SalarySwishSalarySwish
Avatar

The_Ultimate_Pielord

I put math in hockey
Member Since
Jun. 1, 2018
Favourite Team
Carolina Hurricanes
2nd Favourite Team
Arizona Coyotes
Forum Posts
624
Posts per Day
0.3
Forum: Armchair-GMNov. 29, 2019 at 9:32 p.m.
Thread: GO HABS GO
Realistically, Hall this year has a value greater than Breadman last year. Comparable scoring rates over the last few years (in the midst of this down year, Hall's just 3 points off a ppg pace with a shooting percentage less than half his career average. If he scores at more typical rates for him, he'll be a ppg player easy) Hall's analytics are also better than Breadman's and he's got 1 more Hart trophy on his resume. Panarin banked an 11.5 million dollar contract after EvolvingWild's model predicted an 11 million dollar deal. Projecting Hall anywhere under ten million is pretty ludicrous.

I question the decision to ditch Kulak and Tatar. Kulak's put up solid offensive and defensive results over the last few years. He's nobody's idea of a star, but he's a heck of a lot better than, say, Alzner, and is definitely the team's best defensive LHD.

Tatar, meanwhile, is a very solid forward on a decent contract. If you need to clear cap space and ink a big name like Hall, it probably makes more sense to pony up the assets to dump the big bad contracts (Alzner, Price on the off chance anyone will even consider it, maybe Weber) as opposed to dumping 2 guys who's combined GAR at evens since the start of 2017 is actually higher than Hall's (Hall beats them combined by his very impressive 11.9 Powerpaly offence number.)

Montreal could definitely use Hall. I don't think this is how they get him.
Forum: NHLNov. 20, 2019 at 5:48 p.m.
Thread: Someone explain why Puljujarvi is so coveted on here
<div class="quote"><div class="quote_t">Quoting: <b>Brian2016</b></div><div>In terms of a trade for draft picks, what exactly is his trade value? Is he worth a late 1st/early 2nd rounder at this point?</div></div>

I'd probably do anything outside the first round. Pick value decays exponentially, so basically any pick outside the first 20ish is essentially the same. It's a really big market inefficiency, teams can add a ton of pick value by trading all their 2nds and 3rds for 6ths and 7ths.

And I mean, he's a 21 year old who's proven he's good enough to stick at the NHL level, even if he's not really a star. Maybe he can fulfill the potential that got him picked 4th overall. Maybe he just is what he is. But a young guy with some proven ability is always gonna be worth a pick for a team with issues in their forward depth.
Forum: Armchair-GMNov. 20, 2019 at 5:43 p.m.
Thread: Stamkos worst plus minus on team again
<div class="quote"><div class="quote_t">Quoting: <b>Bf3351</b></div><div>You can get your point across with like 300 less words. I'm sorry, what player in recent history has won the selke trophy that hasn't put up a lot of points. Your opinion does not matter. It's the way it is.</div></div>

You're right that points are generally a necessity to win the Selke and without a PHWA membership there's not a ton I can do.Doesn't mean I have to like it tho.

5v5 plus minus does have it's defenders but IMO there are are better options out there. GF% does the same job better in most situations since it controls for TOI and avoids over/under-rating players based on how high/low-event they are, and then it's always a good idea to control for context if you're looking for individual ability. FO% isn't completely worthless in terms of tactics, but it's not a big enough impact to be a significant part of a player's game. The 164 win marker is a little flawed, some faceoffs have more value than others (again, FO% is mostly useful for tactical choices as opposed to roster moves), but if a guy's only skill is winning faceoffs, you shouldn't go hunting for him (looking at you, Jim Benning).

Random, what are you using to get the charts to show up on CF?
Forum: Armchair-GMNov. 19, 2019 at 10:09 p.m.
Thread: Stamkos worst plus minus on team again
<div class="quote"><div class="quote_t">Quoting: <b>Bf3351</b></div><div>Everybody who wins a selke puts up a lot of points dude... Putting up points while being incredible in your own end makes that player even more special because it's hard to do. Kopitar could be the best defensive forward in NHL history if he dropped offense and just shut down the other team...</div></div>

"The Frank J. Selke Trophy is an annual award given 'to the forward who best excels in the defensive aspects of the game.'"-NHL.com <a href="https://www.nhl.com/news/nhl-frank-j-selke-trophy-winners-complete-list/c-287904026?tid=287709666" rel="nofollow noreferrer noopener" target="_blank">https://www.nhl.com/news/nhl-frank-j-selke-trophy-winners-complete-list/c-287904026?tid=287709666</a>

Points are not an aspect of defense.

I will die upon this hill, possibly wearing a Mikko Koivu jersey. I blame the Younggrens.

<div class="quote"><div class="quote_t">Quoting: <b>Bf3351</b></div><div>Ok, but that doesn't make the stat useless. Let's talk about Dion Phaneuf. They've been on good defensive teams in the past, yet horrible defensively and there's a big fat minus next to their name all the time. Does that make +/- irrelevant? +/- isn't always cut and dry. It takes someone with common hockey sense to fill in the blanks. Doughty for example. One of the best defensemen in the league when it comes to his defensive ability. Yet he had a horrible -34 next to his name and if you watched him he deserved most of those minuses by not playing well. I take the entire Kings team in to consideration as well as the terrible partner he had named Derek Forbort. If +/- was pointless it wouldn't be an NHL stat.</div></div>

The issue is that RAPMs do the exact same job but better. So do variants of 5v5 GF%. There's no reason to use +/- with the stats we have now.

A lot of the classic NHL boxcars are weird. Why do the first 2 passers and the goal scorer get credit on a goal, but nobody else on the ice? Why not just the first pass, like in basketball? Why factor special teams goals in only one direction? Why include empty-net stuff at all? These ideas face less scrutiny than any component of RAPMs or even GF% or CF%.

+/- is better than nothing, but it's a LOT worse than modern regression models. It'll catch some players, but it'll miss some too (Taylor Hall's always been good defensively, but his plus/minuses have typically been very bad). It's not quite in "even a broken clock is right twice a day" territory, but there's no reason to use a clock that ticks erratically when a better clock is right there.
Forum: Armchair-GMNov. 19, 2019 at 9:49 p.m.
Thread: Stamkos worst plus minus on team again
<div class="quote"><div class="quote_t">Quoting: <b>hockeyfanatic05</b></div><div>So when Kopitar won it the season he was below ppg, did he still win it based on offense?</div></div>

Compared to competitors, yeah. The league leader in EVD GAR among forwards in 2015-16 was Pavel Datsyuk (at age 37, oh my god that man), who put up 25 fewer points than Kopitar. After him was Nino Nieddereiter (31 below), Patrik Hornqvist (23 below) and Mikko Koivu (34 below). Even if you want to make the argument that the first 3 weren't used as true shutdown forwards, Koivu definitely was. Kopitar, meanwhile, ranked a majestic 56th in EVD GAR, between Nikolay Kulemin and Carl Soderberg.

His most recent Selke win the leader was Koivu, who he had 47 points on. After Koivu were Coleman (67 point difference), Danton Heinen (45 points) and Riley Nash (51 point difference). The leader among non-Koivu guys with &gt;1500 all-sits minutes was Henrik Zetterberg (35 point difference). Kopitar ranked 51st, tied with Blake Comeau and Charles Hudon.

All of this is with a TOI cutoff of 1000 all-situations minutes unless otherwise stated.
Forum: Armchair-GMNov. 19, 2019 at 9:36 p.m.
Thread: Stamkos worst plus minus on team again
<div class="quote"><div class="quote_t">Quoting: <b>The_Ultimate_Pielord</b></div><div>Don't cite trophies as evidence of anything. Kopitar won a Selke for his offence. Vasilevskiy's Vezina was almost literally a participation trophy, he got it for being the goalie for the Lightning more than his actual play. And, most relevantly, we want evidence that is relevant to every player in the league, not the tiny fraction that get these trophies.

The better evidence for +/- based logic being wrong is that +/- is a bad stat that tells us more about the situation a player is placed in than anything about the player themselves. ROR was always a very good player, but in Buffalo was forced to eat tough minutes alongside awful teammates. He's a very good 2-way centre (20th in EVD GAR since the start of the 2016-17 season), but not on the level of a guy like Mikko Koivu (whose EVD GAR is double O'Reilly's over that timeframe). For those wondering, he ranked 13th last year among skaters with over 1000 minutes, while Mark Stone ranked first.</div></div>

Whoops, that was even-strength minutes, not all minutes. ROR actually ranked 31st among skaters with over 1000 minutes, with Stone continuing to hold the top. Interestingly, Jesperi Kotkaniemi ranked #3 when I changed the cutoff. Weird how he didn't really get any Calder buzz.
Forum: Armchair-GMNov. 19, 2019 at 9:31 p.m.
Thread: Stamkos worst plus minus on team again
<div class="quote"><div class="quote_t">Quoting: <b>hockeyfanatic05</b></div><div>Ryan O'Reilly is a career -23, but just won the Selke. So your logic is wrong</div></div>

Don't cite trophies as evidence of anything. Kopitar won a Selke for his offence. Vasilevskiy's Vezina was almost literally a participation trophy, he got it for being the goalie for the Lightning more than his actual play. And, most relevantly, we want evidence that is relevant to every player in the league, not the tiny fraction that get these trophies.

The better evidence for +/- based logic being wrong is that +/- is a bad stat that tells us more about the situation a player is placed in than anything about the player themselves. ROR was always a very good player, but in Buffalo was forced to eat tough minutes alongside awful teammates. He's a very good 2-way centre (20th in EVD GAR since the start of the 2016-17 season), but not on the level of a guy like Mikko Koivu (whose EVD GAR is double O'Reilly's over that timeframe). For those wondering, he ranked 13th last year among skaters with over 1000 minutes, while Mark Stone ranked first.
Forum: Armchair-GMNov. 19, 2019 at 9:23 p.m.
Thread: Stamkos worst plus minus on team again
<div class="quote"><div class="quote_t">Quoting: <b>BOLTLOVER1</b></div><div>Agree on plus minus relative to team .. players and goalies matter alot .. never compare absolutes across NHL .. but plus minus valid within team .. with asterisk for special teams I guess</div></div>

It'd be better as a rate, since guys can't control how much their coach wants to play them, and it'd be better with more context than just relative to team (teammates within the same team can change a LOT. Just look at the with/without Mark Stone numbers for the Sens last year). So we need to analyze 5v5 and special teams differently, factor in contextual factors and change it to a rate: that creates the GF RAPM (and expected goals against RAPM for defense, since the difference between expected and actual goals is the goalie's responsability).

That's the big issue with +/-, even with some controls, it's just a lesser version of other stats. There's occasionally something to be gleaned in the difference between a player's isolated impact and what results on the ice, but for the most part RAPMs are just always better.
Forum: Armchair-GMNov. 19, 2019 at 9:12 p.m.
Thread: Stamkos worst plus minus on team again
<div class="quote"><div class="quote_t">Quoting: <b>BOLTLOVER1</b></div><div>Finally a good reply Pielord .. thxs.. some valid .. some not .. assuming you are familiar with statistics and sample size and the randomness of numbers .. thing such as goalie, # games etc.. even out for everyone over time with enough inputs .. and even minutes should not matter .. one thing you mention above is PP and PK do count but in inverse way .. did not know this. So more PP time leads to potential more short handed goal which does count as minus .. interesting</div></div>

Yeah I have no idea why the NHL does the special teams thing. The volatility of goals does even out over time, but it can take a while. 3-ish years of data is usually the baseline for saying a major difference between goals and expected goals is real and not just a result of luck. Minutes ends up mattering just as a result of scale, if you have a guy play 3000 minutes and his team's +1 goals per 60 over their opponents with him on the ice and another guy gets to +1.5 goals per 60 but only gets 1500 minutes, guy 1 has a +/- of +50 (3000/60*1) while guy 2 has a +/- of +37.5 (so 37 or 38) (1500*60*1.5), despite the fact that he won his minutes by more than guy one.

Quality of teammates can have a huge impact on performance. Tomas Chabot, an excellent young defenseman, is currently rocking an ugly 40.5 GF% at 5v5 this year, with a 45% CF% and 48% xGF%. He's an excellent young defenceman by both the analytics and the eye test and helps his team a lot, but the current Sens are beyond saving, especially when he's been shackled to the great boat anchor that is Nikita Zaitsev. Some guys can carry awful teammates and still achieve success (like Mark Stone), but it's not infrequent for star players to be underwater by raw on-ice metrics because their teammates are too bad to carry.
Forum: Armchair-GMNov. 19, 2019 at 8:59 p.m.
Thread: Stamkos worst plus minus on team again
Stamkos's 5v5 expected goals against RAPM this season is about half a standard deviation below league average, which is an improvement from where he's typically been the last few years (about 1 SD below average). Not good, but at least close to average. Interestingly, his corsi against RAPM is holding pretty steady at about 3/4 of an SD below average, right in line with his averages over the last few years. So he's doing a better job of restricting quality of chances, while being about the same in the quantity he allows.

The poor +/- seems to be more to do with less plus than more minus. He's been below league-average in goals for, expected goals for and corsi for. Part of this is less shooting luck, Stammer's usually had elite goal generation with only decent to good chance generation, and goals are super volatile, but it's still concerning. Especially since Stamkos is about to turn 30, he's been elite his whole career, but time hasn't lost one yet.

He should be fine, but I've been saying the same thing about the Leafs all year and they seem determined to prove me wrong, so yeah. Not ideal.
Forum: Armchair-GMNov. 19, 2019 at 8:49 p.m.
Thread: Stamkos worst plus minus on team again
<div class="quote"><div class="quote_t">Quoting: <b>Bf3351</b></div><div>Are you stupid? When someone is a minus multiply times and by more than -10 it's not a coincidence. They are bad defensively.</div></div>

+/- is SUPER usage dependant, for a few reasons.

1. It's essentially an on-ice stat with no controls for quality of competition (impactful for certain players), zone starts (rarely a major factor, but relevant. Not really gonna hurt Stammer though), score effects (relevant), or quality of teammates (very relevant.)
2. It's strength controls are dumb. PP goals for don't count as pluses but shorties against do, so PP time can only hurt your +/-. Similarly, PK time can only help it (shorties for count but PK goals against don't.) The big one though is that 6v5 GA (goals against with an empty net) count as normal, so players with significant 6v5 icetime get REALLY heavily hurt by it.
3. It's a count stat, not a rate or percentage. This means high icetime is required for significant swings either way. The worst players in the league are usually close to average by +/- because they don't see much icetime, while the bottom of the ladder is populated by guys like Edmonton-era Taylor Hall, star players on crap teams. The fact that such players tend to get lots of PP and EN icetime doesn't help matters.
The other big issue over a small sample is that it's a goal based metric, which makes it really vulnerable to wild swings as shooters/goalies run hot and cold. There's a reason one of the first and most prominent luck stats in hockey (PDO) is just on-ice shooting% + on-ice save%. Goals are super volatile over any sample smaller than 3-ish years for an individual player.

I'll look up some other numbers on Stammer's D this year and see how he stacks up to past years. It's entirely possible he hasn't been good, but PLEASE stop using +/-. 5v5 GF% (the percentage of goals scored that were scored by a player's team with that player on the ice at 5v5) does essentially the same job but cleans up some of the usage stuff. Regression models (like Evolving-hockey's RAPMs) are even better, with the added bonus of separating out offence and defence.

Use better stats!
Forum: Armchair-GMNov. 17, 2019 at 5:32 p.m.
Forum: Carolina HurricanesNov. 17, 2019 at 2:28 p.m.
Thread: We Do Not Need A Number One Goalie
<div class="quote"><div class="quote_t">Quoting: <b>Earfgang</b></div><div>Okay first, a .3 goalie and another .3 goalie don't get you to a .6 goalie, every game would be played at a .3 Wins Per Ten pace so obviously you would be at .3 wins per ten for all 82 games. But this theoretical stat game is a zero ends game because i can give you stats that say one thing you can refute with a different one that says another. SV% doesn't account for quality, Wins above expected doesn't give enough credit to goalies with defensively solid teams, and doesn't take in to account quality, GAA is literally just a team stat, WAR is OK but oversimplifies whats going on and punishes goalies in good defensive systems. Plus all goalies are unicorns, defining them as if they will act consistently year round is disingenuous.

But on another note, the Ducks are proof that over work is definitely a thing, Gibson is the best goalie in hockey period in my opinion, and early in the year the Ducks ride him to being competitive in the early playoff race. But after January his stats begin to really fall off because its not just a volume of games played it is a volume of shots he faces and a very uniquely high volume of quality chances to score against him. Basically Gibson faces a constant barrage of rubber to the face every night for 60+ nights unimpeded by his offensively focused defense. Under this system the Ducks backup gets absolutely torched in the fire that only Gibson is able to handle. Gibson under the Islanders system (Assuming his play remains just as good which isn't true but theoretically speaking) his numbers would be godlike. He's not a better goalie in New York, he is very much a product of environment. The Islanders have a very competent defense and don't let quality chances through often at all. Thus, the reason that you see Robin Lehner, Thomas Griess, and now Varlomov are all putting up numbers that are equal too Gibson's (SV% wise, GAA, GPG, Wins). All flawed stats in their own right but so is Wins above expected. These goalies aren't better than Gibson in any world but their environment makes them play at a level that is equal too a Gibson. Gibson would put up a .935 sv% in NY, Lehner put up a .931% Griess put up a .927%, basically, by putting their money and focus on building up a strong defensive system they have given themselves a goalie that is equal to what Gibson would do every night. This is accomplished by understanding the mechanics of being a goaltender in the NHL. Most guys that are good in the NHL can stop the shots a NYI goaltender faces on a nightly basis, no goalie on the planet not named Gibson or Jesus himself can put up the performances Gibson puts out on a nightly basis in Anaheim.

What I just described to you is the future of hockey, by letting Lehner walk in free agency the Isles demonstrated they understand what they've done, they have created a system where all they need is a goalie that is positionally sound and they will preform the same as if they had Carey Price. Thus they can spend less on a goalie tandem than anyone else have it play better than anyone else, and invest the extra Draft capital and extra money capital on maintaining a good defensive core, and to pay Barry Trotz whatever the hell he wants to keep running it.

Im not arguing simply that two decent goalies are better than one Amazing one, cause thats not true, im arguing that understanding the goalie position and practically eliminating the high danger scoring chances by bolstering your defense will produce a tandem of goalies that, if they are positionally sound, will both play at a level equivalent to a Gibson or a Price in their systems which often leave them out to dry.

This is better for playoffs because your season can't be jack knifed by one player being injured, if one of your goalies gets cold and starts letting in weak pucks you have another that will work equally well, and you have more money and draft picks that you spent in making your defense better so your team can theoretically be better in the process. This is economics merging with hockey to make our game more efficient and smarter, this is the future of the game.</div></div>

Wins above expectation isn't an actual stat (at least not what I was using), it's an entirely theoretical construction I made for the thought experiment: a perfect goalie WAR model with no flaws whatsoever, so that the quality of the goalies could be a constant in order to discuss how that quality being broken down would be. The idea of having the sum of the tandems was to show the total ability of the tandems. You're right that assuming consistency from goalies is a bad plan and that evaluating goaltenders is analytics hell, I conceded those at the end when I said that teams shouldn't invest a lot in goaltending for those reasons, but you'd still rather have a #1 goalie than a platoon.

If a goalie is better entirely because of a different environment, that just means the goalie has an easier job, not that he's become any better/worse. Therefore my theoretically perfect WAR model would still give them credit for the same amount of wins above expected.

I get the point about how a strong team defence can inflate goaltending numbers and make great goaltending unnecessary, and you're right that generally it's more efficient to invest in good skaters and trust that the goaltending won't completely screw you over. But iffy goaltending can definitely doom even teams with elite skaters, the Carolina Hurricanes have learned that lesson the hard way many times with the Cam Ward Experience (and the .888 incident).

Again, my point wasn't that the Canes should give Gibson an 8x8 offersheet, it's that the Canes would right now be better off with a true #1 goalie than with their current tandem situation. They can win with this setup, but it'd still probably be worth making a low-risk trade for Raanta or Kuemper to get a proper #1 if the Yotes are willing to do so (they might be).

Yeah, killing the high-danger stuff really helps a goalie, but Minny's done that better than any team in the league the last few years and Dubnyk still managed to cost them a playoff spot. Goaltending definitely matters.
Forum: Carolina HurricanesNov. 17, 2019 at 2:15 p.m.
Thread: We Do Not Need A Number One Goalie
<div class="quote"><div class="quote_t">Quoting: <b>Earfgang</b></div><div>But if we are talking wins above expected if you have 2 goalies with a .5 WAE vs one with a 1WAE and his partner with a -.5 WAE in a 45 game split they are equal, and favering more games towards your number 1 would make it better than my split. This number is flawed obviously but i do understand your logic mathematically</div></div>

The point was if the total talent level was close to equal, you'd rather it be concentrated into one goalie than spread across two. The example you gave isn't equal talent between the tandems: the even tandem has a total WAE of 1 (0.5+0.5=1), while the unequal one's total is 0.5 (1 + -0.5=1-0.5=0.5). The example I was using had a WAE of 0.5 for the unequal one and 0.6 WAE for the equal one.

There is a limit to how much total talent loss a team should accept in the name of having a true #1, but if the total talent of the tandem is close to equal you're better off with one good/one bad than with two decent.