Monday, October 20, 2014

Quantifying defensive-play using micro-stats



New data-generated defensive-metrics
 
Defensive-play is the most difficult aspect of hockey analytics. Yet it plays a huge part in every player’s impact on the game. Defensive-play is primarily about removing possession from the opposition; as such it is not restricted to the defensive-zone. 

The two main aspects of defensive-play involve positioning, and breaking the opposition’s flow of possession. Positioning is very difficult to quantify, as it is often a non-event. The reason it can be referred to as a non-event is that a well-positioned player will sometimes force the opposing team into choosing a different option. An example of this is seen as a team is trapping in the neutral-zone. A defenseman looking to make a stretch pass will abort that option if he doesn’t see a teammate available. In other words, a well-positioned player in the neutral-zone forced a defenseman not to attempt a play. This didn’t immediately cause a loss of possession, but produced a non-quantifiable impact all-the-same.

A more quantifiable aspect of defensive-play involves the act of removing possession from the other team. There are multiple ways of removing possession including; stick-checks, body-checks, blocked passes, and blocked shots. In other words, any play that breaks the other team’s cycle of possession.  
Much like plays with possession, each of these defensive-plays can be tracked as either successful or unsuccessful. The key factor in each event is whether the attempted play resulted in the opposition losing or maintaining possession.  A list of all the events tracked within my system can be found here.

Tracking individual puck-possession events allows us to produce multiple metrics that can be used to quantify defensive-play. I’ll focus on 4 (12 if broken down by zone) metrics in this post. They include; overall defensive success-rate, successful defensive-plays per-60, failed defensive-plays per-60, and scoring-chances against per-100 defensive-plays.

OVERALL DEFENSIVE SUCCESS-RATE

Answering the question; how successful is a player when attempting to remove puck-possession from the opposition?

This metric represents how successful each player is when attempting to remove puck-possession from the opposition, and is communicated as a percentage (success-rate). It divides the number of successful attempts at removing puck-possession from the opposition by the total number of attempts. Defensive-plays used in this calculation include stick-checks, body-checks, blocked shots, and blocked passes. The metric can also be broken down by zone (offensive, neutral, and defensive-zones). 

Teams that out-chance the opposition at even-strength have produced an average defensive success-rate of 64.5%, while teams that have given up more scoring-chances than they produced have a defensive success-rate of 62.8%.

SUCCESSFUL DEFENSIVE-PLAYS PER-60 MINUTES PLAYED

Answering the question; how often is a player removing puck-possession from the opposition?

This metric represents the number of successful attempts at removing puck-possession from the opposition a player contributes per-60 minutes of even-strength ice-time. Defensive-plays used in this calculation include stick-checks, body-checks, blocked shots, and blocked passes. The metric can also be broken down by zone (offensive, neutral, and defensive-zones).

Teams that out-chance the opposition at even-strength have averaged 154.5 successful defensive-plays per-60, while teams that have been out-chanced have averaged 152.4.

FAILED DEFENSIVE-PLAYS PER-60 MINUTES PLAYED

Answering the question; how often is a player failing to remove puck-possession from the opposition?

This metric represents the number of failed attempts at removing puck-possession from the opposition a player contributes per-60 minutes of even-strength ice-time. Defensive-plays used in this calculation include stick-checks, body-checks, blocked shots, and blocked passes. The metric can also be broken down by zone (offensive, neutral, and defensive-zones).

Teams that out-chance the opposition at even-strength have averaged 85.1 FAILED defensive-plays per-60, while teams that have been out-chanced have averaged 90.1.

SCORING-CHANCES AGAINST PER-100 DEFENSIVE PLAYS
 
(lower is better)
Answering the question; how much of an impact is the player’s defensive-play having on limiting scoring-chances against. 

This metric represents the number of scoring-chances a player is allowing per-100 attempts to remove possession. It helps quantify how efficiently a player’s defensive-play is helping to reduce scoring-chances against. Defensive-plays used in this calculation include stick-checks, body-checks, blocked shots, and blocked passes.

Teams that out-chance the opposition at even-strength have averaged only 5.1 scoring-chances against per-100 defensive-plays, while teams that have been out-chanced have averaged 7.2 scoring-chances against per-100 defensive-touch.

Quantifying defensive-play remains among our biggest challenges in hockey analytics. Data-generated scouting (micro-stats) definitely helps break this aspect of play down into manageable portions. That said, the non-events or position-based defensive-event will never be quantifiable. As such, it will remain an aspect of play best judged by traditional (perception-based) scouting.