MoneyPuck - More robust statistics/metrics for HS hockey

[Wallyworld]

I was thinking of a composite score with all the usual categories plus ones like penalties drawn, shooting %, passing completion % (somewhat subjective as you could have a certain length so your not counting five foot passes etc), Average distance from puck on shift (with video technology we have opportunity for stats that were not feasible in the past.) Avg time of possession (from a sampling of shifts, how long do they hold puck etc.), Giveaways/Turnovers (had puck taken from them or passed directly to opposing player), icing, offsides, penalties, steals, hits, also possibly have slightly less weight for second assist on goal...

Just brainstorming but I find the subject very interesting and with the new technology, I think we need more metrics than just G/A/Pts etc. We saw how baseball evolved in this respect. Once you have this boiled down to a single number you could track and tweak it over time until it actually correlates with predicting future performance as well. That would be the home run, if you had an algorithm that was predictive. Something scouts would pay to have access to maybe???

Anyone else have ideas for other meaningful/predictive categories.

[O-townClown]

Wally, a very good idea. I think you could start where European soccer is now and save a lot of time. Possession in hockey isn't real clear, so maybe a territorial measure.

[blueblood]

Soocer possession is defined by all touches by the team.

So if soccer team A has 600 touches to 400 touches for team B, the possesion stat that is quoted is 60% for team A and 40% for team B. Length of pass does not matter.

Other stats we track for our HS team are:

Quality Scoring chances - shots inside face-off dots to each post
Body play - doesn't always mean checks
Giveaways and takeaways
Special teams tracking (+ for PPG ands SHG scored; - for PPG or SHG allowed).

[O-townClown]

Blue, the phone app I have uses a toggle switch to track time of possession for soccer.

[karl(east)]

Wally, a very good idea. I think you could start where European soccer is now and save a lot of time. Possession in hockey isn't real clear, so maybe a territorial measure.

I've long thought that this would be an interesting stat to keep.

A lot of teams keep their own detailed stats, but don't make them public. At last year's Tournament, Benilde had a complete stat packet that I thought was pretty impressive, with complete data for the team and every player on not only PP/PK and faceoffs, but also such things as shift times and an in-depth shot breakdown. The shot breakdown included shots attempted, percentage that hit the net, missed, were blocked, saved, etc.

No doubt hockey has a lot of room for development on this front, and coaches who use this data well might be able to gain some real advantages. I'm curious to hear if anyone is as advanced on this front as Benilde.

I love the idea of some day developing a single metric or small set of metrics, like baseball now has in things like Runs Created and Wins Above Replacement. We definitely need to move beyond our G/A fixation (especially at the HS level, where these can be quite misleading), and start to work in things like body play, rub-outs, and back-checking.

Hockey is such a fluid sport that it's a lot harder to pin these things down than in football or baseball, and there are some very real limitations on the invention of "great statistics." (Bill James has some great essays on the topic that I'd love to post on here at some point.) But I'm all for trying to figure it all out.

[almostashappy]

I've long thought that this would be an interesting stat to keep.

A lot of teams keep their own detailed stats, but don't make them public. At last year's Tournament, Benilde had a complete stat packet that I thought was pretty impressive, with complete data for the team and every player on not only PP/PK and faceoffs, but also such things as shift times and an in-depth shot breakdown. The shot breakdown included shots attempted, percentage that hit the net, missed, were blocked, saved, etc.
(SNIP)

It would be great to have statistics that recognize more of the things that players do to contribute to team success beyond goals and assists (e.g. blocked shots, intercepted passes, rub-outs). But I would be very cautious if these additional metrics were tracked and reported by the teams themselves. It would be very hard to compare stats between different teams and different players, given the potential for uneven tracking and/or stat-padding. At least with goals and assists, the decision on who did what is out of the coaches hands, and determined by the refs. They don't always get it right, but at least they don't have skin in the game.

Trust me, I know...given my experience as the parent of a lacrosse player. Lax refs don't even decide on assists...it's up to the coaches and an honor system, and it doesn't work well at all. Huge difference on the ratio of assists/goals between different programs, systematic from year to year (given the tighter definition of "assist" in lacrosse, a lot more goals are unassisted). This renders the laxhub's "Points" leaderboard functionally useless.

[observer]

One of the best that gives a lot of information is completed passes. Some teams target 40 completed passes per period. Beyond that who made the successful pass and who received the successful pass. All the passing stats lead to possession and control of the game. That combined with detailed shot stats give you a lot regarding individual players. Also gives the coach good info. regarding where individual players, and team, can improve after each period and the game. When completing 30-40 per period the results are good. When not there's room for improvement.

[Wallyworld]

I would like to see if a computerized program could compute average distance from puck for a player during his shifts. i.e. highlight player and puck and then have computer measure distance every second or so. This may be predictive in terms of those guys that have a nose for the puck or, as Gretzky said, good players move towards where the puck is while great players move to where the puck is going to be.

[Simpleton]

Great topic...agree with many thoughts already expressed, especially the challenges of speed and 'player interdependence' with respect to any measure accurately depicting performance/value. Some experiences/thoughts of my own:

- Need to be careful of mixing too many numbers when it comes to 'individual player' assessment...too much interdependancy to deal with.
- Should probably break discussion into two categories - i) player performance; ii) team performance.
- Personal experience is that Giveaway/Takeaway (G/T) ratio draws the strongest correlation to performance. Ranked players top-to-bottom one season and of all the stats kept, G/T ratio was easily the closest to final ranking. I think it also captures the impact of other proposed stats (ex. rub outs, back-checking, etc.). Think the correlation would have been even stronger if we were able to do it by zone (ie. more points deducted for turnover in your own zone vs. offensive zone).
- Also like 'team-weighted plus/minus' - where you calculate the average plus/minus for the 'average player' on a team, then compare a specific player with the average. In other words...a 'minus-2' on a team with an average of 'minus-12' would get a higher score than a player with a 'plus-2' on a team with an average of 'plus-12'. Followed an 'upper level' player last year who was a 'minus-12' on a team full 'plus-10s'. No surprise when I heard a team rep say he was a significant liability despite lots of offensive upside.
- toughest challenge at higher levels is issue of 'matchups'. If a coach has a third-line checking unit that's consistently matched against the opposing team's top line, then it's hard to avoid 'below average' stats...even though they may be doing a solid job of neutralizing the other team's strength and enabling their team to be successful

Again...great topic.

[O-townClown]

I wonder what NHL scouts do. One night they watch the OHL, the next night a young player that is in the NCAA, the next it is Prep School hockey, and finally an NAHL game. How do we compare the players when we can't treat the level equally?

My guess is that something as simple as tracking how many times the puck crosses the blue line in a game would provide a "speed gauge" so you could quantify that a 3-point night in the NAHL is superior to a 3-point night for some private boarding school in Connecticut.

[Wallyworld]

I would love to do a website with this whole statistical project and eventually post these stats on various teams through season etc. Is there anyone out there with Web design experience who would be interested in collaborating and/or doing it for a reasonable fee?