A Commissioner’s Manifesto (Part 3 – Stat Categories)

by Johnny Fantasy

Setting up a fantasy hockey league is no easy task if you want to do things right.  While it’s impossible for fantasy to exactly track reality, I always wanted to find a league setup that did the fairest job of approximating reality and that would provide the best experience of managing a hockey team of NHL players.   To that end, there are the some basic principles I’ve settled on over the years.  In part 1, we looked at the different platforms available (we settled on Yahoo and the rest of our discussion will focus on that); in part 2 we looked at the different league formats.  Here in part 3, we’ll look at stat categories.

Choosing Statistics Categories

Regardless of what format you’re playing in, the whole feel of the game will come down to the stat categories that are chosen, whether you win by category or you assign fantasy points for each stat.  This is, by far, the most variable aspect of fantasy hockey, and this is often the greatest source of debate among commissioners and players.  This is also an area where the default settings are often inadequate.  While we’ll go into detail about the right balance of categories (with respect to stat category leagues) and point values (with respect to fantasy-points leagues) in future posts, the first question is: does each category even belong in the fantasy game?  Is this something you should get credit for when your players do it?  Let’s break down each of Yahoo’s category options.

Goals (G) – completely uncontroversial.  This category belongs in every league.

Assists (A) – basically uncontroversial.  This category should show up in every league (unless you are playing a stat-categories league and trying to weight goals heavier by including goals and points but not assists).

Points (P) – inexplicably controversial.  I don’t know why, but points are not one of Yahoo’s default stats.  In a fantasy-points league, I suppose this doesn’t matter since you’ll be weighting your goals and assists anyway (although, I still prefer to include it (and adjust goals and assist fantasy-point values accordingly) for the ease of reviewing stats – having the G-A-P stat line gives you the scoring info you need without having to add in your head each time you look at a different player).  But in a stats category league, I have no idea why you wouldn’t want to include another category the rewards actual scoring, especially if you are including several peripheral stats (like shots or PIMs).  I think this category should be in every league (be it for convenience or for the value of having another scoring category).

Plus/Minus (+/-) – inexplicably uncontroversial.  Plus/minus is a stat that has gotten a lot of criticism in respect of real (non-fantasy) statistics.  A great article explaining the history of this stat (and some choice words from GM Brian Burke) can be found here, but to summarize, the argument is that plus/minus is not a great stat because it is only correctly awarded (i.e. given to someone who actually had an effect on the play) about 66% of the time.  So, while it is more often correct than incorrect, it’s pretty inaccurate and it has been falling out of favour, especially among the growing analytics community.  Given that, I’m surprised that this stat is a category in so many leagues (and is even a default stat in Yahoo).  And yet, it is a stat is derived from actual scoring, and so it can be valuable in fantasy, especially in stat-category leagues to balance out peripheral stats.

And speaking of peripheral stats, let’s jump into those. One of the basic premises of the analytics movement in hockey is that goals (and thus points) are rare events, statistically speaking, and are therefore not the best way to evaluate how a team is playing (or, more precisely, predict how they will perform in the future).  Scoring goals matters at the most at the end of the day, but teams can be said to be playing well, and players can be contributing to their teams, even if they aren’t generating goals or assists in any particular game.  Fantasy should reward this in some respect in order to acknowledge good play and to make fantasy more interesting by including more statistics to consider.

Shots on Goal (SOG) – One of the most fundamental peripheral stats is shots on goal.  Indeed, much of the analytics movement is based on shots on goal (Corsi) (and attempted shots on goal (Fenwick), which is not currently an option for fantasy).  This stat should be a category in pretty much every league (unless you are doing a bare-bones league with literally no other peripherals).

Penalty Minutes (PIM) – PIMs were the original ‘toughness’ category.  Yahoo did not originally offer hits or blocks as categories, and so PIMs filled in to represent the physical aspect of the game.  This obviously is a logical problem because, while physical players are important to a team (theoretically…), actual penalties are bad – they result in power plays, scoring opportunities and goals for the other team.  So-called ‘good penalties’ (where an illegal physical play is warranted because it prevents a probable goal) may exist, but they are few and far between and the overall PIM total doesn’t distinguish between those and dumb or careless penalties that are detrimental to a player’s team.  Additionally, as Down Goes Brown recently addressed, PIM totals get inflated by misconduct penalties.  Players who can’t control themselves and get thrown out of games shouldn’t have their fantasy value increased as a result.  While PIMs served as a decent enough “this is the best we have” physicality stat before Yahoo introduced hits (and blocks), now that we have those other categories, PIMs are really obsolete and shouldn’t be used.

Hits (HIT) – While we’re on the subject of ‘physicality’ stats, let’s talk about hits.  Now, hits are not an exact statistic – what one scorekeeper considers a hit may not be considered a hit by a scorekeeper in another arena.  Additionally, there seems that there is a home arena bias, where scorekeepers are sometimes more generous to the home team.  Hopefully recording standards for hits will be improved (or at least otherwise compensated for).  Despite these imperfections, hits are a much better stat for measuring physicality than PIMs.  They track actual physical play, rather than serving as a poor proxy, as PIMs do.  Interestingly though, hits do not seem to have any meaningful correlation to winning or scoring.  Does this mean that hits have no place as a fantasy category?  That is a fair argument, although physical play is a big part of the sport of hockey, and hitting is the epitome of that.  Hitting may be best off being used as a low-value category in fantasy-points leagues – giving a little bonus for physical play, but not to the extent that big hitters who struggle to crack 60 points on a season end up in the top 15 skaters.  Hopefully recording standards for hits will be improved (or at least otherwise compensated for – see this article from NHLnumbers for some interesting stats and ideas:

Blocks (BLK) – The other ‘toughness’ stat.  Overall, I think this is an underutilized stat.  More and more, we are understanding how important preventing shots on goal is to a team’s success.  Blocked shots brings one aspect of shot suppression into fantasy relevance.  Additionally, using blocked shots adds value to defensemen (who typically block more shots than forwards, which can offset there generally lower scoring stats).  The one downside is that this is very much a peripheral stat, and dilutes actual scoring categories in stat-category leagues.

Faceoffs Won/Lost (FW/FL) – Faceoffs are certainly an important part of any hockey game.  Teams that win more faceoffs get possession of the puck more often, which means more shots, scoring opportunities and often goals.  In fantasy-points leagues, these categories can be appropriately scaled to add a bit to the game.  However, if they are given too much weight (or if they are included in stat-category leagues), they can really skew things to favour centres.  Leagues that provide a ‘Utility’ slot will have that spot unofficially transformed into another centre slot since managers will be keen to try to win the faceoff categories.  This will also have the effect of overvaluing centres that also have wing position eligibility.  And of course, this is another peripheral stat, so you have to worry about diluting actual scoring stats in stat-categories leagues.  Overall, I think faceoffs are fine to include in points leagues if their fantasy point values are sufficiently limited, but they seem to be more skewing than valuable in stat-categories leagues.

Some of you may have been thinking: “Why did we jump into peripheral stats already?  What about special teams?  Those are scoring categories!”  Well…

Power play goals/assists/points (PPG/PPA/PPP) – Power play categories seem to be generally unquestioned as good fantasy categories (especially PPG or PPP).  Perhaps this is because they are one of Yahoo’s default stats and managers are used to it.  Perhaps because they are a popular statistic cited for NHL players, and because it is important for teams to be able to capitalize on their opportunities to score goals.  But when it comes to fantasy, we have to ask ourselves: why should a player who scores on the power play be valued higher than a player that scores at even strength?  Unquestionably, the players that score more often should be valued higher than players that score less.  And certainly many of the high-scoring players get a significant chunk of their points on the power play.  But from a fantasy perspective, whether by having a PP stat category, or providing extra points for PP scoring, we are giving players a bonus for scoring when it is easier for them to score.  Should Vladimir Tarasenko (37 G, 8 PPG) or Max Pacioretty (37G, 7 PPG) be valued less than Joe Pavelski (37G, 19 PPG)?  If anything, Tarasenko and Pacioretty should be valued more than Pavelski because they are scoring points for their teams when it is harder to do so (at even strength, rather than on the power play when they have an advantage).  Indeed, that is one of the basics of scoring the plus/minus statistic – you’re supposed to being scoring on the power play; you shouldn’t get bonus points for doing so.  The only argument I can see for their inclusion is to balance out a large number of peripheral stats in stat-categories leagues, but in that case I think it’s probably better to just reduce the peripherals.  Power play points don’t always occur proportionally to points in general, and thus they increase the arbitrariness of a player’s value.  All other things being equal, Tarasenko or Pacioretty are better goal scorers than Pavelski, because they can more often get the job done regardless of whether they have an advantage on the ice.

Shorthanded goals/assists/points (SHG/SHA/SHP) – If there’s no reason to reward a player for scoring when they have an advantage, then surely scoring when at a disadvantage should be worth something extra, right?   Well, generally yes, but the issue with shorthanded scoring is that it is such a rare event that it basically turns into a random stat category.  In stat-categories leagues, teams often win shorthanded scoring categories 1-0, because someone on their team happened to factor into a shorthanded goal that week.  You know how many shorthanded points the league-leader in that category scored last year?  6.  The year before?  6. That’s roughly once every 13-14 games.  Don’t get me wrong, having players that can score while shorthanded is great, and a shorthanded goal can turn the tide of a close game.  Including a ‘random X factor’ like that can add a fun element to fantasy hockey, but in general randomness is not a fair representation of play.  Like many of the other well-meaning categories we’ve looked at so far, I think shorthanded scoring categories work best in a fantasy-points league, where a modest bonus can be awarded for scoring shorthanded, but these essentially random events are not turned into game-changers.

Game-winning goals (GWG) – When it comes to arbitrariness, I’m not sure that there is a better example than game-winning goals.  The phrase ‘game-winning goal’ brings to mind images of sudden death and that ‘clutch’ player doing what they do best and coming through for their team when they need it the most.  Except that’s not how game-winning goals are awarded.   Sure, overtime winners will always be game-winning goals, but you know what else will be?  The 3rd goal scored by the winning team in a 6-2 win.  This is the case regardless of whether the teams were tied 2-2 at the time of that third goal.  A team could go up 6-0 in the first period and then surrender 2 goals in the third, and that third goal will still be the game winner.  So what this comes down to is complete randomness,  with players that score more goals being more likely to score GWGs (as we would expect, since more goals means more chances for them to be game-winners).  Ironically, game-tying goals (if they were recorded properly in the modern NHL) would be a much better stat to capture ‘clutch-ness’ (if such a thing really exists), since a game-tying goal would always be a goal scored to prevent a team from losing.  Like shorthanded scoring, this category could be used in stat-categories leagues to balance out scoring and peripheral stats, but it would still not be ideal to do so since this category is so random.

Shooting percentage (SH%) – The final skater category is the first that is not a cumulative stat (i.e. it is not an accumulation of events, but rather a ratio of goals to shots).  Given that, it is only available on Yahoo as a category in stat-categories leagues.  Shooting percentage can be nice because it rewards goal scoring (and again, could be a category to offset peripherals), but it undermines shooting.   The result is that two players who score the same amount of goals but have different amounts of shots will be valued equally (they would tie for goals, one would win SOG, the other would win SH%), and this can’t be right.  Of these two players, the player generating more shots should be valued higher, because he is creating more scoring chances for his team.  As a result, shooting percentage shouldn’t be used as a fantasy category.

Wins/Losses (W/L) – Wins are a terrible way to value goaltenders.  Hockey is a team sport and the amount of wins that a goaltender can accumulate is wholly dependent on the team they play for.  Carey Price and Henrik Lundqvist would have comparatively abysmal win totals if they played for Edmonton or Arizona.  Mediocre goalies can have stellar careers behind dominant teams (we can call that the “Chris Osgood effect”).  When you’re drafting a goaltender based on wins, you’re basically drafting the team he plays for.   Even fantasy football, where you are literally drafting an entire team’s defense, doesn’t use wins because wins in nearly every sport are based on both offense and defense.  The Contrarian Goaltender has a lot of great posts about evaluating goaltenders, but he spells out the point simply – it is harder for goaltenders on some teams to earn wins than goaltenders on some other teams. Losses are of course the opposite side of the same coin.  Admittedly, most goalie statistics are flawed and affected by the play of the team overall, but wins is the most egregious category in this respect (and is made even worse in terms of arbitrariness by the shootout).  Thus, wins and losses should not be used to award goaltenders in fantasy hockey.

Shutouts (SHO) – Okay, goalies don’t usually single-handedly win games, but what about shutouts?  In that case the goalie has played perfectly and deserves some extra credit.  While I generally agree with that sentiment, shutouts suffer from the same problem that shorthanded scoring categories do – they don’t happen often enough to be considered anything other than random.  In stat-categories leagues, shutouts are typically won 1-0 by the team that happened to have one that week. Like shorthanded categories, this can introduce an element of excitement into fantasy – having a player that can do something to effect a big change in the fantasy score.  But, in general these kind of random categories are not preferred, especially since a goalie earning a shutout will be rewarded in through almost every other possible goalie category.

Games started (GS) – On first glance, games started may seem to not be a great statistic.  Why get points just for showing up?  But, games started does reflect an important aspect of goaltending – endurance, durability and consistency.  A goalies that can give you 60-70 starts during a season is certainly more valuable than one who can only give 40-50.  Further, in most circumstances goalies earn their starts.  A coach is not going to play a goalie who is performing poorly when he has the option to give the backup a chance to do better.  Granted, many starting goalies are firmly entrenched and have a long leash before anyone would consider them at risk of losing their starting job to their backup.  But, that does happen (look at Winnipeg last year, when Pavelec essentially lost the starting job midseason to Hutchinson, only to regain it once Hutchinson hit a rough patch).  Games started does value a goalie’s endurance and consistency, and it’s a cumulative stat, which we will see is harder to find for goalies.

Goals Against/Goals Against Average (GA/GAA) – The next logical step would be to say “okay, let’s value the goalies who let in the least number of goals.”  This is certainly better than wins, but pure goals against is difficult to use in fantasy because it is a cumulative statistic where the lowest number wins.  A manager could simply decide to play only one goalie for one game in a week and have a good chance at winning the category.  Goals against average is better in this respect, since it is a ratio rather than cumulative stat, but still suffers from a similar problem – if a team’s first goalie appearance is good, then the manager is incentivized to bench its goalies for the rest of the week and not risk increasing its GAA.  This could be mitigated by requiring a minimum number of goalie appearances per week, but then teams are at risk of losing all goalie categories if they can’t put together enough appearances (which could happen due to injury, a hot backup challenging one of their starters and/or if certain teams just don’t have many games in a given week).  Additionally, goals against and goals against average are flawed measures of a goalie’s performance because they do not account for how many shots a goalie faces (which of course the goalie has no control over).  For stat-categories leagues, we need to pass on these and find a better category.

Shots against (SA) – This is dumb.  This is not the number of saves a goalie makes; it’s just the number of shots a goalie faces.  Since goalies have no control over this, this should not be used under any circumstance.  If your commissioner chooses this stat category, start thinking of ways to fleece them in trades.

Saves (SV) – Okay, this is a little better.  At least the goalie isn’t getting rewarded for shots that go in the net.  The problem remains though that goalies have little or no effect on the number of shots they take, so this stat is going to be completely subject to the play of the rest of the goalie’s team.

Save percentage (SV%) – So what is left?  How do we value goalies?  Well, let’s put the role of a goaltender simply: how good is a goalie at stopping the shots that he faces?  The imperfect answer to that question is save percentage.  This is imperfect because save percentage is affected by power plays, where the chances of scoring are much higher, even though a goalie has no control over how many penalties its team takes (other than the rare goalie penalty).  Perhaps one day Yahoo will offer 5-on-5 save percentage as a category, which is generally seen as a better measure of an individual goalie’s performance, or some other proposed goalie statistics that are designed to filter out background and situational effects but until then save percentage seems to be the best fantasy has to work with.

But, a problem with save percentage is that it is a ratio and not cumulative statistic, and so we will have to deal with the minimum appearances issue or pair it with another, cumulative goalie category to provide an incentive for managers to play their goalies for the whole of each week.  Shutouts would be great, since it reflects a goalie’s efforts, but as discussed above shutouts are too rare to work well.  Wins, shots against and saves are at the opposite end of the spectrum, with the goalie having little control over the result.  That leaves games started, which at least doesn’t suffer from the same kind of team bias that the other cumulative categories do.

There’s another problem however, since save percentage is a ratio based category, it is only available in stat-categories leagues in Yahoo.  Fortunately, we can essentially replicate it in fantasy-points leagues by using some of the otherwise detestable statistics above, namely goals against and saves.  Assigning the right values (including negative fantasy points, in the case of goals against) can yield a scheme where goalies with high save percentages earn big points, and goalies with low percentages earn few, none or even negative fantasy points.

Advertisements

4 thoughts on “A Commissioner’s Manifesto (Part 3 – Stat Categories)

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s