$50 well spent.

This summary is not available. Please click here to view the post.

My case for a second goalie spot

Kevin's analysis showed the relative value of goalies, under the current scoring system, to be reasonable and in the spirit of DO. I think, however, Matt's rationale for proposing the second goalie spot was not a complaint against the current G scoring and the market forces, but rather a comment on the mechanics of the head-to-head system and the roster. Matt, if I am mis-quoting you in spirit or in specific, jump on it.

In what is becoming a tradition, I will attempt to analyze some data while drinking. Matt's argument was that the single goalie spot exposed managers to the vagaries of game scheduling and starter-backup cycling. A skillful manager, nurturing their team to greatness with special care for the G-spot, would suffer unfairly when their teams' schedule had few games for their stud G, or worse.

Presented below is a summary of (almost) all the match-ups in the 2010-2011 season. For each weekly match-up, I calculated the difference between the number of games played by the goalie of Team A and the number of games played by the goalie of Team B. The column represent the frequency of wins (light bars) or losses (dark bars) for Team A in the weekly match-ups.

If the numbers of games started did not matter, then the distribution of the light and dark bars should be the same and centered around 0. Sadly, Team A's who had more goalie-games did significantly better – having a goalie-game in-hand in a weekly matchup increased the probability of victory by 90% (±44% for the 95% confidence interval). Look how often a team with a 2 goalie-game deficit won the match-up (pro-tip, look for the smallest light bar). And remember, these data ignore the performance of the goalies in each match-up.

Matt's proposal to add a second goalie spot would likely reduce the week-to-week variance in number of goalie-games. How much? I don't know. I could look at the games-played pattern for some or all of the skating positions, which would give an idea on the effect of scheduling (but not the backup issue). I can't seem to find a good source for those data from the league website (Kev?), so that might have to wait.

The current scoring system for goalies is balanced and fair, but I think the mechanics of the h-t-h system is distorting the ultimate value of goalies. A couple of weeks of bad luck with scheduling and a few backup starts and your team is in a big hole, even though your goalie is a 2.0 ppg. Let us design a system that rewards skillful drafting, crafty trading, and epic bullshit, rather than punishing teams for the vagaries of scheduling.

I support Matt's proposal to add a second goalie spot. I think it will diminish this games-started effect I describe here, and returning the focus of the match-up to the players and their performance. We will likely need to re-scale the goalie scoring to reflect the larger inputs of total positional scoring to team performance, but I think with the fantasy wizards McCartney at the helm, we will not be steered wrong.

My case Against a second goalie slot

One idea that was raised at the AGMM this year was to add a second starting goalie slot, bringing our total starting roster to 15 players.  I don't like this idea, and feel strongly (as strongly as people feel about fantasy hockey) that it will change the balance of the positions considerably, as well as the balance of scoring week-to-week.  I understand there to be a few specific reasons to add this slot, so I will address each as I understand them.  Please feel free to point out my mistakes, respond, critique, etc.  Let's get this conversation going.

1) Anyone can get a good goalie so there is no scarcity

I get that we don't plumb the depths of the goalie market as much as we do players. This year, only 5 goalies were kept by teams.  Another 14 were taken in the draft. But we are expanding, so next year we can reasonably expect to take 2-4 more goalies.  More importantly, we already have a substantial spread in the quality of goalies taken and started, which is strong evidence that scarcity is already at work.

By pure subtraction, 11 starting goalies are on the waiver wire.  In actual fact, I only see 6 true starters (and 5 platoon situations). These are for Toronto, Dallas, Winnipeg, Columbus, Florida and the New York Islanders. Only one of those scored more than 100 fantasy points last year and he can reasonably be expected to do worse this year (Kari Lehtonen).  There are flyers, sure, as Dubnyk or Mike Smith might start.  But as it stands, we have left the very worst goalies on the pile and still have a spread from Tim Thomas's 2.9 pts per game last year to Kiprusoff's 1.6.  In a three game week, that's a 4 pt per week spread in fantasy output - larger than even the difference between Ovechkin (1.9) and Steve Sullivan (0.8) last year.  And Steve Sullivan was once called the 'worst player on any team in the whole league' by Jason Bay.

As well, Tim Thomas led all players with his 2.9 per game average, and was only second to Daniel Sedin in total points (a Hart Trophy finalist on the President's Trophy winning team).  Goalies are valuable.  They have absolute value in the league as the highest scoring position and they have relative value to each other in that the top goalies score way more than the bottom goalies.  It already pays to have a good goalie, and some teams don't have one. That is the consequence of scarcity we desire.  It's hard to win with a crappy goalie, just as it's hard to win with a crappy LW core, or crappy D.

2) Maybe the worst goalies are bad, but the top 12 are roughly equal, so scarcity is still not working

Okay, lets just consider starters and ignore the outliers.  If we take Thomas out (because his season was record setting), and go from Luongo and Rinne (2.4) down to the 13th best goalie (a four way split at 1.9), we have a difference of 0.5 per game (or 1.5 per 3 game week).  If we go from Stamkos (second best C) to the 37th best C, we have a difference of 0.6 (or 1.8 per 3 game week).  Same if we go from the 2nd best D to the 49th best D - we come out with a 0.6 difference in per game average.  That's remarkable balance in starters.  Starters are just better than bench players, whether they are at Centre or Defence or Goalie.  LW and RW both have more outliers and so break the mould a little. If we remove the 4 guys at the top who are markedly better than the rest, the difference is again 0.6.  Some funny math, maybe, but just as Thomas skews the numbers for goalies, Hart nominees skew things for players.  Plus, that's why those 2 positions are so valuable.  We all know it.

The point is, this scoring system is remarkably balanced.  I dare say that it's the most balanced system I've played under in fantasy hockey.  If we add a goalie slot, goalies become much more important than they currently are, and honestly, I think much more important than any but the top 6-8 wingers plus Crosby.  I don't want the difference between a Vokoun-Miller pairing and a Backstrom-Montoya pairing having more sway on a week than how your other 13 player match up.  

3) Even the very best goalies are not the difference makers we expect. We can tweak the point scoring system.

I sort of heard this rumbling today and I was surprised.  The top 4 Vezina finalists from last year are the top 4 point getters by per game average.  We are accurately accounting for the best individual goalies, independent of team performance.  Isn't that what we want?  Goalies aren't solely responsible for Wins (a frequent problem in goalie scoring), and their job is to make saves, not make a better % of saves than a goalie he's not even playing against in real life.  Our scoring system is doing very well for goalies.

When I set up the league, one of the things I worked hardest to avoid is over valuing goalies.  This is one of the most common attributes of other fantasy hockey leagues: goalies are king, and goalies from good teams are better than goalies from bad teams.  In our league, this isn't currently true and I like that.


I feel like adding a goalie slot doesn't make more ways to win the league, it makes fewer.  I accept that some teams have a vested interest in generating greater value of goalies as they have invested heavily, but I don't think letting different teams win is the same as affording more teams the chance to win.

I officially vote against adding another starting goalie, and suggest that we consider other ways to create more scarcity at the position if that is desired.  Expansion is a start.

2010/11 Draft Review Part 6: Variance by Manager

In this series, we'll look at how valuable draft picks are by round and to each manager. Who does the best job of drafting, and how likely is any given manager to find a starter in each round?

Okay, so I skipped parts 3-5.  My SPSS license expired and my old excel won't do the graphs I wanted.  Plus, I use excel like Shawn drafts.  But the show must go on. Specifically, this show is about how well each manager did during the 2010 draft.

For the most part, these findings support Ben's Opportunity Cost Model, only with less depth and fewer interesting findings.  Still, this descriptive account is focused on how managers themselves did, irrespective of where their draft picks landed.  So instead of asking whether managers were able to maximize their assets, we're starting with the premise that where you picked doesn't matter.  Let's just compare how we did against each other.

To those who are prone to losing their confidence: you might want to look at this after Sunday's draft.

REMS To Make First Pick... again

For a second year in a row, the REMS have taken first round pick against the odds.  Last year, they beat out Santorum Sloppy Seconds to select Taylor Hall first overall (and later package him for Alex Semin).  Can we expect more youth from these improbable lottery masters?

1st Round Draft Order for 2011/2012:

REMS
Eschatologists
Ramotswe
Cats
Seamus
Guns
Stultifiers
Sloppy Seconds
Eschatologists (From Billcago)
Batarangs
ESR!
Batarangs (From Mercury's)

9 days and counting...

Draft Lottery

This is the Official Draft Lottery for the Definitely Offside 2011/2012 season.

You won't believe the result.

Trading Draft Picks

In the 2009-2010 season, 15 trades were made that involved draft picks.  In that first season, managers were fairly conservative in this area and particularly reticent to deal high round picks. Only 9 of those 15 trades involved a pick inside the top 6 rounds, and only 2 trades involved picks inside the top 4 rounds.

Immediately after the draft on September 26, 2010, I was asked to look into what those draft picks became so that we could retroactively evaluate the trades and perhaps even begin to settle the raging debate about the value of a draft pick relative to present roster players.  With more systematic analysis of the value of draft picks on-going, this review will focus on evaluating trades based on what those draft picks became. [Total fantasy points and/or Fantasy points per game average in brackets]

I've struggled all year to find a better way to present this data.  For now, you're stuck with a list of trades.  I've broken them up into different periods of the season to establish some context for the trade (ditching extra keepers vs. acquiring depth players etc.).  I've also listed the trades according to what each team acquired, rather than what each team gave away.  I've assigned (guessed at?) winners for the stats presented.  I've not considered whether players are/will be kept or their chance at getting better or worse.

So, you think you know hockey

In the interests of full disclosure, I sure-as-fuck do not know hockey. I might know a little about data analysis and in my now-travel-and-distraction-free life, I thought I would dig into last year's draft. Inspired by Kev's thoughtful analysis, I travelled a bit of a different path.

My take was to look at individual picks by their opportunity cost – the value of that pick relative to the most valuable player still available. I chose total point production over the entire season as the measure of value and scored opportunity cost as the difference between the value (i.e. TOT) of that pick and the TOT of the most valuable still available in the draft. Good picks, by my approach, have positive OC values because they had a larger TOT than any remaining players. Conversely, choosing a less valuable player contributes negative OC. I know, I know, opportunity cost is inherently negative, but I started this analysis halfway through a bottle of wine and couldn't be arsed to backtrack.

My first slice through the data treated managers' individual choices as picking the best player available, no matter the position. I haven't discussed drafting strategy with many, but this approach seemed reasonable to me. Skilled drafters would maximize OC by always choosing the best available player, right? The cumulative sum of the OC for all picks will therefore be highest for the best managers.

Well, we all suck, but with varying degrees of suck. Note the scale is in TOT, and remember the final PF for the last season ranged 1050 to 1350. Sullivan & Kev drafted well, according to this analysis, because of their relatively positive OC and the relative flatness of their drafting trend.

The general trend of the cumulative OC showed a general flattening as the rounds progress – in other words the high performers were being removed from the pool reducing the overall OC of missing good picks.

Another draft strategy is to pick by position, so I constructed a second analysis wherein OC would be calculated for each pick from the pool of available players at that position.

Forgive the busy graph, but I think there is some good stuff here. Each bar represents the average of a manager's picks for that position throughout the draft. The best picks would have positive (or less negative) bars. The lines show the range of OC values for each manager's picks at that position, so good managers should have both positive bars and small lines.

Again, we see Sullivan and Kev come out as relatively skilled drafters, with their nearly positive bars and particularly J's small lines.* And we can see who struggled drafting by position, like the Bealls at RW and Dale's woes at D.

Furthermore, the positional analysis shows the rewards of drafting at G. Many of the goalies drafted last year had very large TOT, and disproportionately contributed to the overall draft performance in this OC analysis. Leighton notwithstanding, 10 of out 15 G had TOT >100 and managers seemed to be better able to accurately estimate their value.

Finally, some thoughts on the value of drafting well. This OC analysis did not produce any good evidence for a link between draft performance and team performance. ESR & ESCH ended up at opposite ends of the standing yet performed relative well in the draft. And obviously, this analysis neglects lots and lots of details, like trades and injuries, so stay tuned for more refinements.

B

* I never thought I would be praising J's small lines. 

Test from France