I was listening to an episode of my favorite podcast: You Are Not So Smart the other day, and his guest was Google director of People Operations (Human Resources): Laszlo Bock. You should go listen to the whole podcast, because it was really fascinating. One part that caught my ear was when Mr. Bock was talking about hiring, and how managers hire within Google [20 minute mark]. Their research shows that college sophomores can make the same judgement on an interviewee that trained HR professionals can within just 10 seconds of watching the interview. That is astounding. What Google has found is that that first impression is so strong that the rest of the interview, the interviewer is confirming the biases they've formed within the first 10 seconds of the interview. They are still professional and get good answers and good results from the interview, but their personal bias is so subconsciously overwhelming that they interpret the information they've gathered to fit their bias.
"The single biggest thing, the most valuable thing you can do, is- whoever's doing the interviewing, and especially the hiring manager, do not let them make these decisions" -Lazlo Bock
He then goes on to talk about Google's hiring process. They have interviewers, and then they send it to a hiring comitee. So they solved it with a number of ways. Potential candidates go through a number of interviews. Their answers are compiled. It's given over to an HR comittee that has never met or seen the candidate does the hiring. They've found this to be an extrodinarly effective way of evaluating talent.
So the thought hit me: how can this apply to hockey?
To me, interviewers are very analagous to hockey scouts. That talent evaluators basically make up their minds on people within 10 seconds of seeing them is why I am extremely skeptical of scouts opinions on players. This is true for talent evaluators, and it's almost certainly true for hockey scouts as well, since they are, in fact, human. (This goes for people on the internet whose primary way of evaluating players is watching them too). Scouts aren't useless, and I want to make sure I don't shortchange what they do. There are still people whose inputs I greatly valiue, and who have a lot of insight. The draft vets this out. Better players are generally taken at the top of the draft. Just like the interviewers: Scouts are still good at their jobs and they still pick out talent. It's just that they also let their biases cloud their ability to finally judge them.
So how could hockey apply these principles for talent evaluation? How could hockey take advantage of a scouts knowledge and insight, but try and strip out the bias that clouds their judgement? A place to start, IMO, is follow Google's model as closely as possible. So here's how I'd apply these few things to hockey (ok. I'd do a lot more research before doing this, but this is a start based upon these interviews and other behavioral science principles):
• I'd send scouts to view prospects multiple times. Seperately. I wouldn't really let them converse with each other, and I'd attempt to keep them away from media coverage (and corrupt their own evaluations with others' biases). They would all be responsible on every prospect they've viewed at least 5 times. (obviously potential 6th round picks wouldn't have 20 scouts worth of data, but as much as possible) We'll call these scouts write-ups.
• I'd have an analytics department gather as much data on these prospects as possible. All the relevent stats possible. (and I'd try and see if someone's tracked the advanced ones for all these leagues) These would be a player's analytics write-up.
• I'd have a 3rd group of people combine the analytics and scouts write-ups together. Then, they would strip out as much of the biased language (big body, good motor, etc) as possible. I'd even have them strip out size (size bias is rampant). The player would be marked by "player A, position and handedness" This would be the final write-up, and hopefully it's stripped out the scouts AND analysts biases.
• I'd then send it on to a 4th comittee (maybe GM and Asst GMs, or maybe not). This would be analagous to Google's hiring comitee. They would take these full write-ups and rank the players based on those final write-ups. Then reveal the palyers names on the list.
There's your draft list. Choose the top one available.
There's probably a lot of tweeks to this that could be applied (there's a lot more behavioral science that could be applied). But this, to me, is a starting framework to start applying the very valuable stuff that's being discovered in behavioral science to hockey.
This would take a giant leap of faith because what I just laid out above lays out something that's almost crazy when it's said a different way: The guys deciding who to draft are not going to be watching those players play! That's a concept that's quite a bit different than what hockey does now. I don't see hockey culture being anywhere near ready for this. Hockey culture isn't really known for being on the innovative heels of Google. Any organization that did this would face intense scrutiny (especially from a hockey media that revels in the pride of it's own crustiness).
But, I think the behavior science in hockey is worth investigating, and the steps I laid out above, modeled after Google's hiring process, based on behaviorial science, is a way to get a more comprehensive evaluation process.
This could be a doozy, so we'll jump right in, because there's a lot I want to cover.
Most of the regression talk with the Avs comes in three forms: Goaltending/sv%, Corsi/Fenwick, and Shooting %. There's a couple more areas I don't thik have gotten a lot of ink, so I'm going to talk about them, but I'll start with Shooting %:
Shooting %: Shooting % is a stat that really goes all over the place, from year to year for nearly every team. It's a touchy subject, because the Avs (and pretty much every team who has a solid sh%, really) make the claim they are taking better shots. The Avs were 2nd in the league at 8.77% shooting percentage (5v5 close), and many stats folks have that coming down.
But there is some evidence that the Avs actually do have some shooting % skill. They have some elite offensive shooting talent (Duchene, MacKinnon, O'Reilly, and Tanguay) and there's some numerical evidence too: they appear to be good at Zone Entries, which lead to better quality shots.
I think they will regress some, but maybe not as much as predicted. For a reasonably optimistic (and to avoid contraversy) let's say they don't regress here.
One-Goal Games - The Avs had a +30 goal differential last season but 112pts. +30 is normally reserved for teams in the 100 point range, not so much in the 110 point range. This led me to believe the Avs had some fortune on when they scored goals too. Sure enough, the Avs were a league leading 27-4-8 (.700) in 1-goal games last season. That's a ridiculous record (and completely unsustainable). [The image in that link is broken, but I think the r^2 of .05 makes the point] (Here's another link with a broken link to the really good stuff). (One more link on one-goal games)
If the Avs fall to a .500 record in 1-goal games next season that would be similar to Carolina (20-9-11). That's 51 points rather than the 62 the Avs got last year, an 11-point drop, which, isolating all other affects, would put them at 101 points instead of 112. Much more typical of a team with a +30 GD.
That's a fairly significant point: The Avs could do everything the same as last season and still drop 11 points (or more, if they get some bad luck) just on when their goals happen to come, i.e. when they are able to pull out a 1-goal win.
If you're counting at home, I've put down 7points to goalie regression & 11 to 1G game regression. That's down to 94 pts without shooting regression.
Injuries - I've seen people say "don't forget, Tanguay is back too" when discussing the Avs forwards corps, but really I think the Avs had a very fortunate spell on injuries last season. Here's GP for the Avs top 9 players:
Duchene - 71
Landeskog - 81
MacKinnon - 82
O'Reilly - 80
Stastny - 71
Erik Johnson - 80
Barrie - 64
Parenteau - 55
Varlamov - 63
The Avs basically had 2-3 important injuries last season. Among the Avs most important 7 players (Stats, Duchene, MacK, ROR, Landeskog, EJ, Varlamov) they suffered two minor injuries (Duchene and Stastny) and no major ones. Among their second tier of plyers they did have some significant injuries (Barrie, PA, and Tanguay missed time, and Hejda played injured). That no major pieces of the team were injured, and there was only one really devestating injury, a 3rd liner, is incredible luck.
And injuries are terrible, because they affect everything. An injury to the best player, say Duchene, affects possession and sh%.
It's possible the Avs can go another season without injuries to top guys. (After all, the injury dice have no memory), but do I think it's likely? No.
Which brings us to:
The Avs have young players who will likely get better: Duchene, O'Reilly, MacKinnon, Landeskog, EJ, & Barrie. Now maybe one of those guys stalls a little but yes, they are going to get better. This should really help the Avs possession, and these guys, as stated above, are the Avs key players.
But the Avs are really counting on some old guys too. Iginla, Stuart, Briere, Hejda, Tanguay are not the most important players, but they are pretty important players. They are all on the wrong side of 33, and injury is a big risk with these guys. But so, too, is regression. In fact all of these players have shown signs of regressing already, and one or two may fall off the cliff this season. Regardless, if we're prognosticating young players getting better, it's fair to prognosticate old players getting worse.
Again, the older players are not counted on as much as the young players, and the young players growth will probably is better than the old guys depreciation, but both need to be taken into effect.
Honestly, I don't think Conf III is as good as being projected. Chi is still great. StL limped into the post season, possessionally, and is going with Allen & Brian Elliott as their goalies (Picking up Stats & Gunnerson helps them a lot). .i not convinced. Dallas seems legitametly better (& good). Minnesota had miserable possession numbers, too, last season (worse than the Avs). Preds are the opposite Avs - killer top 3 D , 1 forward (Neal). And Winnipeg still has Pavelec.
The only team in the west who I think is not getting the credit they deserve is Vancouver, who isn't what they were, but were also unlucky last season (and were saddled with Torterella. How'd Alan Vigneault's team do?). The Cali teams are good.
So, my predictions: Top 7 spots in the west:
Chi, Dal, StL
Last WC: dogfight for the 8-seed with Minnesota, Nashville, Phoenix, & Col.
I think COL is the better of those 4, but not by a lot. A bad shooting season or a really significant injury & the Avs fall below all 3.
Going into this preview i was thinking COL-88pts... but i don't think the West is that good outside the top 6 and the Avs should improve.
Final prediction: 94 pts & a 1st round exit to the Hawks.
(I wrote a good portion of this on a tablet, with no spell check. Please forgive any typing or spelling errors)
Key Additions: Brad Stuart, Zach Redmond
Key Subtractions: Andre Benoit, Corey Sarich
The defense is actually a place where I think the Avs improved over a season ago, the problem is, it's a slight improvement and not nearly enough for a team with a defense that is, legitimately, thought of as one of the weakest in the league.
Let's throw some Zone Exit data out there, curtesy of Corey Sznajder (whose doing some AMAZING work over at his Hurricanes blog this summer, and has now been hired by an NHL team). Here's the Avs through the first half (give or take) of last season.
(note: Success% is successful attempts to exit the zone)
First, Let's talk about Erik Johnson: He faced some of the toughest competition in the NHL last season, and ended with a pretty successful CorsiRel of 2.4, and a point line of 9-30-39. That's tremendous. He may not ever be in the same class as Doughty, Keith, Subban, Chara or Weber, but I think he's one of the best 30 defenders in the game, and is a legitimate #1D. He's got a very good Exit Success % (See Above: at least from my limited sample of what I've seen). He's very good.
The problem is: he needs a partner. Jan Hejda is not it. That exit data is brutal (The Sharks data, below, gives his more context). Brad Stuart is not the answer either.
In fact Brad Stuart isn't a very good defenseman anymore. He faced the equivalent of lower 2nd line competition and wasn't used super defensively, slightly more than average, and got buried at ES relatively. The nice part about Brad Stuart is that he pushes other inferior defensemen out of the lineup. The Avs probably won't see 68 gp for Nate Guenin, so that makes them a better team instantly. But Brad Stuart wasn't a top-pairing defenseman in his peak, and he's well past that now. Here’s his Corsi w and w/o last season (Zone start adjusted):
Here's a text link, for blowing it up
And Stuart's data for the two years prior:
And, here's the exit zone data for the Sharks (again, this covers the first 1/2 a season, give or take a game)
Hejda and Stuart are not very good at getting the puck out, marginally better than Scott Hannan. I'm not sure when Hejda broke his hand, and obviously that would affect his ability to get the puck out of the zone, so maybe he gets some slack here. Stuart has no such excuse, though. Stuart is prone to turnovers (more than any Av or Shark not named Ryan Wilson). He doesn't get the puck out well. His one positive is he doesn't ice it. It appears that when Stuart was with the Red Wings 3 seasons ago, he was a pretty decent defender. He wasn't that player for the Avs. Most likely; he got old.
I don't think I've given Nick Holden enough credit. He played decent competition and held his weight, and has terrific underlying numbers. I'm ok giving Holden-Barrie the second pair (and dropping Hejda to 3rd pair). That said, Holden did play in 54 games last season. He should be getting more ice time, and play in 20 more games, so that should improve the Avs defense, some.
I <3 Barrie
If the Avs had a real solid top-pairing guy, I'd probably like their D ok, but as it is there's just a giant hole next to EJ that no one on the Avalanche roster can adequately fill. And Guenin being projected on the 3rd pair is still a disaster. He's not an NHL defenseman.
That said, I think the D is a small improvement. A full season of Nick Holden will help, Brad Stuart isn't very good, but he's a lot better than Guenin, and probably slightly better than Hejda. I don't know what to expect from Redmond (My draft of this had the wrong first name and mispelled last name) and he only played 10 games for Winnipeg last season. He's kind of a Benoit replacement who can PK I guess. Hejda starts the season without a nagging injury. I don't like Hejda-Stuart 1st PK at all.
So, small improvement in the D over last season, but still a not very good one.
Part I was Goalies. This is part II - Forwards
Key Additions - Jarome Iginla, Danny Briere, Jesse Winchester, Ben Street
Key Subtractions - Paul Stastny, PA Parenteau, MA Cliche
Edit: for some stupid reason, I thought the Avs did not re-sign of Cliche. Apologies for the error.
Let's start the position players with the elephant in the room: The Avs Corsi-tied last season was 22nd in the league. Their Fenwick-tied was 23rd. That's not good, and while it's possible to overcome a possession deficiency, it's extremely hard (and usually requires a good dose of... favorable random chance, which the Avs got in spades last season). The Avs needed to increase their possession game in order to maintain some of their success, and it's pretty hard to make the claim the Avs got much better at forward this offseason.
What I see from a lot Avs fans saying: Iginla replaces Stastny, and Briere replaces Parenteau, who wasn't that good anyways
I see something a lot different
Jarome Iginla plays wing. Iginla played perfectly fine minutes for Boston. He was on a line, almost exclusively, with Milan Lucic and David Kreiji, and they weren't on the ice against the toughest players the east had to offer (that honor went to Patrice Bergeron, Loui Eriksson and Riley Smith, with a little Brad Marchand thrown in too). In the last three years he's put up 76-85-161 in 204 GP (.79ppg).
Jarome Iginla is a winger who played against top line competition but took a lot of starts in his attacking zone, relative to his teammates. He had ok possession numbers, but nothing that jumps off the page. He's a possession positive, or at the very worst not an anchor. He had a very good 30-31-61 points line, and he was paired up almost exclusively with David Kreiji and Milan Lucic at even strength. In the last three years he's put up 76-85-161 in 204 GP (.79ppg).
Paul Stastny is a center. He was dynamite for possession and Stastny racked up his points (25-35-60) against some of the toughest competition in the NHL. The Avs forwards who played the toughest minutes last season were: Ryan O'Reilly, Gabriel Landeskog and Paul Stastny. (Apologies to Mitchel and Talbot, who also had tough minutes, but against lesser competition). In the last 3 seasons he's 55-82-137 over 190gp (.72ppg)
Jarome Iginla isn't taking those minutes, especially the defensive ones. He's not a center, and he's not really capable of it. Whose going to replace those minutes in the defensize zone? Ryan O'Reilly already had a difficult workload, does he double down on it? How will that affect his offensive game? John Mitchell is a fine player, underrated in fact, but is he really that shutdown guy? Does anyone think MacKinnon is ready to take those tough draws in his own end against the Toews, Koivus and (sigh) Stastny's that lurk in our division?
I don't see anyone on the Avs roster who can replace what Paul Stastny was brining to the team.
PA Parenteau is a winger. He He played tough competition, (both years), but he was slightly sheltered in terms of zone starts. He was terrific possessionally. Last season was a down season (14-19-33 in 55 GP). But over the last three years he's got a line of 50-93-143 in 183gp (.78 ppg). PA Parenteau may be the most underappreciated Av. He had a fantastic two year spell with the Avs, including a better PPG with the Avs than: Claude Lemieux, Gabe Landeskog, Chris Drury, Adam Deadmarsh, Ryan Smyth and Pierre Turgeon. While those guys aren't HoFers, that's a pretty impressive list. Yes, he had a down season (0.55ppg) last season, but he battled injuries and at 31 with three other very productive years prior it's fairly likely he'll return at least mostly to form next season.
When I look at the possession numbers I see: Iginla is an upgrade on Parenteau and no one replaces Stastny. That's a lot less palatable. You could probably argue that's unfair, because we're comparing only last season to this season, but then that changes to:
Iginla replaces Stastny's offense, No one replaces Stastny's defense, and who replaces PA?
Which brings us to Danny Briere, who had a terrible season (13-12-25). He played middling competition, but did so often from his own zone. He was ok possessionaly, in that role and he is a good bet to slot into a 3rd line with Mitchell and Talbot, I think. In his prime, Danny Briere was another player I think was overlooked in his career. But I don't think he's going to be able to replace PAs contributions to the team. And at 36 it's significantly more likely that he'll never recover from his poor season. Ove the last three seasons he was 35-55-90 over 173 gp (.53ppg). The absolute best case is he almost covers for the loss of PAs career worst season. At 36 the most likely scenario is he bolster a weak bottom-6, and not much else.
The one clear win for the Avs is replacing reducing MA Cliche's role with Jesse Winchester. Cliche may have been the worst non face-punching regular NHL forward last season. Winchester maybe be a below ave 4th liner, but he's an upgrade over Cliche. That said, the 4th line gets, what, 7 min a night? Downgrading in the top 6 and upgrading on the 4th line is still a significant downgrade.
Overall, I think the Avs forward corps took 2 steps backwards this season, and I don't think they're going to be able to improve on either their Goals For, Goals Against, Corsi or Fenwick next season. For a team in the 20s in possession, that's big trouble.
(Yes, there are young players on the Avs that will improve. I'll get to that in the Part - IV 'Others')
I've been insanely pessimistic about the Avs season all summer, and I have only revealed the source of my pessimism in bits and pieces over twitter. So I figured I needed to give a comprehensive guide to why I don't think the Avs are going to be very good this season: (and why I'm going to enjoy the season anyways). This is intended to be a 5 part series, but it could balloon to six or drop to four depending on how much I have to say. The intended parts are:
I - Goaltending
II - Forwards
III - Defense
IV - 'Other'
V - Recap
So here we go:
Key Additions: Reto Berra (kind of, he was here last seaosn but played 2 games).
Key Departures: JS Giguere
Semyon Varlamov: Is coming off of a great season professionally, in which he was second in Vezina voting and sported a sparkling .927 sv% (including a .933 ES sv%). At 25, this should be a cause for optimism, right?
Well, yes and no. He went a long ways to establishing himself as a bona-fide starting goaltender in this league, and one a team can lean on (this was very much in doubt before last season). But expecting him to repeat his perfomance from last season is unrealistic. League average sv% for starters (35+ GP) last season was .9167%. This includes other goalies such as: Henrik Lundqvist (.920), Jon Quick (.915), Ryan Miller (.918), Luongo (.919).
If Varlammov regresses to a .919 sv% (which would be above average for starters): That's 17 more goals given up next season, on regression alone. It's important to note that Varlamov may not regress to there, it's entirely possible that he regresses only a little... but it's equally possible he could also have an 'off' year next season too (his sv%'s the three season's prior to last were all at or below league starter ave). Goalies are fickle and unpredictable: there's too much random chance in goals (even over the course of a 2,000 SA season).
My baseline on Varly: an ~17 GA drop next season, a .919 sv%. That would still be an excellent season for him. (Of course the error bars on that are large. 1 st dev on average is a range of .910-.923. For the Avs that would mean a GA of 35-9 more. 35 is terrible, nine isn't so bad.) Still, I think everyone would be happy with Varly in the .919 range.
Reto Berra: The most contraversial backup goalie in the league. The avs paid a steep price (2nd rounder) for a guy who has not performed well at the NHL level. He's got a career sv% of .893 over only 31 career starts. Now many people have pointed out, that's a very small sample size to judge from, but really bad players don't usually get to have enough games to have a large sample size, either. And Berra's international appearences for Switzerland and in other leagues don't inspire much donfidence that he's very good either.
So he's an unknown, and Allaire and Roy are using him as a replacement. It's also worth noting that if he doesn't cut it, he'll likely be replaced by Aittakaillio or Pickard.
Last season Giguere had a sparkling .913 sv %, which was also good enough for "below average starter". Even if Reto Berra improves/regresses 12 points in save %, that's a net loss in the backup goalie position. How much?
Giguere faced 608 shots last season. Replacing that with a .905sv% adds 5 to the GA column, not terrible but that translates roughly to 2 points in the standings. (again, that's assuming Reto Berra/backup by comittee can improve a whopping .012 in his sv%, which is pretty optomistic for anyone to improve)
So, I expect -22 in Goal Differential (with large error bars) based on goaltending alone (which translates to ~7 points in the standings). Again it's worth noting my regressions are, IMO, optimistic with assumptions that:
- Varly is an above average NHL starter
- The Backup Goalies performing well above their (admittedly Small Sample Size) career averages.
(A side note on goalies:
The season Varlamov had last year doesn't even guarantee he is an above average starter [I think he is, but it's well within the realm of possibilities that it was a fluke year]. If you are a betting man regarding who you think will turn out the best, I'd suggest using uk's leading sports betting website.
Here's a partial list of guys who had seasons as good as Varly (.014 sv% points above average): Dan Ellis, Manny Legace, Rick Dipíetro, Ron Tugnett, Dan Ellis, Manny Fernandez, Arturs Irbe, Roman Cechmanek. Of course, Henrik Lundqvist, Dominik Hasek, Roberto Luongo, Tim Thomas all have those kind of seasons too. The point is: having one great season isn't really uncommon for mediocre goalies.
Very good goalies do it a lot. Varly had a very good season, he needs a couple to be in a spot where you can pen him in for an above ave sv%. Goalies are fickle. Never ever forget this. )
Those who argue with Advanced stats usually do so because of the eye test. The general vibe given off in these conversations is that a trained eye is better at evaluation because it can pick up nuance that numbers don't capture.
Here's why I find "the eye test" extremely unreliable, in ove video about a card trick: