Wayne Gretzky and Mike Bossy Time on Ice

[MadLuke]

And the time remaining in the game is really important to consider as well - The goal scored when three behind is more important the more time is remaining in the game, and the goal scored when one ahead is more important the less time is remaining (to an extent).

Yes and getting a team ahead by 2 goals is a bigger deal than making a 5 goals deficit down to 4....

Maybe just some scoring when the goal differential is 3 goals or less, 2 if there is 2 minutes or less left to the game or something like that.

Looking at the .json of a gamelog from the nhl api it seem like it would be theoretically possible to automate this

 

[seventieslord]

Yes and getting a team ahead by 2 goals is a bigger deal than making a 5 goals deficit down to 4....

Maybe just some scoring when the goal differential is 3 goals or less, 2 if there is 2 minutes or less left to the game or something like that.

Looking at the .json of a gamelog from the nhl api it seem like it would be theoretically possible to automate this

I've seen WPA models that are very good - they use statistics derived from all games, to determine the win probability added by each goal. For example, if there are 5 minutes left and you're up by a goal and you score, you just added, for example, 20% to your win probability (from 78% to 98%). Buf if you score with 5 minutes left to go from 5 down to 4 down, your probability just went from 0.1% to 0.2%, so you did nothing.

 

[Hockey Outsider]

I've seen WPA models that are very good - they use statistics derived from all games, to determine the win probability added by each goal. For example, if there are 5 minutes left and you're up by a goal and you score, you just added, for example, 20% to your win probability (from 78% to 98%). Buf if you score with 5 minutes left to go from 5 down to 4 down, your probability just went from 0.1% to 0.2%, so you did nothing.

I was going to make the same comment. This is a concept that has a ton of potential, and I don't think it's really been explored in depth.

 

[seventieslord]

I was going to make the same comment. This is a concept that has a ton of potential, and I don't think it's really been explored in depth.

Jfresh has been doing it for the last few playoffs, but there's a ton of opportunity to take it back further, as far back as we want to go, really, since the game sheets are all public now.

 

[Hockey Outsider]

Jfresh has been doing it for the last few playoffs, but there's a ton of opportunity to take it back further, as far back as we want to go, really, since the game sheets are all public now.

I started on this analysis around 20 years ago. I don't remember which season it was on - probably 2004, but it could have been 2003 or 2006. The math behind calculating the win probability isn't overly difficult(*). The real obstacle was the difficulty in finding all the data in a useable format. Ultimately, I wasn't able to get it to work. It should be much easier now (keeping in mind that "easier" doesn't necessarily mean easy).

(*) The biggest question is - do we calculate win probabilities based on the strength of a specific team, and their opponent, or do we do the calculations based on the league average? In other words - let's assume that a team scores the go-ahead goal, with 10 minutes left in the third period. I'm making up these numbers, but to illustrate the point - that might give us 0.25 win probabilities added (ie the team goes from a 50% to 75% expected win percentage). But if the 1982 Islanders do that against the Kings, it might push the win percentage to 85%, because it's unlikely that such a bad team could come back against such a strong opponent. So maybe that goal is really worth 0.35. (On the other hand, despite it being a tie, the Islanders might have had a 60% win probability before the goal anyway). My guess is both methods will end up very similar, but there could be distortions for players on teams that are either very strong, or very poor.

 

[Staniowski]

Mark Messier's numbers when trailing are incredible, as were his linemate Glenn Anderson's. So many times when the Oilers were trailing, it was Messier's line scoring to get them back in the game. I'm not sure I know why, but it is likely one of the reason's for Messier's very high reputation as a playoff performer.

The basic reason is what I've discussed before - Messier, like Lemieux, was better equipped than Gretzky to deal with better defense, tougher opposition. Messier was one of the most powerful skaters ever, plus physically strong, could hit and play defense, and was fearless. Anderson wasn't much different.

When I think about the Oilers scoring big goals, I think first of Messier and Anderson, not Gretzky and Kurri. I think anybody who watched the Oulers would say the same.

 

[Staniowski]

Ya, I have seen multiple sources where Arbour/Torrey/players relate how the Isles had that big 116pt regular season filled with individual accolades but flamed out in the playoffs. They realized they needed to leave some in the tank and so they took their foot off the gas in the regular seasons from then on. They also improved some of their depth to allow the minutes to be spread around though.

The Oilers had a similar realization during their dynasty as well where they would jump out into a playoff spot and then coast the rest of the way before dialing back up in the playoffs.

A very good strategy in a time when most teams made the playoffs.

It was the Islanders' depth that was the key. They really didn't have it prior to 1980. They dont win any Cups without Goring, Morrow, Langevin, and the Sutters.

 

[The Panther]

The basic reason is what I've discussed before - Messier, like Lemieux, was better equipped than Gretzky to deal with better defense, tougher opposition.

Dude, you need to give this up. It's becoming like your life's mission to discredit the greatest forward of all time.

When I think about the Oilers scoring big goals, I think first of Messier and Anderson, not Gretzky and Kurri. I think anybody who watched the Oulers would say the same.

Forgive the essay, but I have to do this.... The following are all playoff games that the Oilers won during the 1980s and with Gretzky, Messier, and Kurri all as teammates:

1981 vs Montreal:
Game 1
Kurri 2 goals (both assisted by Gretzky who had 5 points)
Messier 0 points
Game 2
Kurri 1 goal (third period clincher)
Messier 0 points
Game 3
Gretzky 3 goals (4 points) to clinch series win
Messier 0 points

1981 vs. Islanders
Game 3 (Oilers already down 0-2 in series)
Gretzky 3 goals, Kurri 1 goal
Messier 1 secondary assist

1982 vs. Kings
Game 2 (Oilers down 0-1 in series)
Kurri 1 goal (late in third to send it to overtime)
Gretzky 1 goal (overtime winner)
Messier 0 points (and minus 2 on 2 Kings' goals)
Game 4 (Oilers down 1-2 in series after blowing 5-0 lead in game three)
Kurri 1 goal (ties game in first period; Oilers go on to win)
Messier 0 points

1983 vs. Winnipeg
Game 1
Gretzky 4 goals, 5 points
Messier 1 assist
1983 vs. Calgary
Game 2
Kurri 1 goal (opens scoring in 1st minute)
Messier 1 secondary assist (on meaningless late third goal)
Game 3
Gretzky 4 goals, 3 assists
Messier 3 goals (two with primary assist to Gretzky)
Game 5 (Oilers clinch series)
Gretzky 1 goal (3 points)
Messier 1 goal
1983 vs. Chicago
Game 1
Gretzky 1 goal (5 points)
Messier 1 goal (3 points)
Game 4 (series clincher)
Gretzky 1 goal
Kurri 1 goal
Messier didn't play

1984 vs. Winnipeg
Game 1
Gretzky 1 goal
Messier 0 goals
Game 4 (series clincher)
Kurri 1 goal
Messier 1 goal

1984 vs. Minnesota
Game 1
Gretzky 1 goal
Kurri 2 goals (both primary assists to Gretzky)
Messier 0 goals
Game 3
Gretzky 2 goals (second of game and penalty-shot closer)
Kurri 1 goal
Messier 1 goal

1984 vs. NY Islanders
Game 4
Gretzky 2 goals (incl. opener)
Messier 1 goal
Game 5 (Cup clincher)
Gretzky 2 goals
Kurri 1 goal
Messier 0 goals

1985 vs. Kings
Game 1
Kurri 1 goal (series opener)
Messier 0 points
1985 vs. Jets
Game 1
Kurri 1 goal (and Gretzky EN goal)
Messier 0 goals
Game 2
Gretzky 1 goal (opener)
Messier 1 goal (scores late in third when game already over)
Game 3
Gretzky 1 goal (game winner, late in third)
Messier 0 goals
Game 4 (series clincher)
Gretzky 3 goals (and 4 assists)
Kurri 3 goals (and 1 assist)
Mesier 1 goal (in third period when game was already over)

1985 vs. Chicago
Game 1
Gretzky 1 goal
Kurri 2 goals
Messier 0 goals
Game 2
Kurri 3 goals (2 on Gretzky primary assists)
Messier 0 goals
Game 5
Gretzky 2 goals
Kurri 3 goals
Messier 1 goal
Game 6 (series clincher)
Kurri 4 goals (all 4 with primary assists to Gretzky)
Messier 2 goals

1985 vs. Philadelphia
Game 2
Gretzky 1 goal (Oilers only score 2 legit goals)
Messier 0 points
Game 3
Gretzky 3 goals (and 1 assist)
Messier 1 assist (on a Gretzky goal)
Game 4
Gretzky 2 goals
Messier 1 assist

1986 vs. Vancouver
Game 1
Gretzky 1 goal
Messier 0 goals
Game 2
Gretzky 1 goal
Kurri 1 goal
Messier 1 goal
Game 3 (series clincher)
Gretzky 1 goal
Messier 1 goal

1986 vs. Calgary
Game 4
Gretzky 3 goals (and 2 assists)
Messier 1 assist
(This is the only series the Oilers lost from 1984 to 1988. Gretzky scored 5 goals and Messier 2.)

1987 vs. Kings
Game 2
Kurri 4 goals (all with Gretzky assists)
Messier 2 goals
Gretzky 1 goal
Game 3
Kurri 2 goals
Messier 2 goals
Game 4
Kurri 1 goal (and Gretzky 5 points)
Messier 1 assist

1987 vs. Winnipeg
Game 3
Kurri 2 goals
Messier 1 goal
Game 4 (series clincher)
Gretzky 1 goal
Messier 0 goals

1987 vs. Philadelphia
Game 1
Gretzky 1 goal (opens scoring)
Kurri 1 goal
Messier 0 goals
Game 2
Gretzky 1 goal (opens scoring)
Kurri 1 goal (overtime winner... largely set up by Gretzky)
Messier 0 goals
Game 4
Kurri 1 goal (and Gretzky 3 assists)
Messier 0 goals
Game 7 (Cup clincher)
Kurri 1 goal (Cup winning goal)
Messier 1 goal (opener for Edm.)

1988 vs. Winnipeg
Game 1
Kurri 1 goal
Messier 1 goal
Game 4
Kurri 2 goals (both assisted by Gretzky)
Messier 0 goals
Game 5
Gretzky 1 goal
Kurri 2 goals
Messier 2 goals

1988 vs. Calgary
Game 1
Kurri 1 goal (game winner)
Gretzky 1 goal (clincher)
Messier 1 goal (opener)
Game 2
Gretzky 2 goals (incl. overtime winner)
Kurri 1 goal (game tier)
Messier 1 goal (pass bounced off Flames' D and went in on the PP)
Game 3
Kurri 1 goal (third period clincher)
Messier 0 goals
Game 4
Gretzky 1 goal
Messier 1 goal

1988 vs. Detroit
Game 1
Kurri 1 goal (and Gretzky 3 points)
Messier 0 goals
Game 4
Kurri 2 goals (incl. overtime winner)
Messier 0 goals
Game 5 (series clincher)
Gretzky (EN goal)
Kurri 1 goal
Messier 1 goal

1988 vs. Boston
Game 1
Gretzky 1 goal (series opener)
Messier 0 goals
Game 2
Kurri EN goal
Gretzky 1 goal (game winner in third)
Messier 1 goal
Game 4 (or is it 5?) (Cup clinching game)
Gretzky 1 goal (Cup winning goal)
Messier 0 goals

___________________________

So, I guess you just forgot about those specific games, eh?

 

[overpass]

Okay, so (assuming @overpass's calculation of minutes played in various situations and scoring of Bossy/Gretzky is correct), we have the following idea based on the team-Dynasty period of each player. (Correct me if I'm wrong):
-- When Isles/Oilers were tied: Advantage Gretzky
-- When Isles/Oilers were ahead: Advantage Gretzky
-- When Isles/Oilers were behind: Advantage Bossy in points-per-minute

Do I have this correct?

Yes.

Now, is it not possible that Gretzky's large number of points when 'ahead' actually contributed rather a lot to the Oilers' not being behind? That is, if the team builds up a big lead, they aren't going to get behind.

Absolutely. I would expect most of them did contribute to some degree. There weren't that many 10-3 goals.

I can tell you that I broke the stats down further into "insurance goals" (scored with a 1-2 goal lead), and "blowout goals" (scored with a 3+ goal lead.

Gretzky had a point on 67 of 144 insurance goals by the Oilers (47%), and a point on 29 of 60 blowout goals (48%).

Bossy had a point on 32 of 120 insurance goals by the Isles (29%) and a point on 12 of 62 blowout goals (11%).

It's possible that some of the blowout goals also had value in the high-scoring 80s, But they're probably the least valuable goals, with the possible exception of goals scored while being blown out. If you wanted to label some goals as valuable, you could do worse than excluding those blowout goals. Which would leave each of Bossy and Gretzky with the following dynasty scoring lines.

Bossy: 72 GP, 54 G. 45 A, 99 P, 0.75 G/GP, 1.375 P/GP
Gretzky: 87 GP, 48 G. 101 A, 149 P, 0.55 G/GP, 1.71 P/GP

So 1/3 of Gretzky's point per game edge in the playoffs game was scored on blowout goals. The rest of his advantage was on insurance goals.

Another point might be: How many of the respective points-scored-when-behind were relevant in a team victory? I mean, is it not possible (I'm just asking) that one player scored a lot when down, but his team usually lost those games (i.e., kind of empty points, in the end)?

To put it another way, scoring points to add to a 2-0 and 3-0 lead and then winning are actually more valuable points than scoring two points when down 6-2 and losing 6-4.

There are various ways to look at these things...

The stats definitely need some interpretation and I agree you could draw different conclusions.

The one stat that really surprised me was that the Islanders, as a team during the dynasty years, scored at a slightly higher rate with the lead than the dynasty Oilers did. And that was with Bossy scoring on a relatively low percentage of those goals with the lead. They went 10 forwards deep in those situations, with Duane Sutter, Wayne Merrick, and others scoring a lot of their points with the lead.

So does that mean that Bossy had better team support with the Islanders than Gretzky did?
Or does it mean that the Oilers could have done just as well by playing Gretzky 22 minutes/game as 27 minutes/game, even if Gretzky scored fewer points with the lead and their third line contributed more? (I think they actually did this to some degree in the 1984 playoffs).

This is the kind of analysis we need more of!

I don't know if you did all this manually or if you used some sort of script to scrape the data, but I wonder if it could be further refined to situational time, since we know what penalties were called and when.

I mean, I guess I know that the answer is yes it can be further refined, but whether you did this manually or with a script, it would make collecting the data significantly more difficult.

Manually enough that I'm not going to split out the situational time.

But I can split out the EV/PP/SH points pretty easily.

Mike Bossy, dynasty playoff scoring
Bossy trailing: 919 team TOI, EV (5 G, 2 A, 7 ), PP (8 G, 14 A, 22 P)
Bossy tied: 1388 team TOI, EV (13 G, 8 A, 21 P), PP (8 G, 9 A, 17 P)
Bossy leading: 2109 team TOI, EV (15 G, 10 A, 25 P), PP (10 G, 9 A, 19 P)

Rates (using total team TOI, so a higher PP than EV rate means that he literally scored more PP points in that situation)
Bossy trailing: 0.33 EVG/60, 0.46 EVP/60, 0.52 PPG/60, 1.44 PPP/60
Bossy tied: 0.56 EVG/60, 0.91 EVP/60, 0.35 PPG/60, 0.73 PPP/60
Bossy leading: 0.43 EVG, 0.71 EVP/60, 0.28 PPG/60, 0.54 PPP/60

Bossy was incredibly productive on the power play when trailing. And his power play production declined even further with the lead. This is why I'm hesitant to label his power play scoring as inferior in any way. The timing was extremely valuable. Al Arbour probably spread the power play time around more when the Islanders were leading.

He wasn't very productive at even strength while trailing, his even strength scoring was best while tied.

Wayne Gretzky, dynasty playoff scoring
Gretzky trailing: 1066 team TOI, EV (4 G, 14 A, 18 P ), PP (3 G, 10 A, 13 P), SH (1 A, 1 P)
Gretzky tied: 1562 team TOI, EV (9 G, 21 A, 30 P), PP (6 G, 9 A, 15 P), SH (2 G, 2 A, 4 P)
Gretzky leading: 2606 team TOI, EV (21 G, 41 A, 62 P), PP (8 G, 20 A, 28 P), SH (3 G, 5 A, 8 P)

Rates (again using total team TOI)
Gretzky trailing: 0.23 EVG/60, 1.01 EVP/60, 0.17 PPG/60, 0.73 PPP/60, 0.06 SHP/60
Gretzky tied: 0.35 EVG/60, 1.15 EVP/60, 0.23 PPG/60, 0.58 PPP/60, 0.08 SHG/60, 0.15 SHP/60
Gretzky leading: 0.48 EVG/60, 1.43 EVP/60, 0.18 PPG/60, 0.64 PPP/60, 0.07 SHG/60, 0.18 SHP/60

Gretzky's EV goal and point production was lowest when trailing and highest when leading. Power play scoring was more even, and his rate of point production was best when trailing.

 

[overpass]

Depending on how labour intensive collecting this data is, there are all kinds of possibilities this opens up with respect to Win Probability Added (WPA) models.

Currently all goals scored are being placed into three buckets - behind, ahead, and tied. But those are generalizations. For example, a goal scored when you're one ahead is more important than one scored when you are three behind.

And the time remaining in the game is really important to consider as well - The goal scored when three behind is more important the more time is remaining in the game, and the goal scored when one ahead is more important the less time is remaining (to an extent).

And this is all data that is available. Without knowing how it's collected, I just don't know how realistic it is to hope it can be collected for numerous players over a period of many years.

Agreed. I've played around with a Poisson distribution model for WPA, which is based on score, time remaining, and the expected number of goals per game. The Poisson distribution assumes that a goal is equally likely to be scored at any time, which isn't accurate for hockey due to power plays and late game strategies, so I think the model could be improved.

(*) The biggest question is - do we calculate win probabilities based on the strength of a specific team, and their opponent, or do we do the calculations based on the league average? In other words - let's assume that a team scores the go-ahead goal, with 10 minutes left in the third period. I'm making up these numbers, but to illustrate the point - that might give us 0.25 win probabilities added (ie the team goes from a 50% to 75% expected win percentage). But if the 1982 Islanders do that against the Kings, it might push the win percentage to 85%, because it's unlikely that such a bad team could come back against such a strong opponent. So maybe that goal is really worth 0.35. (On the other hand, despite it being a tie, the Islanders might have had a 60% win probability before the goal anyway). My guess is both methods will end up very similar, but there could be distortions for players on teams that are either very strong, or very poor.

I found the same. I wanted to use a model that estimated the number of goals likely to be scored by each team for that particular game. But then you might have the game start with one team having a 70% probability to win. Players on strong teams have a hard time getting a lot of WPA because the model is basically giving out the same amount of WPA to each team.

I ran the numbers for some of the top goal scoring seasons of all time. Brett Hull's 90-91 stood out as by far the most valuable from a WPA perspective.

 

[Hockey Outsider]

Agreed. I've played around with a Poisson distribution model for WPA, which is based on score, time remaining, and the expected number of goals per game. The Poisson distribution assumes that a goal is equally likely to be scored at any time, which isn't accurate for hockey due to power plays and late game strategies, so I think the model could be improved.



I found the same. I wanted to use a model that estimated the number of goals likely to be scored by each team for that particular game. But then you might have the game start with one team having a 70% probability to win. Players on strong teams have a hard time getting a lot of WPA because the model is basically giving out the same amount of WPA to each team.

I ran the numbers for some of the top goal scoring seasons of all time. Brett Hull's 90-91 stood out as by far the most valuable from a WPA perspective.

You may have already seen this, but here's an interesting paper by Alan Ryder.

His conclusion is hockey is "Poisson enough" to be modelled by that probability distribution. There are a few challenges (scoring is lower during the first minute of each period, scoring is much higher during the final two minutes of the third, and overtime can't follow this distribution). But it seems to be close enough for a rough model.

(Interestingly, Poisson "works" even when you don't directly factor in ES/PP/SH time. Presumably that's because it's taken into account, albeit indirectly, by the overall scoring environment).

 

[overpass]

You may have already seen this, but here's an interesting paper by Alan Ryder.

His conclusion is hockey is "Poisson enough" to be modelled by that probability distribution. There are a few challenges (scoring is lower during the first minute of each period, scoring is much higher during the final two minutes of the third, and overtime can't follow this distribution). But it seems to be close enough for a rough model.

(Interestingly, Poisson "works" even when you don't directly factor in ES/PP/SH time. Presumably that's because it's taken into account, albeit indirectly, by the overall scoring environment).

Yeah, it's definitely close enough for a model, a model doesn't have to be perfect.

I would just make a note that the Poisson model might undervalue late goals with a small lead a little, because it doesn't recognize that the trailing team has a higher chance of tying the game with the goalie pulled. In other words, it might undervalue late-game empty net goals. And Gretzky scored a lot in those situations compared to, say, Bossy and Brett Hull, so the Poisson might undervalue Gretzky's scoring a little.

I ran Bossy and Gretzky's playoff dynasty scoring through the Poisson WPA model. Quite simply, the model estimates the team's chance of winning before the goal was scored, estimates their chance of winning after the goal was scored, and the difference is the Win Probability Added (WPA) for that goal.

WPA puts the highest value on goals scored late in close games. The most valuable are overtime goals, which take the game from a 50% chance of winning to a 100% chance of winning, so they are worth 0.5 WPA. The least valuable are goals scored late in games that aren't close. Goals scored early in games are moderately valuable, and the value doesn't depend as much on the score of the game, because there's still a lot of unknowns and a lot of hockey left to play.

Here are the results for Bossy and Gretzky, based on their point scoring (i.e. the value of the goals on which they scored the goal or got an assist).

Wayne Gretzky: 87 GP, 178 points, 23.66 WPA (average point = 0.133 WPA).
Mike Bossy: 72 GP, 111 points, 16.35 WPA (average point 0.147 WPA).

Looking on a per-game basis, Gretzky was 33% ahead of Bossy in dynasty PO points per game (2.05 to 1.54). He was 20% ahead in WPA/game (0.272 to 0.227). So the WPA model does close the offensive gap in value between Gretzky and Bossy somewhat.

Just to give you an idea of how much WPA goals in different situations are worth, here's a breakdown by game situation. Again, using the dynasty playoff point scoring for Gretzky and Bossy.

Bossy
Scored a point down by 2+ - 9 points, 0.98 WPA, 0.109 WPA/point
Scored a point down by 1 - 21 points, 4.46 WPA, 0.212 WPA/point
Scored a point while tied - 38 points, 6.88 WPA, 0.181 WPA/point
Scored a point while up 1 - 21 points, 2.67 WPA, 0.127 WPA/point
Scored a point while up 2 - 10 points, 1.01 WPA, 0.101 WPA/point
Scored a point while up 3+ - 12 points, 0.35 WPA, 0.029 WPA/point

Gretzky
Scored a point down by 2+ - 8 points, 1.06 WPA, 0.133 WPA/point
Scored a point down by 1 - 24 points, 4.85 WPA, 0.202 WPA/point
Scored a point while tied - 50 points, 9.90 WPA, 0.198 WPA/point
Scored a point while up 1 - 49 points, 5.69 WPA, 0.116 WPA/point
Scored a point while up 2 - 18 points, 1.48 WPA, 0.082 WPA/point
Scored a point while up 3+ - 29 points, 0.68 WPA, 0.023 WPA/point

As we would expect, goals/points that tie the game or break a tie are the most valuable.

Gretzky tended to score points later in the game than Bossy, which explains why his points while tied were a little more valuable on average than Bossy's and why his points while leading were a little less valuable than Bossy's. For example, he had 3 OT points to Bossy's 1, and those are the most valuable with a value of 0.5 WPA.

Here's an illustration of the difference in late-game scoring, looking at points scored in the last 10 minutes of regulation and OT. Gretzky had a lot of relatively low-value points while his team was leading late, but his 5 points while tied late (including 3 in OT) gave him the edge in late value over Bossy.

Gretzky in final 10/OT
Scored a point down by 2+ - 2 points, 0.076 WPA, 0.038 WPA/point
Scored a point down by 1 - 2 points, 0.816 WPA, 0.408 WPA/point
Scored a point while tied - 5 points, 2.227 WPA, 0.445 WPA/point
Scored a point while up 1 - 8 points, 0.324 WPA, 0.040 WPA/point
Scored a point while up 2 - 8 points, 0.199 WPA, 0.025 WPA/point
Scored a point while up 3+ -9 points, 0.012 WPA, 0.001 WPA/point
All points in final 10/OT - 34 points, 3.654 WPA, 0.107 WPA/point

Bossy in final 10/OT
Scored a point down by 2+ - 3 points, 0.196 WPA, 0.065 WPA/point
Scored a point down by 1 - 3 points, 1.184 WPA, 0.395 WPA/point
Scored a point while tied - 2 points, 0.800 WPA, 0.400 WPA/point
Scored a point while up 1 - 4 points, 0.300 WPA, 0.075 WPA/point
Scored a point while up 2 - 3 points, 0.056 WPA, 0.019 WPA/point
Scored a point while up 3+ -2 points, 0.004 WPA, 0.002 WPA/point
All points in final 10/OT - 17 points, 2.54 WPA, 0.149 WPA/point

And points scored in the first 10 minutes of the game. Their value varies less with the score because there is still a lot of time to play.

Gretzky in first 10
Scored a point down by 2+ - 0 points
Scored a point down by 1 - 4 points, 0.577 WPA, 0.144 WPA/point
Scored a point while tied - 15 points, 2.069 WPA, 0.138 WPA/point
Scored a point while up 1 - 4 points, 0.450 WPA, 0.113 WPA/point
Scored a point while up 2 - 1 points, 0.110 WPA, 0.110 WPA/point
All points in first 10 - 24 points, 3.206 WPA, 0.134 WPA/point

Bossy in first 10
Scored a point down by 2+ - 1 points, 0.140 WPA, 0.140 WPA/point
Scored a point down by 1 - 2 points, 0.464 WPA, 0.155 WPA/point
Scored a point while tied - 13 points, 1.894 WPA, 0.146 WPA/point
Scored a point while up 1 - 1 points, 0.139 WPA, 0.139 WPA/point
Scored a point while up 2 - 0 points
All points in first 10 - 17 points, 2.637 WPA, 0.155 WPA/point

Finally, why did Bossy usually have slightly higher average WPA/points than Gretzky even in the same situations? Because I used RS team GF and GA and opponent GF and GA to calculate the scoring environment. Bossy played for teams that scored and allowed fewer goals, and against opponents that scored and allowed fewer goals. So each Bossy goal was a little more valuable on average due to the lower scoring environment, when compared to Gretzky.

 

[Hockey Outsider]

Yeah, it's definitely close enough for a model, a model doesn't have to be perfect.

I would just make a note that the Poisson model might undervalue late goals with a small lead a little, because it doesn't recognize that the trailing team has a higher chance of tying the game with the goalie pulled. In other words, it might undervalue late-game empty net goals. And Gretzky scored a lot in those situations compared to, say, Bossy and Brett Hull, so the Poisson might undervalue Gretzky's scoring a little.

I ran Bossy and Gretzky's playoff dynasty scoring through the Poisson WPA model. Quite simply, the model estimates the team's chance of winning before the goal was scored, estimates their chance of winning after the goal was scored, and the difference is the Win Probability Added (WPA) for that goal.

WPA puts the highest value on goals scored late in close games. The most valuable are overtime goals, which take the game from a 50% chance of winning to a 100% chance of winning, so they are worth 0.5 WPA. The least valuable are goals scored late in games that aren't close. Goals scored early in games are moderately valuable, and the value doesn't depend as much on the score of the game, because there's still a lot of unknowns and a lot of hockey left to play.

Here are the results for Bossy and Gretzky, based on their point scoring (i.e. the value of the goals on which they scored the goal or got an assist).

Wayne Gretzky: 87 GP, 178 points, 23.66 WPA (average point = 0.133 WPA).
Mike Bossy: 72 GP, 111 points, 16.35 WPA (average point 0.147 WPA).

Looking on a per-game basis, Gretzky was 33% ahead of Bossy in dynasty PO points per game (2.05 to 1.54). He was 20% ahead in WPA/game (0.272 to 0.227). So the WPA model does close the offensive gap in value between Gretzky and Bossy somewhat.

Just to give you an idea of how much WPA goals in different situations are worth, here's a breakdown by game situation. Again, using the dynasty playoff point scoring for Gretzky and Bossy.

Bossy
Scored a point down by 2+ - 9 points, 0.98 WPA, 0.109 WPA/point
Scored a point down by 1 - 21 points, 4.46 WPA, 0.212 WPA/point
Scored a point while tied - 38 points, 6.88 WPA, 0.181 WPA/point
Scored a point while up 1 - 21 points, 2.67 WPA, 0.127 WPA/point
Scored a point while up 2 - 10 points, 1.01 WPA, 0.101 WPA/point
Scored a point while up 3+ - 12 points, 0.35 WPA, 0.029 WPA/point

Gretzky
Scored a point down by 2+ - 8 points, 1.06 WPA, 0.133 WPA/point
Scored a point down by 1 - 24 points, 4.85 WPA, 0.202 WPA/point
Scored a point while tied - 50 points, 9.90 WPA, 0.198 WPA/point
Scored a point while up 1 - 49 points, 5.69 WPA, 0.116 WPA/point
Scored a point while up 2 - 18 points, 1.48 WPA, 0.082 WPA/point
Scored a point while up 3+ - 29 points, 0.68 WPA, 0.023 WPA/point

As we would expect, goals/points that tie the game or break a tie are the most valuable.

Gretzky tended to score points later in the game than Bossy, which explains why his points while tied were a little more valuable on average than Bossy's and why his points while leading were a little less valuable than Bossy's. For example, he had 3 OT points to Bossy's 1, and those are the most valuable with a value of 0.5 WPA.

Here's an illustration of the difference in late-game scoring, looking at points scored in the last 10 minutes of regulation and OT. Gretzky had a lot of relatively low-value points while his team was leading late, but his 5 points while tied late (including 3 in OT) gave him the edge in late value over Bossy.

Gretzky in final 10/OT
Scored a point down by 2+ - 2 points, 0.076 WPA, 0.038 WPA/point
Scored a point down by 1 - 2 points, 0.816 WPA, 0.408 WPA/point
Scored a point while tied - 5 points, 2.227 WPA, 0.445 WPA/point
Scored a point while up 1 - 8 points, 0.324 WPA, 0.040 WPA/point
Scored a point while up 2 - 8 points, 0.199 WPA, 0.025 WPA/point
Scored a point while up 3+ -9 points, 0.012 WPA, 0.001 WPA/point
All points in final 10/OT - 34 points, 3.654 WPA, 0.107 WPA/point

Bossy in final 10/OT
Scored a point down by 2+ - 3 points, 0.196 WPA, 0.065 WPA/point
Scored a point down by 1 - 3 points, 1.184 WPA, 0.395 WPA/point
Scored a point while tied - 2 points, 0.800 WPA, 0.400 WPA/point
Scored a point while up 1 - 4 points, 0.300 WPA, 0.075 WPA/point
Scored a point while up 2 - 3 points, 0.056 WPA, 0.019 WPA/point
Scored a point while up 3+ -2 points, 0.004 WPA, 0.002 WPA/point
All points in final 10/OT - 17 points, 2.54 WPA, 0.149 WPA/point

And points scored in the first 10 minutes of the game. Their value varies less with the score because there is still a lot of time to play.

Gretzky in first 10
Scored a point down by 2+ - 0 points
Scored a point down by 1 - 4 points, 0.577 WPA, 0.144 WPA/point
Scored a point while tied - 15 points, 2.069 WPA, 0.138 WPA/point
Scored a point while up 1 - 4 points, 0.450 WPA, 0.113 WPA/point
Scored a point while up 2 - 1 points, 0.110 WPA, 0.110 WPA/point
All points in first 10 - 24 points, 3.206 WPA, 0.134 WPA/point

Bossy in first 10
Scored a point down by 2+ - 1 points, 0.140 WPA, 0.140 WPA/point
Scored a point down by 1 - 2 points, 0.464 WPA, 0.155 WPA/point
Scored a point while tied - 13 points, 1.894 WPA, 0.146 WPA/point
Scored a point while up 1 - 1 points, 0.139 WPA, 0.139 WPA/point
Scored a point while up 2 - 0 points
All points in first 10 - 17 points, 2.637 WPA, 0.155 WPA/point

Finally, why did Bossy usually have slightly higher average WPA/points than Gretzky even in the same situations? Because I used RS team GF and GA and opponent GF and GA to calculate the scoring environment. Bossy played for teams that scored and allowed fewer goals, and against opponents that scored and allowed fewer goals. So each Bossy goal was a little more valuable on average due to the lower scoring environment, when compared to Gretzky.

Thanks for putting that together. A longer-term project would be to automate the calculation, so the data could be fed in for each season, and we'd get the results within having to do any manual calculations. I have some ideas on how to set it up, but not sure if it's feasible yet. I think it would also be interesting from the opposite standpoint - which goalies added/detracted the most from their team's chances of winning with timely saves (or goals allowed).

 

[authentic]

Yeah, it's definitely close enough for a model, a model doesn't have to be perfect.

I would just make a note that the Poisson model might undervalue late goals with a small lead a little, because it doesn't recognize that the trailing team has a higher chance of tying the game with the goalie pulled. In other words, it might undervalue late-game empty net goals. And Gretzky scored a lot in those situations compared to, say, Bossy and Brett Hull, so the Poisson might undervalue Gretzky's scoring a little.

I ran Bossy and Gretzky's playoff dynasty scoring through the Poisson WPA model. Quite simply, the model estimates the team's chance of winning before the goal was scored, estimates their chance of winning after the goal was scored, and the difference is the Win Probability Added (WPA) for that goal.

WPA puts the highest value on goals scored late in close games. The most valuable are overtime goals, which take the game from a 50% chance of winning to a 100% chance of winning, so they are worth 0.5 WPA. The least valuable are goals scored late in games that aren't close. Goals scored early in games are moderately valuable, and the value doesn't depend as much on the score of the game, because there's still a lot of unknowns and a lot of hockey left to play.

Here are the results for Bossy and Gretzky, based on their point scoring (i.e. the value of the goals on which they scored the goal or got an assist).

Wayne Gretzky: 87 GP, 178 points, 23.66 WPA (average point = 0.133 WPA).
Mike Bossy: 72 GP, 111 points, 16.35 WPA (average point 0.147 WPA).

Looking on a per-game basis, Gretzky was 33% ahead of Bossy in dynasty PO points per game (2.05 to 1.54). He was 20% ahead in WPA/game (0.272 to 0.227). So the WPA model does close the offensive gap in value between Gretzky and Bossy somewhat.

Just to give you an idea of how much WPA goals in different situations are worth, here's a breakdown by game situation. Again, using the dynasty playoff point scoring for Gretzky and Bossy.

Bossy
Scored a point down by 2+ - 9 points, 0.98 WPA, 0.109 WPA/point
Scored a point down by 1 - 21 points, 4.46 WPA, 0.212 WPA/point
Scored a point while tied - 38 points, 6.88 WPA, 0.181 WPA/point
Scored a point while up 1 - 21 points, 2.67 WPA, 0.127 WPA/point
Scored a point while up 2 - 10 points, 1.01 WPA, 0.101 WPA/point
Scored a point while up 3+ - 12 points, 0.35 WPA, 0.029 WPA/point

Gretzky
Scored a point down by 2+ - 8 points, 1.06 WPA, 0.133 WPA/point
Scored a point down by 1 - 24 points, 4.85 WPA, 0.202 WPA/point
Scored a point while tied - 50 points, 9.90 WPA, 0.198 WPA/point
Scored a point while up 1 - 49 points, 5.69 WPA, 0.116 WPA/point
Scored a point while up 2 - 18 points, 1.48 WPA, 0.082 WPA/point
Scored a point while up 3+ - 29 points, 0.68 WPA, 0.023 WPA/point

As we would expect, goals/points that tie the game or break a tie are the most valuable.

Gretzky tended to score points later in the game than Bossy, which explains why his points while tied were a little more valuable on average than Bossy's and why his points while leading were a little less valuable than Bossy's. For example, he had 3 OT points to Bossy's 1, and those are the most valuable with a value of 0.5 WPA.

Here's an illustration of the difference in late-game scoring, looking at points scored in the last 10 minutes of regulation and OT. Gretzky had a lot of relatively low-value points while his team was leading late, but his 5 points while tied late (including 3 in OT) gave him the edge in late value over Bossy.

Gretzky in final 10/OT
Scored a point down by 2+ - 2 points, 0.076 WPA, 0.038 WPA/point
Scored a point down by 1 - 2 points, 0.816 WPA, 0.408 WPA/point
Scored a point while tied - 5 points, 2.227 WPA, 0.445 WPA/point
Scored a point while up 1 - 8 points, 0.324 WPA, 0.040 WPA/point
Scored a point while up 2 - 8 points, 0.199 WPA, 0.025 WPA/point
Scored a point while up 3+ -9 points, 0.012 WPA, 0.001 WPA/point
All points in final 10/OT - 34 points, 3.654 WPA, 0.107 WPA/point

Bossy in final 10/OT
Scored a point down by 2+ - 3 points, 0.196 WPA, 0.065 WPA/point
Scored a point down by 1 - 3 points, 1.184 WPA, 0.395 WPA/point
Scored a point while tied - 2 points, 0.800 WPA, 0.400 WPA/point
Scored a point while up 1 - 4 points, 0.300 WPA, 0.075 WPA/point
Scored a point while up 2 - 3 points, 0.056 WPA, 0.019 WPA/point
Scored a point while up 3+ -2 points, 0.004 WPA, 0.002 WPA/point
All points in final 10/OT - 17 points, 2.54 WPA, 0.149 WPA/point

And points scored in the first 10 minutes of the game. Their value varies less with the score because there is still a lot of time to play.

Gretzky in first 10
Scored a point down by 2+ - 0 points
Scored a point down by 1 - 4 points, 0.577 WPA, 0.144 WPA/point
Scored a point while tied - 15 points, 2.069 WPA, 0.138 WPA/point
Scored a point while up 1 - 4 points, 0.450 WPA, 0.113 WPA/point
Scored a point while up 2 - 1 points, 0.110 WPA, 0.110 WPA/point
All points in first 10 - 24 points, 3.206 WPA, 0.134 WPA/point

Bossy in first 10
Scored a point down by 2+ - 1 points, 0.140 WPA, 0.140 WPA/point
Scored a point down by 1 - 2 points, 0.464 WPA, 0.155 WPA/point
Scored a point while tied - 13 points, 1.894 WPA, 0.146 WPA/point
Scored a point while up 1 - 1 points, 0.139 WPA, 0.139 WPA/point
Scored a point while up 2 - 0 points
All points in first 10 - 17 points, 2.637 WPA, 0.155 WPA/point

Finally, why did Bossy usually have slightly higher average WPA/points than Gretzky even in the same situations? Because I used RS team GF and GA and opponent GF and GA to calculate the scoring environment. Bossy played for teams that scored and allowed fewer goals, and against opponents that scored and allowed fewer goals. So each Bossy goal was a little more valuable on average due to the lower scoring environment, when compared to Gretzky.

So that’s on top of his ice time being 35-40% less, which would indicate the gap between them is drastically smaller than the raw totals would indicate. If this is enough to consider Bossy the better goal scorer, it also would seem that in Gretzky’s place he could’ve scored 170-80 points or something… I mean if we’re also factoring in that he wouldn’t have the same defensive responsibilities that could actually be lowballing it.

 

[overpass]

So that’s on top of his ice time being 35-40% less, which would indicate the gap between them is drastically smaller than the raw totals would indicate. If this is enough to consider Bossy the better goal scorer, it also would seem that in Gretzky’s place he could’ve scored 170-80 points or something… I mean if we’re also factoring in that he wouldn’t have the same defensive responsibilities that could actually be lowballing it.

Yeah, I could see a 170+ point season in the early 80s if Bossy played for a coach who really fed him all the minutes and offensive opportunity. But then his career is probably even shorter, and I doubt he would hold up for four playoff runs.

Rather than making regular season points the benchmark, I'd rather just say that Bossy's scoring was extremely valuable in a way that isn't captured by the usual metrics.