Dobber Investigates the role of advanced statistics in fantasy hockey

Elliotte Friedman posted a column Wednesday night on the increased use of advanced metrics in the NHL. In it, he opines that different teams have been using their own statistical measuring tools for the last several years, not just for evaluating how to use players, but how to pay them. Advanced stats are also starting to play a role in fantasy hockey.

Although there will probably never by a fantasy league that awards points for "Corsi" or "Qual Comp", advanced statistics can be used to get a better handle on the quality of player that you are investigating. But the tightrope that fantasy owners must walk is finding that balance. How much stock do you put into these metrics? If you lean on them exclusively, you will suffer - mark my words. But if you ignore them completely, you will fall behind your fellow GM's.

From Friedman's post:

Oakland A's general manager Billy Beane, the hero of Moneyball, blamed his team's playoff failures on "luck." That's a cop-out.

Actually, it's not. It's an excuse, but a legitimate one. Luck is not a "copout", it's a variable that can't be measured. And it's a negative influence more than a positive one. All the analysis in the world can put together a great team, but it can't measure an injury. Or a coaching mistake. These metrics can help you with the odds and the probabilities, but really that's all they are - odds and probabilities. And unless those odds are 1/1 and the probability is 100%, the method will never be foolproof.

I received my degree in statistics in 1997, with a minor in mathematics. And although much of what I learned has long since been forgotten, there is no denying that numbers interest me and statistics fascinate me. But, thanks to my training, I have a firm understanding of what stats mean and what their limitations are. I understand the importance of sample size and the randomness of certain variables. The average Joe does not.

"Statistics are like a lamp post to a drunk: Useful for support but not for illumination" - Brian Burke

Burke's not right…but he's not wrong, either.

Advanced stats give you a probability based on past performance. But as for using them as a future predictor, there are many weaknesses. They represent a measure of what could be done - not what will be done.

It is impossible to measure all of the outside factors that come into play. These are rarely taken into account by amateur statisticians who constantly fall back on advanced stats when arguing the merits of a player. Factors include:

1. Coaching. This includes:

i. Ice Time. All the stats in the world can tell you how great this player is, but if he's not put on the ice for more than six minutes a game, he's not going to pan out in fantasy hockey.

ii. PP Time and special teams in general. Getting on the first power-play unit versus not seeing any power-play time at all could mean the difference between 45 points and 65 points in a season.

iii. Does the player even play the game? If the roster is healthy, some very capable rookies and journeymen wind up in the press box. You can stick a piece of paper in front of the coach's face outlining Player X's beautiful Corsi rating or low PDO statistic…but do you think he'll care?

iv. Linemates. I always refer to Steve Sullivan's time in Toronto. Advanced stats were prevalent back then, but you could just see that he was a talented and productive hockey player. I'm sure if you could look at the advanced stats, they would look pretty sweet. But he played a lot of hockey with Kris King and Tie Domi…so the production was disappointing. What if he got through waivers back in 1999? I'll tell you what would have happened - he'd be an AHL journeyman, like so many Keith Aucoins. But he didn't get through waivers, Chicago claimed him. And he went on to post 64 points in 73 games and never looked back.

***Note: get the latest line combinations - free - right here ***

v. Coaching Philosophy. Advanced stats may say that Craig Smith will improve his numbers significantly as this season moves forward, but chances are he will never be anything close to a point-per-game player in a Barry Trotz system. Not even a 0.80 point-per-game player.

2. Luck. This includes:

i. Injuries. I don't need to expand on this one.

ii. A single mistake. Example - zigging when he should have zagged, costs the team the game and puts the player in the doghouse. The event itself isn't luck, but rather a lapse of judgment on the player's part - which in itself is a skill (or lack thereof). However, the timing is luck. How the opposition responds to the mistake is luck. If they miss the net, perhaps all is forgiven.

3. Intangibles. This includes:

i. Heart. Find me an advanced stat that measures this?

ii. Clutch Performance. To measure this, you almost need to break off important games from regular games and analyze the advanced stats therein. A lot of players, over a large sample size, would boast the same numbers. Some players would actually decline for the bigger games. And some players - it's easy to guess who - would see a bump in their advanced stats.

4. Scouting. This includes:

i. evaluation of skating

ii. evaluation of effort

iii. evaluation of strength

iv. evaluation of shot

v. evaluation of defensive responsibility

If a scout feels that Player X can't skate, or is too small, then he'll never get a proper opportunity at the NHL level and it doesn't matter what his Relative Corsi number is.

Craig MacTavish said it best in Friedman's article:

"If you rely solely [on analytics], you do it at your own peril. Nobody's figured out a fail-safe method, not in baseball, football, hockey or soccer. But you're foolish not to be [looking at them], either."

As advanced statistics become more widely accepted and more widely used, their impact as a predictor will increase. Right now, the impact of advanced stats as a predictor helps - but is much closer to "nil" than many think. You can evaluate how good a player is in certain situations, but what good is that if the coach doesn't agree with you?

Use these metrics to back up your argument, or your beliefs. But use them the same way you would use other numbers, such as "goals", or "ice time". If you spend too much time and put too much emphasis in possession numbers and zone starts, you'll make yourself look foolish. It's a balance. Don't lose sight of the big picture.

If you do this "balancing act" right, then with each passing year the (let's say) 10% edge that advanced stats give you will become 20% because coaches are starting to buy in. From Friedman's article we see that Boston, Dallas, Edmonton, Tampa Bay, Vancouver and Washington are already using advanced stats in some capacity. When all 30 teams use them, the value to poolies will increase. But advanced metrics will never be 100% accurate. And thank goodness for that because we'd all have to quit fantasy hockey. 

Follow me on Twitter - @DobberHockey

Comments (10)

Subscribe to this comment's feed

Dobber said:

... Tiger - All three of your points, I don't understand...Some problems that I have with your points: 1. TOI/PP - that has nothing to do with anything I said. Nothing. I'm referring to time on the ice, period. While some hardcore Adv. Stats followers shout Corsi at me, I look at the player and see he played 8 minutes. His Corsi can be sparkling, but it matters little if the coach decides not to play him. 2. PDO measures luck very well. But can it predict a broken leg? That's the kind of luck I was referring to. I think my article was pretty clear on that point, but you didn't catch it. 3. Intangibles are the most relevant thing in fantasy hockey, when it comes to keeper leagues and prospects. I like advanced stats and they help me a lot. But you're arguing with my article as if I'm against them. I'm just preaching caution that they should not be used exclusively... and that those who do use them need to completely understand their limitations. Jamie - my point on clutch is being overblown. I list 50 reasons for 'caution' and not to take advanced stats as the bible. Clutch is just one of those reasons. So we're talking about 2% of my point. A small factor, but collectively taken with the rest of them makes a huge difference. Ryan - your quote "I understand that you shouldn't lean 100% purely on stats, but I think they should play a much bigger role than most poolies give it credit for." Fully agree. But my point is about advanced stats and how the basic stats are being ignored in favor of advanced stats. Some argue how great a player is because of this Corsi number and this Zone starts number vs. Zone finish...and they don't see how the guy has three points in 20 games. So they should play a much SMALLER role then certain poolies give them credit for. The conclusion you should draw from this article is: If you don't pay enough attention to advanced stats, then you need to. Now. And if you pay too much attention to advanced stats to the detriment of the basic stats, then you need to dial it back.

report abuse vote down vote up

March 09, 2013
Votes: +0

TigerUnderGlass said:

... A couple problems I have with this article: re: TOI/PP - advanced stats takes this into account much more completely than any other method of evaluation. It's the reason so many are calculated on a per/60 basis. Most advanced stats are alse very specifically divided by game situation ie. 5X5, 5X4, 4X5. re: luck - advanced stats are really the only solid means of identifying luck, good or bad, so I don't understand why you'd identify luck as something they don't cover very well. re: Intangibles - highly irrelevant to fantasy hockey, plus one has to actually believe in clutch first. Even still, in the end a persons "clutchness" or lack thereof, would be easily identified by most advanced metrics if it existed.

report abuse vote down vote up

March 08, 2013
Votes: +1

MikeTSchmidt said:

... I'd agree that there are very few advanced metrics that offer much of anything for fantasy hockey enthusiasts. For example, Mikael Granlund's Corsi numbers match what you can see on the ice -- he's not ready for prime time. Watching games will tell you more than enough about the guys on your fantasy roster. But there are a handful of stats that can give you an edge over the owners in your league who completely disregard them. PDO is a nice, simple number that HELPS (key word) an owner understand luck and statistical regression. Zone starts are a good numerical indicator of a player's role and how he's being used by his coach. I even find it helpful to look at a player's performance against even strength to see how it compares to his overall production. Is it better? Is it worse? How does it compare with how he's being used by his coach? More information (as long as its properly understood, contextualized and weighted properly) is always better than less information -- especially when being used to make decisions on value. No one should make all their roster decisions based on advanced stats. No one should completely disregard them, either. You are absolutely right when you say it's a balancing act.

report abuse vote down vote up

March 08, 2013
Votes: +0

Jamie @ Fantasy Hockey Geek said:

... That's a great point, Ryan. While you make some good points, Dobber, I don't think fantasy hockey players should generally concerned, for example, with whether a player is "clutch" in key games. For a fantasy hockey player, it doesn't matter whether with a player scores 5 points in a big game or a mean-nothing game. In that respect, we can afford to approach player evaluation from a much higher level than "real life" GMs.

report abuse vote down vote up

March 08, 2013 | url
Votes: +0

Ryan Ma said:

Mixing between "real-life" and "fantasy hockey" Also another big factor that you have to take into consideration is that the people you quoted (Beane, Burke, MacTavish), have a much different prerogative than us poolies. They're in the "real-life" scenarios, of everyday hockey, they see each other at the rink, they travel together, they eat together, they see the on ice difference that a person who has heart does for a team. They can see a direct effect that a big hit or a fight does to change the momentum of a team. So I can fully understand why they would say that "stats isn't the be all and end all." Because there are much more intangible factors behind success in "real life" hockey than there is in fantasy hockey. However, in the realm of fantasy hockey, we're all about the numbers. We don't measure "heart", we don't measure effort, strength or defensive responsibility. Which is where I think you're devaluing the how stats play a role in fantasy hockey. Fantasy hockey is all about stats , real-life hockey takes more into consideration of the "bigger picture." At the end of the day, fantasy hockey is about who gets the most goals, most assists, the better GAA, or SP... That's what helps you win. That's where advanced stats can help you build a better picture of what's going on. I understand that you shouldn't lean 100% purely on stats, but I think they should play a much bigger role than most poolies give it credit for. Once you have the stats, looking at the interpretation of them is the key.

report abuse vote down vote up

March 07, 2013
Votes: +0

Ryan Ma said:

... Well I'm big into my stats, the three that I use are ice-time, SOG and PP TOI (now I'm transitioning over to percentage of team power play opportunities). The thing with stats is that they are quantifiable, and if used correctly can be used as an explanation as to why things are occurring. I've have a few numerous debates with Dobberities (MT319) regarding this factor, the thing is how do you take "heart", "skating mechanics", "offensive talent" into account into statistics? Those are all qualitative factors, how do you judge someone has more heart? How do you judge that player X has better skating mechanics as player Y... You're arguing from your point of view of interpretation, and it's hard to develop a solidified argument. You can't however argue from numbers, you can't argue that player X is garnering 20:20 in ice-time with 4:02 on the power play... Granted you can always fudge the numbers to make it represent what you want it to say... But if used correctly they are a very powerful tool to help you gain a "bigger picture". The big thing that I 100% agree with angus about is their predictive value. If someone is shooting at 35%, but they for their entire career is shooting as 12%, well what should the stats tell you moving forward? So yes they can tell you what happened, now what the challenge for us big into those stats, is understanding why they happened and if it will continue to happen or will it change... That's the job for us, or Fantasy Hockey Power Play (plug!), to interpret. Advanced stats are great, but as Angus said, the three that is probably the most important is TOI, SOG and PP TOI, as all three of those factors have a huge correlation factor to point production.

report abuse vote down vote up

March 07, 2013
Votes: +0

angus said:

	... The big question I have with the advanced stats and fantasy hockey is their predictive value. Advanced stats are great at telling us what happened, but can they tell us what will happen? I have enjoyed diving into this field over the past two years. I use it for fantasy hockey, but not significantly. I still rely a lot on ice time, shooting percentage, and so on.
	report abuse vote down vote up

March 07, 2013
Votes: +0

Pengwin7 said:

Excellent Well written, one of my favourite articles this year. I like the order in which they are written too. Time-on-ice (TOI) is indeed the starting block for player evaluation. How is he doing in his even-strength time? How much capacity does he have to play? Most top forwards play 20min/game, most top defense play 25min/game. Knowing how much a player is currently getting vs. what's left is key. I also focus on even-strength production which correlates with the "capacity" to increase production by getting that PP-time (gravy). And linemates are KEY factors that are often overlooked. Some people like to say "JoeX can produce all by himself". For the elite, like Crosby, true. For 99% of hockey players, linemates can have as much as a 20pt impact on scoring. I think advanced statistics are going to help coaches figure out how to best use their players in roles, most defensively. I don't think any coaches are too far out-to-lunch with how to find playing time for their most offensively talented players... so there really isn't much, IMO, that advanced statistics can help us with. Really well written here. A+, HUGE thumbs up.

report abuse vote down vote up

March 07, 2013
Votes: +0

Dobber said:

... Very true, but that also ties in with 'clutch' - which won't be reflected in stats of 'regular' games but rather in 'key' games. Also, anyone throwing around the word "heart" is referring to evaluating the potential of younger players - so the sample size is smaller and with increased ice time and improved responsibility the 'heart' factor comes to the forefront - and only at that point would it get better captured in stats.

report abuse vote down vote up

March 07, 2013
Votes: +0

Jamie @ Fantasy Hockey Geek said:

	... An argument that many people fail to see is that, ideally, stats should take into account intangibles. That is, if a player has 'heart', or is 'clutch', or is a good skater, or whatever, those intangibles should naturally be reflected in that player's stats, both basic and advanced stats. Otherwise, if a player's stats aren't boosted by his "heart", what value is that heart?
	report abuse vote down vote up

March 07, 2013 | url
Votes: +1