draft position, and historical benchmarking across 47 feeder leagues.

Every NHL draft pick across 61 draft classes has a CarProd tier assignment. The Career Production model classifies players using career games played, era-adjusted production, and historical benchmarking across 52 feeder leagues — updated automatically throughout the season.

The ReDraft Room lets you re-draft any year with full hindsight. CarProd Tables rank every NHL skater by sustained career production and peak performance. Prospect Probabilities project draft-eligible players using historical comparable outcomes. The Trade Calculator analyses trades using CarProd tiers, pick equity models, and a searchable database of 2,495 real NHL trades with GM attribution.

Nothing here is opinion. The model scores players based on comparable historical outcomes. When we say a player had a low probability score, we mean that historically, players with that profile convert at a low rate. We never speculate beyond what the data supports.

Our Principles

Data only. No opinions.

Every claim is a fact from the dataset. We report what historical probability says about players with comparable profiles. We never speculate beyond what the data supports.

Scouts are people, not targets.

The scouts and front office staff who appear in our analysis are treated with respect. Historical record and recognition – never evaluation or criticism of individuals.

Built to use, not just read.

Interactive tools, not static reports. The ReDraft Room, CarProd Tables, Trade Calculator, and probability models let you explore the data yourself – the same dataset that powers our consulting work.

Always current.

The pipeline scrapes 52 leagues daily, rebuilds weekly, and updates trade history automatically. Every tool on the site reflects the most recent data available – not a snapshot from months ago.

The Dataset

What we built from.

61 draft classes of NHL outcomes across 52 feeder leagues and all 32 organisations – with automated daily updates through the 2025—26 season.

Last updated

April 8, 2026

Draft Classes

Full data from the 1964—2025 drafts. Career outcome tracking through the 2025—26 NHL regular season.

Feeder Leagues

OHL, WHL, QMJHL, SHL, Liiga, KHL, NCAA, USHL, BCHL, and 43 additional leagues across Europe and North America. Scraped daily.

12,883

Draft Picks

Every pick across all 61 draft classes with steal/bust flags, pick-before/after context, and career GP tracking.

487K+

Player-Season Records

Unified stats across all 52 leagues with bio enrichment (DOB, nationality, height, weight) from 7 legal data sources.

3,400+

Scouts & Staff

Unique GMs, scouts, coaches, and front office staff tracked across 32 NHL franchises with EP-matched profiles.

30K+

Staff Records

Every GM, scout, and coaching hire across 32 NHL franchises since 2008. Draft attribution linking staff to picks.

2,495

NHL Trades

Every post-lockout trade (2005—present) with GM attribution on both sides. Searchable in the Trade Calculator.

Organisations

All 32 NHL franchises with scouting staff records, CarProd tier analysis, and automated daily data collection.

Methodology

How the model works.

The probability score is a single number – the model's estimate that a player reaches 200 NHL regular season games from their draft year. Here's how it's built.

See the Probabilities

Step 01

League tier assignment

Each of the 47 feeder leagues is assigned a historical conversion tier based on historical NHL conversion rates at equivalent draft positions. The SHL and Liiga rank at the top. The KHL is treated separately due to age and circumstance factors that differ from development leagues.

Tiers recalibrate annually with the addition of each new completed draft year.

Step 02

Age-adjusted production score

A player's production is adjusted for age within their birth year cohort. A 17-year-old producing at a level typical of 19-year-olds in the same league receives a higher adjusted score than the raw numbers suggest. This is one of the model's most predictive variables.

Birth year cohort windows: Sep 15 — Sep 14 (matching NHL eligibility rules).

Step 03

Historical comparables

For each prospect, the model identifies all historical players with similar league tier, age-adjusted score, and position. The proportion of those players who reached 200 NHL games is the raw probability estimate before further adjustment.

Minimum comparable pool: 30 players. Pools below this threshold receive a confidence flag.

Step 04

Position and draft year calibration

Position carries different conversion baselines – historically, centres and defence have higher conversion rates than wingers at equivalent scores. Draft year cohort size also affects probability, as larger draft classes produce more competition for roster spots.

Step 05

Weekly update cycle

Scores recalculate every Tuesday using the most recent available game data from all 47 leagues. Significant statistical shifts – a major scoring streak, a sudden drop in ice time – are reflected within one update cycle.

The model does not incorporate injury information, trade data, or any non-game-performance inputs.

Note

What this is not

A probability score is not a ranking. Two players with the same score are equally likely to make the NHL – the model makes no preference between them. A score of 40 is not "bad" – it means roughly 40% of historically comparable players reached 200 games. Many excellent NHL careers began with scores in that range.

Model Definitions

Three lenses on player value.

Each model answers a different question about a player's career. Together they give a complete picture — peak talent, sustained production, and projected outcome.

CarPeak

How productive was the player at their best?

Best 3 NHL seasons by points production, topped up to 200 GP82 (games played normalised to 82-game seasons). CarPeak captures a player's ceiling — the version of themselves that showed up when everything was clicking. A player with elite peaks but inconsistent seasons will score higher on CarPeak than on CarProd.

Answers: What is this player's ceiling?

CarProd

How productive were they across their whole career?

GP82-weighted average of every qualifying NHL season. CarProd rewards consistency — a player who produces at a second-line level for 12 seasons scores higher than one who has 3 elite years and 5 mediocre ones. It is the most stable measure of what a player actually was, not what they could have been.

Answers: What is this player's sustained value?

CarProj

What level of career production do we expect a prospect to reach?

Where CarProd is the full career and CarPeak is peak performance, CarProj is the forward-looking projection at draft time. Using historical probability tables built from 47 feeder leagues and 60+ draft classes, CarProj assigns each prospect a percentage probability of making and staying in the NHL, along with a predicted tier label — the model's prediction of what a prospect will become.

Answers: What will this prospect become?

Tiers

The tier system

All three models classify players into the same 6-tier structure for consistency: Elite (franchise-level, perennial All-Star), Top Line / Top Pair (first-line forward or top-pair defenceman), 2nd Line / 2nd Pair (second-line contributor), 3rd Line / Bottom Pair (depth scoring), 4th Line / Depth (checking line), and Depth (minimal NHL impact). A player can have different tiers across CarPeak and CarProd.

Contact

Get in touch.

Questions about the model, Pro access, media, or consultancy work.