Brassard

[slipknottin]

What's the flaw?

I'll be sure to pass it along to the author.

It only looks at players that played until they were 37.

It doesnt include all the players who were gone/retired/injured at 35, 36, or 37.

And in any case, he keeps splitting up his sample size, so who knows what kind of sample size he actually is using by the end. He, of course, does not say.

Far from scientifically rigorous. I would label it garbage, in fact.

 

[Thirty One]

It only looks at players that played until they were 37.

It doesnt include all the players who were gone/retired/injured at 35, 36, or 37.

And in any case, he keeps splitting up his sample size, so who knows what kind of sample size he actually is using by the end. He, of course, does not say.

Far from scientifically rigorous. I would label it garbage, in fact.

The players who didn't make it to 37 are included. They are dragging down the GP totals.

The original sample is 78 players. To get the smaller ones, you have to divide it by two or three. Ask a friend to help, if need be.

 

[Fletch]

The players who didn't make it to 37 are included. They are dragging down the GP totals.

The original sample is 78 players. To get the smaller ones, you have to divide it by two or three. Ask a friend to help, if need be.

can you further break it down by 6'4" forwards who have accumulated more than >1000 career PIMs? If the sample size is not statistically significant, drop it down to 6'3" forwards with > 900 career PIMs.

Good work, by the way. Didn't fully read, but I like the analysis. Personally, I can't predict when a player will begin to decline, and I would not be afraid to sign a player after he's turned 35. I believe the team needs to have a balance of contracts and players in various age/experience buckets for it to make sense, and, of course, he needs to fit in positionally and contractually. While I think Thornton > Brassard, and I've actually been a Thornton fan since he came up, I don't think it's the right fit. At the moment, and perhaps I haven't given it enough thought or been convinced enough as of yet.

 

[slipknottin]

The players who didn't make it to 37 are included. They are dragging down the GP totals.

The original sample is 78 players. To get the smaller ones, you have to divide it by two or three. Ask a friend to help, if need be.

And by doing pts/game you ignore the problem of a player not playing isnt actually producing points. (another fatal flaw)


And reducing it from all players to just post 94 players is a sample size of how many. (who knows, you didnt say, actual studies have to show the data, you realize)


You essentially cherry picked your way through the data to prop up an unsupported conclusion.

 

[Infinitum Vitae*]

Oh boy here we go... Ima grab my popcorn

 

[Thirty One]

Where does it say those players are included? (hint, it doesnt say that because it doesnt include them)

And reducing it from all players to just post 94 players is a sample size of how many.

Ill let you do your random division, as if thats what the author did, lol.

You are being really dumb.

Spoilers: I wrote that.

Why would I have to specifically say that those players are included? I said which players are included, why would you make random assumptions about further exclusions? Players who didn't play past 34 are included (i.e. Bobby Clarke's last season was his age 34 one).

The samples are split evenly in groups around the median (as stated). For pre-94 it was a 37/41 split.

 

[Thirty One]

can you further break it down by 6'4" forwards who have accumulated more than >1000 career PIMs? If the sample size is not statistically significant, drop it down to 6'3" forwards with > 900 career PIMs.

It will be tiny. Of the sample only 9 were 6'3" or taller.

 

[slipknottin]

You are being really dumb.

Spoilers: I wrote that.

Why would I have to specifically say that those players are included? I said which players are included, why would you make random assumptions about further exclusions? Players who didn't play past 34 are included (i.e. Bobby Clarke's last season was his age 34 one).

The samples are split evenly in groups around the median (as stated). For pre-94 it was a 37/41 split.

I edited my post.

But if it does include them, then you cant go by points/game played. Because you are EXCLUDING all those players from your data set on goals.

 

[Thirty One]

I edited my post.

But if it does include them, then you cant go by points/game played. Because you are EXCLUDING all those players from your data set on goals.

That's why there's a total games played bar too, to show the attrition rates.

 

[slipknottin]

Now if we look at your data of players pts/gp <0.875

You have a 40% reduction in games played going from age 34 to age 36.

That is a 40% reduction in total points right off the bat, not including actual points/game reductions.

Which if you only count the players still playing, was still a dropoff of 15%

Now I dont know the numbers behind your math, but you are looking at something like a 50% dropoff in points scored by players going from age 34 to 36.

Something you completely missed in your own data

Can do that same thing with any of your data sets, they are all flawed.

For the <1 & >.875 you get a 30% reduction in games, and a 12% drop in points. Probably something like a 40%+ dropoff.

 

[slipknottin]

That's why there's a total games played bar too, to show the attrition rates.

It doesnt show what you think it shows.

It is entirely misleading.

 

[Fletch]

It will be tiny. Of the sample only 9 were 6'3" or taller.

so it would be even smaller if you further narrowed to guys with light hair...just kidding of course. It is a bit of an art and you've done a good job getting it to a science.

 

[Thirty One]

It doesnt show what you think it shows.

It is entirely misleading.

It shows exactly what I think it shows. It shows the points per game of players and their attrition rates.

 

[slipknottin]

It shows exactly what I think it shows. It shows the points per game of players and their attrition rates.

It shows something like a 30-50% reduction in performance going from age 34-36.

But as I figured, you ignored that in your conclusion.

 

[slipknottin]

so it would be even smaller if you further narrowed to guys with light hair...just kidding of course. It is a bit of an art and you've done a good job getting it to a science.

Other than being completely incorrect in his conclusions.

 

[Tawnos]

I think you're just asking the wrong question there, 31. I have no problem with the data analysis itself. More relevant there would be the percentage of 35-37 year olds who increase production, show flat numbers (+5/-5%), 10% drop in production, 20% drop in production, 30% drop, etc, etc, as compared to the year before.

But then again, even that would not necessarily apply to Joe Thornton, since none of these statistics are about Joe Thornton. Using other player's numbers to predict what he will do is somewhat pointless.

 

[Thirty One]

It shows something like a 30-50% reduction in performance going from age 34-36.

But as I figured, you ignored that in your conclusion.

30-50% drop off in Thornton's points equals to something like a 40-50 point scorer. You know, exactly what Brassard is..


And thats not including the ~60% chance Thornton doesnt play at all.

I like how you accuse me of data mining but then mine through my data to pick out the data that's most favorable to your point.

So comparing discounted future point production of Thornton to undiscounted future production of Brassard, you get similar results? Neat.

But then again, even that would not necessarily apply to Joe Thornton, since none of these statistics are about Joe Thornton. Using other player's numbers to predict what he will do is somewhat pointless.

How would this be more relevant?

But then again, even that would not necessarily apply to Joe Thornton, since none of these statistics are about Joe Thornton. Using other player's numbers to predict what he will do is somewhat pointless.

lol ok.

 

[slipknottin]

I like how you accuse me of data mining but then mine through my data to pick out the data that's most favorable to your point.

No, Im saying you are looking at games lost and points independently. When you cant.

From your total data set, players going from age 34 to age 37 play ~30% fewer games, while their points in the games they do play are reduced 10-15%.

You need to combine both of those data points.

Players who arent playing dont produce any points, and you are completely ignoring that.

Essentially you are concluding that players only drop 10-15%, while completely ignoring the fact that they play 30% fewer games.

If you redo the study with point totals for the season, it would tell far more.

Frankly I dont care how many ppg someone has when they are going to miss half the season.

 

[Thirty One]

No, Im saying you are looking at games lost and points independently. When you cant.

From your total data set, players going from age 34 to age 37 play ~30% fewer games, while their points in the games they do play are reduced 10-15%.

Those arent independent.

Players who arent playing dont produce any points, and you are completely ignoring that.

K.

I'm done here.

 

[slipknottin]

K.

I'm done here.

Like I said, there is a fatal flaw. You cant base your conclusion on half of the data and say their ppg doesnt drop much, while ignoring that the games played drops by 30%+

If you combine your data points to get an accurate real world model, you are probably looking at something like a 30-50% dropoff in Thorntons points per season.

Which puts him right back at a 40-50 point player, which Brassard already is

 

[Tawnos]

How would this be more relevant?


lol ok.

I'm assuming the first question, you meant to quote the first paragraph. And it would be more relevant because it would layout a likelihood that the player maintains a certain level of production. Way more valuable than a blunt PPG analysis.

Still, the application is fairly minimal. The data set isn't representative of Thornton, because Thornton is an individual with a ton of things that make him different from any of the other players in the sample, from injury history all the way down through genetics. There's a boatload of uncertainty with a player his age that is entirely dependent on him and him alone. Any bulk analysis of other players just doesn't have much value.

 

[Thirty One]

I'm assuming the first question, you meant to quote the first paragraph.

lol yes.

Still, the application is fairly minimal. The data set isn't representative of Thornton, because Thornton is an individual with a ton of things that make him different from any of the other players in the sample, from injury history all the way down through genetics. There's a boatload of uncertainty with a player his age that is entirely dependent on him and him alone. Any bulk analysis of other players just doesn't have much value.

If that's the case, there shouldn't be any discussion of his future. We shouldn't assume that he will decline at 35, or any age for that matter, because he is totally unique.

 

[Tawnos]

lol yes.


If that's the case, there shouldn't be any discussion of his future. We shouldn't assume that he will decline at 35, or any age for that matter, because he is totally unique.

No one doesn't decline. The later in the career, the more likely it is. The individual part of it is when. It's going to happen. I'd rather not risk it happening while he's here after giving up assets.

I also don't think he's a good fit here, but that's a whole other topic.

 

[Thirty One]

No one doesn't decline. The later in the career, the more likely it is.

What are you basing this on?

 

[Tawnos]

What are you basing this on?

Life.