Rumor: Leafs significantly expanding their analytics department

[IBeL34f]

Never said there was with the Leafs, I leave that to posters on here.

Alright. Seems like a weird thing to bring up if you don't think there's any concern of it occurring within the Leafs organization, but fair enough.

 

[Menzinger]

I'm not the one who said this:

"and a good chunk of that $50m would go towards hiring and training people to collect data properly....and not only on future games, but to go back and watch game tapes for the last 10yrs, re-collect the data more thoroughly and accurately,"

That is saying that the current data has not been collected properly, accurately or thorough enough. If you ran a company, is that the data you would base decisions on?

Again, you can believe that this data has value but can be improved.

Nobody is ever making a decision that is 100% based on stats. But for example, if your existing formation has say a 80% accuracy rate, and via innovation you can get to a 90% accuracy rate, that’s worth it. And doesn’t necessarily invalidate what was used in the past, it’s just more effective than it was before.

 

[Notsince67]

I think some forget the data is there to assist the human evaluation, not the other way around.

Just like social media was suppose to assist human interaction instead of replace it

 

[RLF]

Alright. Seems like a weird thing to bring up if you don't think there's any concern of it occurring within the Leafs organization, but fair enough.

It was in a reply from a conversation with another poster. If you read through the conversation, you would see the comment is in line with what was being talked about.

 

[RLF]

Again, you can believe that this data has value but can be improved.

Nobody is ever making a decision that is 100% based on stats. But for example, if your existing formation has say a 80% accuracy rate, and via innovation you can get to a 90% accuracy rate, that’s worth it. And doesn’t necessarily invalidate what was used in the past, it’s just more effective than it was before.

People make evaluations based on stats on here all the time. It is done daily. Again, I didn't make the statement, nor have I said that stats do not have value. It is the extent of the value that I question at times...and not from the Leafs. Nylander has been a great example of what value some put on stats.

 

[IBeL34f]

It was in a reply from a conversation with another poster. If you read through the conversation, you would see the comment is in line with what was being talked about.

I did read the conversation - The poster you replied to suggested that it was unwise for Dubas to spend $50M on an "idea/theory", and you said that any advantage was good, but, if we rely only on data when making player personnel decisions, that that would be a mistake, in your opinion.

I was just curious if there was anything about the project that suggested that that would ever be a concern, since it seems like one of the last things on earth that this management group would do.

 

[Sweet Leaf]

If the previous analytics department recommended a 7M AAV deal over 6 for Nylander I don’t think 50 million $ is enough to fix it.

 

[grapes55]

I think it's delusional to think this could be anything other than a really good thing

 

[Notsince67]

I did read the conversation - The poster you replied to suggested that it was unwise for Dubas to spend $50M on an "idea/theory", and you said that any advantage was good, but, if we rely only on data when making player personnel decisions, that that would be a mistake, in your opinion.

I was just curious if there was anything about the project that suggested that that would ever be a concern, since it seems like one of the last things on earth that this management group would do.

I once ran a quant analytic department that crunched terabytes of information. It would have required zero extra people if we moved to petabytes. It is amazing what someone can do with 4 or 5 people. 50MM is a bit puzzling. Sounds like significant capex expenditure with operating expenditure aggregated for 10 years. I'm really not sure what kind of white elephant they are building if that is an annual operating spend.

 

[IBeL34f]

I once ran a quant analytic department that crunched terabytes of information. It would have required zero extra people if we moved to petabytes. It is amazing what someone can do with 4 or 5 people. 50MM is a bit puzzling. Sounds like significant capex expenditure with operating expenditure aggregated for 10 years. I'm really not sure what kind of white elephant they are building if that is an annual operating spend.

Doesn't sound like anyone knows exactly what's going on yet. I'm intrigued to find out more.

 

[RLF]

I did read the conversation - The poster you replied to suggested that it was unwise for Dubas to spend $50M on an "idea/theory", and you said that any advantage was good, but, if we rely only on data when making player personnel decisions, that that would be a mistake, in your opinion.

I was just curious if there was anything about the project that suggested that that would ever be a concern, since it seems like one of the last things on earth that this management group would do.

Not sure why you think it is odd then. The point was, any extra info can be good. It has nothing to do with if it seems like they plan to use it in a certain way, just if it was the plan in the future to rely only on it(as no one knows how they are going to use it for sure), I don't think that would be a good plan. Seems reasonable to mention in the scope of the conversation. Didn't think it would cause issues on here...

 

[Mickey Marner]

The $50,000,000 figure is just throwing shit at the wall, pay no mind to it. You couldn't spend that much on hockey analytics if you wanted to. The industry isn't large enough to command such a dollar value.

 

[WillNy29]

did I miss something, I wasn't aware of the Oakland A's winning the World Series under Billy Bean?

until you realize the red sox took the same model threw more money at it and won their first championship in decades...

 

[rumman]

until you realize the red sox took the same model threw more money at it and won their first championship in decades...

yes, we've all seen the movie, the clip that was posted didn't relate to the Leafs, they're hardly hurting for money.

 

[zeke]

I once ran a quant analytic department that crunched terabytes of information. It would have required zero extra people if we moved to petabytes. It is amazing what someone can do with 4 or 5 people. 50MM is a bit puzzling. Sounds like significant capex expenditure with operating expenditure aggregated for 10 years. I'm really not sure what kind of white elephant they are building if that is an annual operating spend.

1. collecting the data (both from future games and re-doing past games via tape, both from the nhl and from every minor league in the world).
2. creating brand new and improved metrics

 

[Notsince67]

1. collecting the data (both from future games and re-doing past games via tape, both from the nhl and from every minor league in the world).
2. creating brand new and improved metrics

Right now do the math. Assuming salary and benefits of 130k, 40 people is 5.2Mm of your 50mm. How many people you figuring they plan to hire?

 

[zeke]

Right now do the math. Assuming salary and benefits of 130k, 40 people is 5.2Mm of your 50mm. How many people you figuring they plan to hire?

eh the $50m was never a figure that made sense for 1yr. I assume that's a longterm number.

 

[Notsince67]

eh the $50m was never a figure that made sense for 1yr. I assume that's a longterm number.

That was my point.

 

[Cap'n Flavour]

I once ran a quant analytic department that crunched terabytes of information. It would have required zero extra people if we moved to petabytes. It is amazing what someone can do with 4 or 5 people. 50MM is a bit puzzling. Sounds like significant capex expenditure with operating expenditure aggregated for 10 years. I'm really not sure what kind of white elephant they are building if that is an annual operating spend.

If you ran a quant department you should probably understand that the complexity of data matters more than the sheer volume. If the NHL is really going to put accelerometers and do position tracking with any kind of accuracy on players and the puck, that's going to be a much more complicated data set than simple time series. It will require a lot of time and effort just to clean the data and begin to make sense of it, let alone develop useful metrics to inform high-level decisions.

 

[thewave]

Again. This means nothing if you don't have good negotiators doing contracts. Sure they can find good players a little better, then when RFA comes they overpay run out of cap and have to trade people? Sounds like a recipe for high turnover.

 

[Notsince67]

If you ran a quant department you should probably understand that the complexity of data matters more than the sheer volume. If the NHL is really going to put accelerometers and do position tracking with any kind of accuracy on players and the puck, that's going to be a much more complicated data set than simple time series. It will require a lot of time and effort just to clean the data and begin to make sense of it, let alone develop useful metrics to inform high-level decisions.

I ran the risk analytics area of a big bank. We looked down to the instrument (credit card) level to build our models. pretty sure the performance characteristics of a number of instruments for a few million customers are bigger than what hockey has. You don't need a lot of quant heads to do this stuff.

 

[Cap'n Flavour]

I ran the risk analytics area of a big bank. We looked down to the instrument (credit card) level to build our models. pretty sure the performance characteristics of a number of instruments for a few million customers are bigger than what hockey has. You don't need a lot of quant heads to do this stuff.

I don't think anyone here is talking about hiring a dozen people to do linear regression on tables from Corsica but to actually do play-by-play analysis, presumably using whatever tracking data the NHL produces next season. It will take a lot of effort just to process the input data. Even if you had accurate positions and accelerations, it's a lot of effort to turn that data (at what time resolution? who knows) into an interpretable product, depending on what you imagine doing with it. Presumably they'll need people analyzing the video after each game because I can't imagine the NHL is planning to put accelerometers into every player's stick and even if they did, that's not enough to reconstruct every play.

Soccer has been tracking pass completion stats for what, 10-20 years now? They also show running distance during a game, etc. The most the NHL shows publicly is the odd shot/skating speed, but a rich team like the Leafs can easily afford to do a lot more with some imagination. Hell, for all we know they could've been tracking their own players during home games and there's an awful lot you could learn from that. It's not like the mechanics of skating are well understood, for example. The Bruins have apparently been compiling biometrics in practice for at least a couple of years now, according to this article.

Anyway, my expectation is that the NHL will not release anything terribly interesting to the public except perhaps for gambling purposes and all we'll see is the odd 'player x skated 5 miles this game and reached a top speed of 45 km/h!' during broadcasts.

 

[Notsince67]

I don't think anyone here is talking about hiring a dozen people to do linear regression on tables from Corsica but to actually do play-by-play analysis, presumably using whatever tracking data the NHL produces next season. It will take a lot of effort just to process the input data. Even if you had accurate positions and accelerations, it's a lot of effort to turn that data (at what time resolution? who knows) into an interpretable product, depending on what you imagine doing with it. Presumably they'll need people analyzing the video after each game because I can't imagine the NHL is planning to put accelerometers into every player's stick and even if they did, that's not enough to reconstruct every play.

Soccer has been tracking pass completion stats for what, 10-20 years now? They also show running distance during a game, etc. The most the NHL shows publicly is the odd shot/skating speed, but a rich team like the Leafs can easily afford to do a lot more with some imagination. Hell, for all we know they could've been tracking their own players during home games and there's an awful lot you could learn from that. It's not like the mechanics of skating are well understood, for example. The Bruins have apparently been compiling biometrics in practice for at least a couple of years now, according to this article.

Anyway, my expectation is that the NHL will not release anything terribly interesting to the public except perhaps for gambling purposes and all we'll see is the odd 'player x skated 5 miles this game and reached a top speed of 45 km/h!' during broadcasts.

Sure but people tracking information would be at reduced salaries from my assumption. My point was from an operating expenditure perspective, 50MM looks way too high for a single year. That would employ hundreds of people. I'm also pretty knowledgeable about the technology infrastructure costs required and it doesn't make sense. That number is a multi-year spend.

 

[666]

I think the Leafs current analytics must be broken that is why they're investing $50 mil starting next year.

The Leafs playing at the cap ceiling needed to add a #1C for $11 mil in Tavares and dump $850k McElhinney to promote $750k Sparks to get 46 wins and 99 points.

Carolina playing at the Cap floor grab McBackup and now have 45 wins and 97 points, while the NYI lose their star franchise player also have a bottom 5 Cap and have 47 wins and 101 points.

One team loses a star #1 C and their results go UP and one team adds a star C and their team results go Down. One team tosses a backup goalie on waivers and their team results go Down and another team grabs that goalie off the scrap heap and their team results go UP.

I'd love to know what kind of advanced analytics can be produced that don't already exist today to help explain this situation above?

Coaching.

 

[Cor]

The $50,000,000 figure is just throwing **** at the wall, pay no mind to it. You couldn't spend that much on hockey analytics if you wanted to. The industry isn't large enough to command such a dollar value.

You can fairly easily between having a large enough staff with salaries, getting software created, paying different teams for their video to use, so on and so forth.