Kris Letang vs Eric Desjardins

[Hockey Outsider]

You're mixing up two guys here.

Dom Luczyczyn has a proprietary formula he's developed, first for his own gambling and then writing for the Athletic. He does have cards, but probably not the one everyone is thinking of. He's the guy who writes the excessively hot takey articles, and I personally find that the more he talks in his own words, the less I like anything he has to say.

Jack Fraser, or JFresh, doesn't seem to do any advanced math at all that I can identify, and instead curates statistics work done by Patrick Bacon (whose formula is also black box as far as I can tell) and Corey Sznajder (who just watches and counts things, very easy to understand) and then does the graphic design work. He also does informal fan polls on Twitter to quantify public opinion. He appears to be moderately fluent in most areas of hockey discussion (stats, Xs and Os, fan trash talk) but not really an expert in any of them.

I think the think 70s was referring to with Bergeron at 100 and OV at zero was directly from SportLogiq and not one of these writer/creators.

I double checked and you're right. My comments (mostly) apply to Luczyczyn, not JFresh. Thanks for catching that.

I think my comments about how people blindly follow the cards applies to both of them. (But in fairness, it's not necessarily the authors' fault that people misinterpret their work).

 

[Michael Dal Swolle]

The issue I have with JFresh is, as far as I'm aware, he's never released full details on how his method works. We're essentially supposed to accept the results as a matter of faith.

I've been critical of Point Shares (from hockey-reference.com) - but at least they're transparent enough to explain their methodology. That way, when there are nonsensical results (Phil Housley > Doug Harvey), we can talk specifically about why the formula doesn't work.

I've seen dozens of examples on HFBoards where people post a JFresh card not even to support an argument, but as the argument in its entirety. I know his cards haven nice looking graphics, but it's the opposite of intelligence to quote an opaque, indecipherable system, and pretend that it's authoritative.

Generally, people who create all-in-one stats systems (such HR.com, Iain Fyffe, Alan Ryder, Bill James for baseball, etc) tend to be transparent about limitations in their model, and highlight any key assumptions. JFresh has done very little of that. I can only speculate that he doesn't want to be too self-critical of his own work, as he's actively trying to sell monthly subscriptions. Ryder and Fyffe showed their work because they wanted to convince others that they were the best. JFresh doesn't want to challenge the authoritativeness of the product he's selling.

And make no mistake - it's a good business model. JFresh releases slickly-produced cards, and writes articles designed to engender controversy (ie why Draisaitl shouldn't have been a Hart finalist in 2020). Then he sells data and articles in exchange for a monthly subscription. I can't fault JFresh. He seems to have found a way to monetize hockey analytics in a way that few others have. But I've never taken his cards seriously due to the lack of transparency, and the obvious conflict of interest.

According to my statistical analysis, Connor McDavid has the worst Flongle Score Card™ in the league and should be going unclaimed on waivers. Tyler Myers on the other hand has one of the highest Flongle™ Scores of all time and should be running away with the Norris vote.

Obviously, I won't tell you what Flongle Score™ measures or how it's calculated because that's proprietary, but you should definitely keep paying me $10/month on Patreon.

 

[seventieslord]

Alexander Barkov
Mark Stone
Anthony Cirelli

Auston Matthews
Mika Zibanejad
David Pastrnak

Alexander Ovechkin
Patrick Kane
Phil Kessel

analytics says that one of these groups of offensive forwards is in the top 5% defensively, another is in the bottom 5% defensively, and another is approximately average defensively. You already know which group is which.

I thought we were done shaking our fists at the clouds about analytics years ago. Of course we trust what our eyes tell us. We can also trust what stat trackers' eyes tell them, as they watch what happens on the ice and then record what happened. Where do you think statistics come from, anyway? Why do you think analytics and "eye test" analysis match up so well nowadays? They're so closely aligned it's honestly getting boring. At this point I'm only really interested in odd cases where analytics and perception/eye test differ, and understanding how/why that managed to occur.

 

[Johnny Engine]

So, while doing a bit of catch-up on some threads and remembered this one, and some Twitter drama a couple of weeks ago that clarified some information on what we were just talking about. Some of this might be a bit garbled in the retelling, but I don't really care to go back to it, and want to focus on the conclusion.

My understanding is that the drama began when Dom Luczyczyn called out Andy and Rono, a pair of Czech brothers who also make analytics cards, for not crediting the source for their microstat data. I get the impression that Andy and Rono are not taken particularly seriously in that community, and the commentary they offer in tandem with their numbers often comes off as betraying a shallow understanding of the topic. This may be due to their limited English making them appear less articulate, and that's not their fault, but you still get the kind of "so and so is showing he's a true number one defenseman" kind of broad strokes stuff that doesn't strike me as all that informative.

Anyway, they responded by claiming that their source did not in fact want to be named, and alluded to them being an upstart in the field who perhaps weren't ready for prime time. People got into talking about the price for the data, and it turns out this data is something like seven times cheaper than what the Athletic is paying their supplier for Luczyczyn to work with. This made certain people suspicious, as something as high-effort as data collection for the entire NHL seems like something you wouldn't just do secretly and on the cheap. Rachel Doerrie, a former NHL employee, alluded to the existence of cheaper sources that "weren't as good" and sort of left it at that.

Jack Fraser, for his part, sources the microstat data from Corey Sznajder, who is just one guy who started tracking everything on his own volition - I personally think that's really cool and the world really needs these guys. Sznajder apparently used to volunteer his data to Andy and Rono, but does not anymore for reasons I don't think ever came up. Speculation arose that the mystery source of data was possibly plagiarizing Sznajder, which I don't think anyone directly blamed on Andy and Rono. Things died down from there, with the usual punditry about whether Dom was 1) a defender of ethics in hockey analytics or 2) an abrasive jerk or 3) yes.

So, my takeaways on this, are:
1) the phrase "the analytics say" is near nonsense, as not only are there many sources for information of this kind, but that it goes through multiple steps, including data collection, number crunching, and graphic presentation.
2) some of the stuff you see is better than others.
3) not everything is JFresh, and he's not the Kleenex of analytics. That's one guy, and his output includes the work of peers including Sznajder and Bacon, who's method is not as black-boxed as I implied upthread, and also sounds delicious.
4) this is exhausting, no wonder so many of you guys hate this.

 

[Wee Baby Seamus]

According to my statistical analysis, Connor McDavid has the worst Flongle Score Card™ in the league and should be going unclaimed on waivers. Tyler Myers on the other hand has one of the highest Flongle™ Scores of all time and should be running away with the Norris vote.

Obviously, I won't tell you what Flongle Score™ measures or how it's calculated because that's proprietary, but you should definitely keep paying me $10/month on Patreon.

the choice of players is funny, because commentators who were taking advanced stats seriously were saying that Tyler Myers blew at a time when people were using the eye test to basically say "he's big so he must be a 1D."

 

[jigglysquishy]

I do find the cards helpful for players I don't get a chance to watch a lot. I prioritize Jets games, and then teams with big stars (Oilers) and whatever is on HNIC. I might only get to watch a team like the Panthers once a year before the playoffs. So ya, the cards help close to knowledge gap.

It's nice to compare cards for players I watch a lot. I see a lot of Morrissey and Scheifele and Hellebucyk. I get to see a lot of McDavid and MacKinnon. I see very little of Reinhart or Panarin. So if I know the limitations based on players I know then I'll know limitations on players I don't.

 

[Stephen]

desjardins was fantastic. people will disagree but through the second half of the '90s he was the clear number four dman of his generation, behind pronger, lidstrom, and blake, but before niedermayer became niedermayer and chara was anything. (this assumes that leetch is of the previous generation, which we can argue about methodologically, but let’s not.)

i liken it to pietrangelo when he was making second all-star teams. behind the norris guys, karlsson, weber, doughty, keith, but consistently in the mix at the top of the next tier.

to me, peak desjardins impacted the game more than peak letang, although letang was more of a gamebreaker, as HO nicely puts it. i would take pietrangelo over letang, so i guess i take rico over him too.

The fourth best defenseman of his generation with fuzzy definition of what a generation constitutes is too hot a take to leave alone, especially the Leetch thing... One year older than Desjardins and won a Norris in the middle of the late 90s, retired the same year as Desjardins, or Al MacInnis, who was 5 years older but snuck in a Norris before Y2K retiring right before the lockout...

I want to say he was a Ryan Suter+, with a higher profile and counted on to be a Mr. Everything on two higher profile teams in Montreal and Philadelphia.