Keith Pelley said the new head of hockey operations for the Maple Leafs needs to be datacentric. Much confusion, some deliberate, has resulted from this term. The presser where Pelley spoke was not about specifics, and he wasn't reading out the skills and experience required to apply from a job ad, but what he was talking about was not skill-specific.

Datacentric doesn't mean having a degree in statistics or a related field. It doesn't mean a direct and personal involvement with what is called the analytics movement. It doesn't mean you're a hacker, a theoretician, or the person who directly collects or analyzes data. It's not about understanding the chart the data science team makes without the explanation that comes with it. It's about how you think.

Evidence-based decision making was really the key term used by Pelley, which is part of the answer to what datacentric means, one of the most important parts, but it's not the whole story.

You can go a long way towards effective evidence-based decisions just by memorizing the standard list of errors in reasoning, and committing yourself to not doing them. Throw in an understanding of some of the basic ideas like regression, how variance makes small amounts of data unreliable, how predictivity falls away when the target group gets smaller (this is the same thing said twice) and you're better than a guy who just shouts out confidently what is just obvious common sense. One thing I think is critical, you need to listen to Craig Simpson more. I'll explain that in a bit.

This is how I can play the piano. I spent a lot of time learning the language of music, memorizing songs in this instructional book I had, and practicing playing them. However, like Data, I have no innate musical skill or ear for the sounds. And all I could ever do was just perform these learned steps. I was never a musician. I can never be that. But I could play Let It Be and you'd recognize it.

I'm not sure, because humans are diverse and unknowable in many ways, but maybe some or many or a lot of people can't ever think the way that makes memorizing all those fallacies beside the point. A genuine and sincere belief in the supernatural is a big barrier. It feels like a choice to me, but I'm not the oracle of all knowing here. My feelings on this could be wrong.

Let's talk baseball for a second. I used to watch the Blue Jays back when they won the World Series, and they were fun. Baseball itself was pleasant, but I do not know how anyone who does not knit becomes a baseball fan – that's human diversity again. I cared not at all about stats. Not even the ones your mother Buck Martinez gives you that don't do anything at all. I didn't need that to enjoy the games. Eventually, I stopped caring about the Jays for a long time.

I watched them in the playoffs last fall because they were fun. I enjoyed myself and also fell asleep halfway through several games. I ignored all discussion of anything beyond the score of the game, and who was doing what cool thing. I picked out who I liked mostly based on unexamined gut reactions and I cheered for my favourites.

You can watch hockey this way, and if you like doing that, if it's what makes you happy, keep it up. But you should understand something: I am so very not debating with you the ideas of evidence based decision making and datacentric approaches as the one and only correct way to run a hockey team. This is the only way you can win without just getting lucky.

If you want to passionately talk at me about the intangible wonder of Vladdy or Springer and how deeply important that is, how much their emotions and affect matters, how inspiring they are, and how Springer's hockey dude ability to play on a crapped out knee was what gave the team the edge – can you hear the soaring orchestra? I mean, you bet, this is evocative stuff. But if Springer ain't hitting no dingers, he ain't never going to be there to be an example of fortitude and commitment. And finding the guy who is going to keep hitting dingers is what evidence based decisions lead you to.

You're allowed to have emotions while you're building a hockey team from data and evidence. You're allowed to use your judgement and consider all those human intangibles. This is really important, actually, and because all that stuff is important, that's doubly the reason why the boss needs to think in a certain way, not just have memorized what the sunk cost fallacy is.

The basis of this way of thinking really is a full and complete commitment to the idea that the universe is not interested in your hockey game. It is not a sentient being. You must understand, and stop kicking against forevermore the fact that hockey is the most random professional team sport. Reductionist thinkers and people who like to argue recreationally will come up with retorts about how the outcome of a game is not "just luck" which, no shit.

This is a key reason why the boss of the whole show has to be the right kind of thinker. They can't be the person making that stupid argumentative taunt all the damn time. They have to know without being told what it means to be making decisions that lead to winning in a sport where the margin for error is so very, very tight. This is what the stakes of the whole game are, and if you don't even get that, you should get out of the way.

If you start with the correct basic mode of thinking about hockey, you're ready to take that meeting with the data science team. What that team is doing is nothing even remotely similar to what fans do with a list of on-ice Corsi numbers, an isolated impact chart or a ranking on a team list of Expected Goals %. That's all about looking for some evidence to decide how to feel. Or more likely to decide how stupid the team is for not doing what the purple bar says they should do.

What the real thing is all about is trying to describe a hockey game, its players and its play in simple enough terms that you can talk about it to each other and get smarter by turning your brains to the task of optimizing the efforts at winning.

We do this simplification and have done this for thousands of years, and some people believe there was some kind of cognitive spark we lit in our minds deep in the past that allowed us to do a thing we think other animals don't do: think in the abstract.

An abstraction is a lie. Or maybe more properly a false thing that's close enough to true to be useful. We use them all the time. There are so many of them that have been created by humans over decades that were used to put these words on a screen for you to read that it boggles the mind to think of them all. Abstractions are how we create things that we use to make a computer do a thing.

What gets called analytics or advanced stats (two stupid terms) is a pile of abstractions of a hockey game. It's a bunch of bits, each one a picture of something that we fit together like Lego until the result is what we want – a better chance to win. Over and over until we know what's more likely to work and what isn't.

Every team in the NHL now has years of puck and player tracking data. Masses of it. And they've had time to begin to analyze it and try to find abstractions made from that data that can tell you how to play, who to play, when to play, and where to spend and where not to.

It's not even the hockey ops boss's job to decide which method of drawing abstractions they use. What's important is that the boss is in a position to approach them with a full understanding of how you decide if they are useful as evidence, and how you know what to do with that evidence once you have it.

Pelley said that evidence-based decisions are always right, and this, alas, is simply not true. They are the right ones to make, but they can still give you the wrong answer, and understanding the weaknesses of what hockey data can tell you, or at the very least, knowing that there will be weaknesses but that your gut feelings won't suss them out is key.

These abstractions of a player or of game concepts are like a snapshot. They exist outside of time, they have no movement, no context, and that's good in many ways. The emotional response to the player is removed, so the understanding is not coloured by that. That is a dangerous thing, though, because it dehumanizes the player. The context gets supplied by the viewer, and a very good datacentric approach is self aware of these limitations in the information and in their own response to them.

Craig Simpson is here, let's listen to him for a second... "Expected Goals! Ha ha, expected by who?" The boss has to know the answer to that one. This is more complex that just knowing what correlation is not. The answer to that one comes first from an understanding that it's not entirely a stupid question. Because expecting Expected Goals in a given game or set of games or in a given instance is not what those model outputs are telling you to do. They aren't telling you who was good or bad in the game last night either.

What is description, what's the difference between a thing the player did and who the player is, what is repeatable, predictable, all of these concepts need to be there in the boss who is given the snapshot to look at by the data science team.

You and I, here right now, we don't know what's been going on in the Leafs basement (done up just like their mom's house, so them nerds feel at home – go away, Craig, we're done with you) but the vehemence with which Pelley emphasized this aspect of a new hire, and his talk about the other MLSE teams being ahead of the Leafs, make me think they've let that side of things backslide a little. Or more properly, the rest of the league has caught up to their early innovation.

The current man in charge, Daryl Metcalf, has a stellar reputation, but I can't believe that Leafs roster, that coaching hire, that system, those trades were made by someone who consistently took in what Metcalf's crew had to say. You have to think a certain way about those human intangibles and the context you're bringing to your decision making to not end up doing something stupid by ignoring the evidence in favour of what you're just so sure is the right way to play the game.

Part of the way of thinking that's required is knowing your own fallibility. It's also knowing in your gut, not just your head, that you need to be most skeptical of the things you want to be true. Particularly those cultural things not actually related to winning.

Look, let's be clear here. I haven't got one single clue how effective any team has been at using all that NHL data. For all I know half the analytics departments are doing "that cloud looks like a bunny" stuff, making spurious correlations graphs, or worse, retroactively constructing evidence for the choice the boss wants to make. Managing up happens everywhere.

There are some failsafes to the use of data-based evidence that have to be inside the brain of the boss. No memo with six charts from the basement crew is going to be enough if the mind that's receiving it has an Outlook folder labelled "Critical" where he puts all the stuff he's never going to read. At the same time, the boss has to be aware of the fallibility of the information he's getting, and the risks of imperfect abstractions, because make no mistake, they are all imperfect.

One of the ways of thinking that's very particular to hockey is the idea of the process vs the results. It's like going back in time to talk about this stuff because Sheldon Keefe, not Kyle Dubas, was the guy who really, really got this. He got this so well he was open about his own hesitation over how far you let something play out.

An example. Let's do a five-forward power play.

First try, the puck gets past the forward covering the blue line and the other team gets a scoring chance.

Response: oh, that'll never work.

Congrats, you are not datacentric and cannot have this job.

No, but you see, it's obvious, the forward can't defend, you can't do this dumb idea because it will never work, stop now.

Congrats, you also don't make evidence-based decisions.

Let's try it for a few games, and whoops, someone scored a shorty.

Result: that's the last time we'll do of that. You can't be giving up goals on the PP.

Congrats, not only should you never run or coach a hockey team, you actually created artificial reasons for your decision that you know are false.

Now, hold onto your hats because I'm going to talk about sexism. In my experience a well-socialized man, even today in the 21st century, has been at least partially convinced that the way you have to go about things is exuding confidence and just being sure of yourself. Making up some bullshit you would know is bullshit if someone else said it and saying it with confidence so real maybe you really believe it is often considered the sign of a good leader. And leadership is important and intangibles and blah, blah blah.

The right answer from the boss of the team running the five-forward PP is this: we don't know if this is working or not, why the hell are we test piloting this in the damn NHL? What the hell is wrong with you? Get the fucking AHL team to do it for a whole fucking season and then maybe we'll know. Better yet, bribe some fucking league to have half the teams do it and half use the normal one.

See, you can be decisive and sure of yourself while making only evidence based decisions! It takes practice, and it takes the way of thinking to be innate, so that when you bust out in frustration your reasoning process stays cold. It can't be a coat you put on or off. It has to be who you are.

Where this becomes crucial is when the thing clouding the mind of the boss is hate or love. If the boss hates a player or really loves them, and they don't have the ability to see that weakness, than it can override their commitment to the hard cold truth, and they will make crucial mistakes.

I'm not sure which is worse, love or hate, for leading you astray from the path of logic and rationality. Actually maybe it's fear. This sounds like this way of thinking is emotionless, but it isn't. It's balanced and tempered, maybe even constrained. And that's hardly a new concept for the NHL, where "it's a business" is the all-purpose explanation for hard cold decisions that players don't like. You can actually make those tough calls about things other than money.

The NHL, and sports in general, are an odd place to find someone who has the way of thinking to be a really great head of hockey operations and also be all the other things the job demands. Sporty Spice but mathy usually ends up working for the business side of a team or a gambling company because the operations department is where "I just think" analysis has reigned supreme for a century.

Anyhow, that's what I think datacentric means and why it's not a nice to have, it's a must have if you want to win and keep winning in an unfair world.

Analysis