What would be the "dead sea scrolls" of hockey data?

[thom]

The stories Dick Irvin Jr must know has to be incredible the players that use to walk in his dad's house he must know the demons of all of the dead players and many who are still alive.Dick Don Cherry and Howie meeker are 3 treasures who won't be be here much longer they are canadian and hockey icons.

 

[BraveCanadian]

The old estimates are far from perfect, but they ballpark it enough to where we can safely say Gretzky wasn't playing 30 minutes per game.

In the early 80s he played a ton.

What does the sheet have him at during say 80-84?

There are tons of references to him playing an average of 27-30 minutes a game back then.. double shifting with multiple sets of wingers etc.

 

[seventieslord]

In the early 80s he played a ton.

What does the sheet have him at during say 80-84?

There are tons of references to him playing an average of 27-30 minutes a game back then.. double shifting with multiple sets of wingers etc.

I don't have the sheet handy, but it’s no more than 25.

This argument coming back up is really disappointing. It fails just like it did with Orr. Simply, if you want to believe Gretzky was playing 27-30 minutes, the goals for and against don’t change, making the frequency of goals when he was on the ice lower. And we can all agree that goals happened much more frequently when he was on the ice (mostly for, but also against)

 

[BillyShoe1721]

In the early 80s he played a ton.

What does the sheet have him at during say 80-84?

There are tons of references to him playing an average of 27-30 minutes a game back then.. double shifting with multiple sets of wingers etc.

I don't have the sheet handy, but it’s no more than 25.

This argument coming back up is really disappointing. It fails just like it did with Orr. Simply, if you want to believe Gretzky was playing 27-30 minutes, the goals for and against don’t change, making the frequency of goals when he was on the ice lower. And we can all agree that goals happened much more frequently when he was on the ice (mostly for, but also against)

They've got him consistently around 25-26 minutes during most of the '80s.

 

[BraveCanadian]

I don't have the sheet handy, but it’s no more than 25.

This argument coming back up is really disappointing. It fails just like it did with Orr. Simply, if you want to believe Gretzky was playing 27-30 minutes, the goals for and against don’t change, making the frequency of goals when he was on the ice lower. And we can all agree that goals happened much more frequently when he was on the ice (mostly for, but also against)

Even if it is the case that the first hand accounts are incorrect, estimates based off calculations that no one has aren't going to convince me very well.

They've got him consistently around 25-26 minutes during most of the '80s.

It would be interesting to see how the numbers were calculated so that we could see how it accounts for Gretzky's rather unique penalty killing numbers in the 80s.

 

[TheDevilMadeMe]

Even if it is the case that the first hand accounts are incorrect, estimates based off calculations that no one has aren't going to convince me very well.




It would be interesting to see how the numbers were calculated so that we could see how it accounts for Gretzky's rather unique penalty killing numbers in the 80s.

Why can't both be true? I'm sure he often played 30 minutes per game when the Oilers were behind. And he probably also often played 20 minutes per game during Oilers blowouts - there are definitely tons of first hand reports that talk about the Oilers cutting down ice time for their stars during blowouts. And what does it average to?

I do agree with you about one thing though - it's pretty deplorable that nobody has the exact calculation used in the estimates.

 

[seventieslord]

Why can't both be true? I'm sure he often played 30 minutes per game when the Oilers were behind. And he probably also often played 20 minutes per game during Oilers blowouts - there are definitely tons of first hand reports that talk about the Oilers cutting down ice time for their stars during blowouts. And what does it average to?

I do agree with you about one thing though - it's pretty deplorable that nobody has the exact calculation used in the estimates.

I've seen these calculations at some point. They're not complex. The situational numbers are based on (very) reasonable estimates of time the team played in each situation – based on PP opportunities for and against. From there it looks at what percentage of time each player was likely to have played in each situation based on what percentage of the GF/GA they had.

I distinctly remember someone explaining the “fudge factor†or whatever you want to call it, was to multiply first liners/pairings by 1.2 and second liners/pairings by 1.1 to reflect real life observations and the higher frequency of goals that tend to happen when they’re on the ice.

I’m sure this was posted at some point. An “exact formulaâ€, maybe not, but we all know enough algebra to build that formula ourselves.

 

[ot92s]

Even if it is the case that the first hand accounts are incorrect, estimates based off calculations that no one has aren't going to convince me very well.




It would be interesting to see how the numbers were calculated so that we could see how it accounts for Gretzky's rather unique penalty killing numbers in the 80s.

this is the correct stance bravecanadian. Based on first hand accounts and contemporary comments i believe he was probably around 30min/game in the early 80's.

I still haven't seen this ad hoc formula which supposedly put the matter to rest.

 

[BraveCanadian]

Why can't both be true? I'm sure he often played 30 minutes per game when the Oilers were behind. And he probably also often played 20 minutes per game during Oilers blowouts - there are definitely tons of first hand reports that talk about the Oilers cutting down ice time for their stars during blowouts. And what does it average to?

I'm sure both are true to an extent. It will also vary a lot depending on which year we look at.

I know I have seen plenty of references to his ice time in the early 80s being 27+ minutes a game and seen Sather's quoted philosophy of how to handle Gretzky is to play him as much as possible.

Once they were a dynasty level club I am sure they were coasting a lot more.

My main issue is the certainty with which people quote an estimate that is based on a formula with a known subjective factor in it which is in turn based on a calibration to the late 90s/early 2000s player usage.

I am sure we can all agree that shift lengths and ice time distribution changed quite a bit from the mid/late 80s on.

Other factors like the introduction of TV timeouts would also have a small impact for example.

The fudge factor alone means that you are picking which players you believe were the "first liners" or "second liners" and then applying the adjustment to make it fit what you expect.

By definition that implies that all teams use the same setup and all players in that role are used the same and that simply isn't true.

In the early 80s Gretzky played a ton.

Are the estimates in the ballpark? Sure I think we can reasonably say they are.. I just don't think they are the be all and end all. And I can easily see them being off by a couple of minutes.

I do agree with you about one thing though - it's pretty deplorable that nobody has the exact calculation used in the estimates.

Right. A lot of people base a good chunk of their player evaluations and other work on ice time. It is incredible that these numbers are being taken at face value.

If it is so easy then someone should be able to recreate the results fairly quickly.

My feeling is that no one wants to do that because then the flaws will be more visible.

 

[TheDevilMadeMe]

Are the estimates in the ballpark? Sure I think we can reasonably say they are.. I just don't think they are the be all and end all. And I can easily see them being off by a couple of minutes.

It's quite possible they are off by a minute or two. Even if they are, a couple minutes error still doesn't bring Gretzky up from the estimated 25-26 minutes to 30 minutes

 

[BraveCanadian]

It's quite possible they are off by a minute or two. Even if they are, a couple minutes error still doesn't bring Gretzky up from the estimated 25-26 minutes to 30 minutes

It does bring it to 27-30, though, which is what many people were saying at the time.

 

[Bear of Bad News]

It does bring it to 27-30, though, which is what many people were saying at the time.

27-28, actually.

 

[BraveCanadian]

27-28, actually.

Sigh.

I meant it would put the number into the range people were citing in media at the time:

ie. 27-28 is included in the range 27-30 for the pedants among us.

 

[Bear of Bad News]

Sigh.

I meant it would put the number into the range people were citing in media at the time:

ie. 27-28 is included in the range 27-30 for the pedants among us.

Ah, I see - thank you for clarifying.

I was interpreting it as justification to bring it closer to 30.

 

[Killion]

... for the pedants among us.

... are you suggesting some of us are precisionists, perfectionists, anal retentive Sir? Why, I oughtta

 

[overpass]

Rough ice time estimates are pretty straightforward. No need to reverse-engineer an existing method. And in fact we may not want to use a method to estimate Gretzky that was derived for use with the general population of players.

Take Gretzky in 1981-82.

Special teams ice time is really easy. Gretzky was on the ice for 75 of Edmonton's 86 PPG or 87%. The PP is probably less efficient without him so let's say he get's 80-85% of PP ice time. Gretzky was on the ice for 14 of 67 PPGA, or 21%. But he also scored 6 G and 2 A while shorthanded, so he may be killing overmatched second PP units. Let's say he's on anywhere from 20-30% of the time on the PK. I don't feel like estimating team special teams ice time so let's assume PP and PK are equal on the team level. Gretzky was on the ice for about 50-60% of special teams ice time.

At even strength Edmonton scored 310 and allowed 220.
Gretzky was on the ice for 190 ES+SH GF and 109 ES+SH GA. Remove a few goals to eliminate SH, leaving 182 ESGF and 102 ESGA for Gretzky.

Let's use a few different assumptions.

1. Goals are scored at the same rate when Gretzky is on the ice as when he is off the ice. Edmonton just scores more of them with him on the ice. So because 54% of ES goals are scored with Gretzky on the ice, he played 54% of ES ice time.

2. Gretzky's defensive effect is zero and his offensive effect is hugely positive for Edmonton. In this case his GA alone provide the best estimate. So because Gretzky was on the ice for 46% of ESGA he was on the ice for 46% of ES ice time.

3. Gretzky was an incredible defensive player but average offensively. GF alone provide the best estimate. Gretzky was on the ice for 59% of GF and 59% of ES ice time. (Yes this is a stupid assumption but Sports Illustrated wrote that Sather said Gretzky was playing 38 minutes a game in 1981-82. For that to be true the assumptions given here would have to be true.)

4. Gretzky cheated for his offence, costing his team GA. The team allowed GA at a 10% higher rate when he was on the ice. So he was on for 46% of GA but only 43% of ice time.

Which assumption is true? Well, Edmonton actually allowed fewer goals than the league average in 1981-82 (220 ESGA compared to 239 league average) so I doubt Gretzky was worse defensively than the average first line centre. But if teams tend to allow more GA with their top line on the ice, assumption 4 could still be close. I'm going to throw out assumption 3 as stupid.

Anyway, let's say that reasonable estimates for Gretzky's ES ice time range from 43% to 54%, depending on your assumptions. (The higher-end assumptions suggest Gretzky as a puck-possession-dominant player who lowers GA as much as boosting GF by denying the other team the puck, rather than a gambling attacker who liked to trade chances.) And his special teams ice time is 50+%. So it's not inconceivable that Gretzky played 30 minutes a game in 1981-82. The lower-end estimates would put him around 26 minutes a game.

One factor that may or may not be considered - bottom sixers tend to play things safe in the ice time era and the estimates are probably based on that assumption, with lower GF and GA rates for lesser players. Did this assumption hold in the 1970s and 1980s?

 

[Ohashi_Jouzu*]

this is the correct stance bravecanadian. Based on first hand accounts and contemporary comments i believe he was probably around 30min/game in the early 80's.

I still haven't seen this ad hoc formula which supposedly put the matter to rest.

According to first hand and contemporary accounts, he was over 30 minutes per night. Like, well over. Wouldn't be hard to test, for someone with stock video lying around. Take one game, any game (they were pretty much all the same, but limit yourself to games where the Oilers blow out teams early, when Gretzky might be expected to sit more toward the end of the game - set the low bar), and stopwatch for yourself how long you see Gretzky on the ice. I expect it would be difficult to turn up consecutive games (not marred by injury, or whatever) in the earliest 80s where Gretzky got less than 30 mins on the ice.

I was watching a rebroadcast 80s Oilers game on the NHL network once, when dad came downstairs and ended up watching the rest with me. Some time during the 3rd period he turns to me and asks, kind of tongue in cheek: "They never really took him off the ice, did they?" Seemed like it was just up to Wayne whether he was good to go or not. Good to go -> get on. Not good to go -> get back on next whistle. lol.