Predicting NHL Player Contracts Matt Cane CaneMatt Hockey Graphs Puck Predicting Contracts: Why? Long Term salary cap planning Teams have limited resources to identify and pursue free agents Useful for players to know what
"Predicting NHL Player Contracts Matt Cane" is the property of its rightful owner. Permission is granted to
download and print the materials on this website for personal, non-commercial use only, and to display it
on your personal computer provided you do not modify the materials and that you retain all copyright
notices contained in the materials. By downloading content from our website, you accept the terms of this
agreement.
Presentation Transcript
01
Predicting NHL Player Contracts Matt Cane
@Cane_Matt
Hockey Graphs | Puck++<br>
02
Predicting Contracts: Why? Long Term salary cap planning
Teams have limited resources to identify and pursue free agents
Useful for players to know what statistically similar players have been paid
Don’t want to leave money on the table
Can help identify teams who overpay vs. market, or find cheap free agent talent
Can help identify players whose price (dollars) is different from their value (wins)
Bonus: Provides something for strangers on the internet to yell at you about when you suggest that their favourite player is overpaid @Cane_Matt | puckplusplus.com | hockey-graphs.com 2<br>
03
Past work Run models over the past 3 off-seasons
2014-15: Linear Regression, UFA Skaters only
2015-16: Beta Regression, UFA + RFA Skaters
2016-17: Random Forest, UFA + RFA + Goalies
Other models:
Manny Perry: k-Nearest Neighbours
Luke Solberg (@evolvingwild): Regression w/ WAR/Game Score/TOI
Chris Watkins (@yolopinato): Regression
Carolyn Wilke: Salary Cap Bands @Cane_Matt | puckplusplus.com | hockey-graphs.com 3<br>
04
The 2016-17 Model Uses Random Forest to predict adjusted cap hit percent at time of signing:
Adj. Cap Hit Percent = (Cap Hit – Min. Cap Hit)/(Max Cap Hit – Min Cap Hit)
Model player salary using:
Prior Year and 3-Year Total Stats from NHL.com
Past Contract History (Last Cap Hit)
Contract Timing (Before a player hits FA or after)
Free Agent Status (UFA vs RFA, Buyout) @Cane_Matt | puckplusplus.com | hockey-graphs.com 4<br>
05
What’s new? Predict both term and salary
Include term as a predictor in the salary model
Use Z-scores by season rather than absolute totals as predictors
Model buyouts separate from other players, assume buyout contracts are for 1 year @Cane_Matt | puckplusplus.com | hockey-graphs.com 5<br>
06
Is the new model any good? For all Free Agents signed after July 1st, 2017 (excluding extensions before FA):
Term model predicts 49.4% of contract lengths correctly @Cane_Matt | puckplusplus.com | hockey-graphs.com 6<br>
07
What does the new model struggle with? Many things!
Predicting the super high-end players
Players like Connor McDavid and Auston Matthews tend to be undervalued
For most players, as term increases predicted salary increases
Some cases this makes sense (bridge vs. long-term deals)
Other cases we’d expect more term would mean less money
The Olds
Harder to model older players since many retire @Cane_Matt | puckplusplus.com | hockey-graphs.com 7<br>
08
How can we use this data?<br>
09
Who were the most overpaid players this off-season? (aka GM errors) @Cane_Matt | puckplusplus.com | hockey-graphs.com 9<br>
10
Who were the most underpaid players this off-season? (aka model errors) @Cane_Matt | puckplusplus.com | hockey-graphs.com 10<br>
11
Predicting Next Contract Term The model is fairly certain Matthews will get a long-term deal, but Marner’s prediction is divided. @Cane_Matt | puckplusplus.com | hockey-graphs.com 11<br>
12
Long-Term Cap Planning (Or are the Sens at risk of losing Erik Karlsson?) Karlsson is a Free Agent in 2019-20
Projected Contract: 8 years, 7.6M
Current Cap Commitments:
2019-20: $39.7M (6F, 2D, 2G)
Other Key Free Agents:
~21.3M to retain
Assuming 80M Cap in 2019-20:
11.5M for ~4 forwards and ~3 defencemen @Cane_Matt | puckplusplus.com | hockey-graphs.com 12<br>
13
Future Work Improving Term Model
Tends to default to 1 year deals in too many cases
More “subjective” variables:
3-Star Selections
NHL Award Voting
Captaincy
Term with Team
Better Adjustments for Injuries and Early Career players
Second year players don’t have the same three year stat profiles
Connor McDavid’s injury changes his stat profile
Including advanced metrics and “one number” metrics for player value
Corsi, WAR, K, etc. @Cane_Matt | puckplusplus.com | hockey-graphs.com 13<br>
