Pavel Bure's ice time in 2000-01 season

[overpass]

Edit: I found that all games where a forward recorded over 30:00 TOI had that TOI edited down, usually cut in half. This change was made on both nhl.com and hockey-reference sometime in 2023.

As a result, any forward who legitimately played more than 30 minutes in a regular season game from 1997-98 to 2022-23 now has incorrect TOI stats. Each game of 30+ TOI that was edited subtracted at least 15 minutes from the player's TOI.

I'm pretty sure the ice time stats for Pavel Bure's 2000-01 have changed, on both nhl.com and hockey-reference, at some point in the past few years. And I think the current stats are wrong.

Has anyone else noticed this issue with any players?

Recently I saw that nhl.com and hockey-reference both have Bure with an average time on ice of 23:14 for the 2000-01 season. But I swear I remember his average TOI for that season was displayed as 26:52 in the past. Here's a link to an old HF post that gives the same number, so I'm not just imagining things.

Here's Bure's 2000-01 game log from hockey-reference. Just looking through here, I see his median TOI was 24:32, well above the average of 23:14. Here's a breakdown of listed his TOI/game, so you can see how it's a skewed distribution.

29-30 TOI: 3
28-29 TOI: 7
27-28 TOI: 6
26-27 TOI: 10
25-26 TOI: 12
24-25 TOI: 6
23-24 TOI: 8
22-23 TOI: 5
21-22 TOI: 2
20-21 TOI: 2
19-20 TOI: 1
18-19 TOI: 5
17-18 TOI: 1
16-17 TOI: 6
15-16 TOI: 8

The part in green looks like a normal distribution centred around 25 or 26 TOI per game, tapering off around 30 TOI at the top and 20-22 TOI on the bottom. But then there's a weird second distribution where Bure had a bunch of games with under 20 TOI, including 14 with 15-17 TOI.

Let's look at a couple of these games where Bure played under 20 minutes.

March 3, 2001, 2-2 vs ATL

• The game was 65 minutes long, so TOI for skaters would be 325 if there were 5 skaters on at all times. Florida's skaters only sum to 288 TOI.
• Florida took a few penalties and played a fair bit of 4-on-5 so their skater TOI should be about 310. 22 minutes missing.
• Bure is listed with 18:09 in ice time, 8th on the team, behind D-men Svehla, Hedican, Simpson, Barrie and forwards Kozlov, Nilson, Sillinger.
• Bure was 3rd in shifts with 35, behind only Svehla and Hedican. If he really played only 18:09, he averaged 31 seconds per shift
• Bure took 13 shots and scored 0 goals.

Feb 14, 2001, 4-2 vs PHX

• Game was 60 minutes long. Florida took 4 minors and allowed 1 PPG, so team TOI should sum to about 293
• Team TOI sums to only 271.
• Bure is listed with 16:35 ice time, 7th among FLA skaters and 4th among FLA forwards (behind Kozlov, Nilson, Niedermayer)
• Bure is listed with 29 shifts, #2 on FLA behind Svehla's 30, and #1 among FLA forwards. If he played 16:35, his shifts averaged 34 seconds.
• Bure took 3 shots and scored 3 goals.

There's over 20 minutes of missing TOI for FLA in both games. And Bure's TOI is weirdly low, even though he played a lot of shifts. I'm pretty sure Bure's TOI is not correct for either of these games, or for many others.

I would say probably any game where Bure had an average TOI/shift of under 45 seconds is wrong.

Date	Tm		Opp		G	A	PTS	+/-	SH	S%	SHFT	TOI	TOI/SHFT
2001-02-28	FLA	@	NYR	L	0	0	0	-1	10	0	24	16:56	42​
2001-03-20	FLA	@	MTL	T	1	1	2	1	5	20	26	18:08	42​
2001-03-28	FLA		MTL	T	0	0	0	-1	4	0	27	18:30	41​
2001-03-16	FLA		PIT	L	2	0	2	-1	12	16.7	25	16:39	40​
2001-02-24	FLA	@	NYI	L	1	0	1	-2	6	16.7	25	16:25	39​
2001-03-17	FLA		TOR	L	3	0	3	1	7	42.9	25	16:15	39​
2001-04-05	FLA	@	WSH	L	0	0	0	-1	2	0	24	15:20	38​
2001-02-26	FLA	@	NJD	L	2	0	2	-2	5	40	24	15:18	38​
2001-03-07	FLA		SJS	T	1	0	1	-1	8	12.5	34	20:06	35​
2001-01-24	FLA	@	WSH	L	0	0	0	0	6	0	26	15:12	35​
2001-02-14	FLA		PHX	W	3	0	3	1	3	100	29	16:35	34​
2001-04-02	FLA		BUF	L	0	1	1	-1	1	0	29	16:35	34​
2001-02-16	FLA		BOS	W	1	1	2	2	3	33.3	29	15:41	32​
2001-02-21	FLA	@	PIT	L-OT	1	0	1	1	4	25	29	15:40	32​
2001-01-12	FLA		CAR	T	1	0	1	0	4	25	29	15:08	31​
2001-03-03	FLA	@	ATL	T	0	1	1	0	13	0	35	18:09	31​
2001-03-23	FLA		WSH	W	2	0	2	2	4	50	32	15:13	29​
2001-03-02	FLA		ATL	L	1	1	2	-1	9	11.1	35	15:40	27​

What about Bure's 1999-00 season? There are a few weird TOI totals in there as well, where he played a lot of shifts but not a lot of TOI. Including one game where he is listed with 29 shifts and 0 TOI. Overall I think the 1999-00 numbers have fewer errors than the 2000-01 numbers, but I think Bure's listed TOI is low for that season too.

Do you guys agree that something is wrong here? Any thoughts on how/why it happened, and have you noticed anything similar for other players?

Click to read more...

 

[Bear of Bad News]

I haven't noticed this specifically with Bure but I have noticed some of the first "official" years of TOI data have games with 0:00 TOI (but you can't tell unless you drill down) and then the season TOI is too low because those are weighted in.

 

[overpass]

Looking at the Wayback Machine for Bure's player page, it appears the change was made between May and November of 2023. Bure's average ice time for his time in Florida dropped from 25:24 to 23:18.

May 2023

November 2023

 

[overpass]

And I was able to pull up Bure's game logs in the Internet Archive and find the differences. Notice that every game originally had more than 30 minutes TOI for Bure, and now has much less. And notice that now Bure's TOI for a game tops out at 29 and change.

My guess is that they subtracted 15+ minutes of TOI from every game where Bure had more than 30 minutes TOI. Maybe it was some kind of data validation process where they were trying to remove incorrect high outliers, but didn't realize that they were cutting off legitimate high values.

Date	Tm		Opp	SHFT	TOI (previous)	TOI (current)
2000-01-17	FLA		PHI	32	30:07:00	15:04
2000-01-21	FLA	@	ATL	28	30:25:00	15:13
2000-02-14	FLA	@	MTL	25	30:32:00	15:16
2000-04-01	FLA		TBL	29	31:45:00	15:53
2001-01-12	FLA		CAR	29	30:16:00	15:08
2001-01-24	FLA	@	WSH	26	30:24:00	15:12
2001-02-14	FLA		PHX	29	33:09:00	16:35
2001-02-16	FLA		BOS	29	31:21:00	15:41
2001-02-21	FLA	@	PIT	29	31:20:00	15:40
2001-02-24	FLA	@	NYI	25	32:50:00	16:25
2001-02-26	FLA	@	NJD	24	30:35:00	15:18
2001-02-28	FLA	@	NYR	24	33:52:00	16:56
2001-03-02	FLA		ATL	35	31:19:00	15:40
2001-03-03	FLA	@	ATL	35	36:17:00	18:09
2001-03-07	FLA		SJS	34	40:12:00	20:06
2001-03-16	FLA		PIT	25	33:18:00	16:39
2001-03-17	FLA		TOR	25	32:30:00	16:15
2001-03-20	FLA	@	MTL	26	36:15:00	18:08
2001-03-23	FLA		WSH	32	30:25:00	15:13
2001-03-28	FLA		MTL	27	37:00:00	18:30
2001-04-02	FLA		BUF	29	33:09:00	16:35
2001-04-05	FLA	@	WSH	24	30:39:00	15:20

 

[overpass]

Looking at all forwards on nhl.com and on hockey-reference, it looks like there are no longer any records of games where a forward had over 30 minutes TOI, with the exception of MacKinnon and Rantanen in one 2024 game. I think they must have done some kind of data validation which edited the TOI down for all games in which a forward had 30+ TOI recorded. Possibly before that 2024 game.

NHL Stats

The official source for NHL Stats including skaters, goalies, teams stats and more.

www.nhl.com

https://stathead.com/hockey/player-game-finder.cgi?request=1&order_by=goals&timeframe=seasons&year_min=1998&player_game_min=1&player_game_max=9999&ccomp%5B1%5D=gt&cval%5B1%5D=30&cstat%5B1%5D=time_on_ice&pos=F

But, per nhl.com, there were 20 instances where a forward had between 29:50 and 30:00 in TOI in a game! So the hard cut at 30 was obviously a bad decision and has led them to post some incorrect numbers.

I've sent feedback to hockey-reference to see if they're interested in fixing this issue. If anyone has a contact at nhl.com, would you mind bringing this issue to their attention?

 

[BraveCanadian]

Not a good look on their part, but great catch @overpass

 

[67 others]

Wow. They chopped Ray Bourque's TOI too

 

[overpass]

Wow. They chopped Ray Bourque's TOI too

They did! Good find.

The following games from 1998-99 regular season had Bourque's TOI reduced in the game log since 2019.

Rk	Date	G	Age	Tm		Opp		G	A	PTS	+/-	PIM	S	S%	SHFT	TOI (2019)	TOI (2024)
14	1998-11-07	14	37-314	BOS	@	PIT	T	0	0	0	0	0	4	0	33	35:12:00	17:36
15	1998-11-08	15	37-315	BOS	@	CAR	W	0	2	2	0	4	6	0	27	33:14:00	16:37
19	1998-11-21	19	37-328	BOS		WSH	W	0	2	2	1	0	7	0	30	39:20:00	19:40
40	1999-01-16	40	38-019	BOS		TBL	T	0	1	1	0	0	8	0	30	37:21:00	18:41
45	1999-01-30	45	38-033	BOS	@	PIT	L	0	0	0	-1	0	1	0	23	33:42:00	16:51
47	1999-02-02	47	38-036	BOS		COL	L	0	0	0	-1	0	1	0	25	34:36:00	17:18
48	1999-02-04	48	38-038	BOS		NYI	L	0	0	0	-1	0	6	0	21	34:29:00	17:15
49	1999-02-06	49	38-040	BOS	@	PHI	T	0	1	1	0	0	2	0	34	35:09:00	17:35
50	1999-02-07	50	38-041	BOS		NYR	W	0	1	1	-1	0	2	0	23	32:10:00	16:05
51	1999-02-09	51	38-043	BOS	@	EDM	W	0	1	1	1	2	1	0	30	33:53:00	16:57
52	1999-02-12	52	38-046	BOS	@	CGY	L	0	3	3	0	0	4	0	26	32:05:00	16:03
55	1999-02-21	55	38-055	BOS	@	CHI	W	1	1	2	1	0	3	33.3	28	36:24:00	18:12
62	1999-03-07	62	38-069	BOS		NYR	L	0	0	0	-2	0	9	0	25	35:14:00	17:37
63	1999-03-09	63	38-071	BOS		FLA	W	1	1	2	1	0	6	16.7	26	33:01:00	16:31
71	1999-03-27	71	38-089	BOS	@	TOR	T	1	0	1	-1	0	4	25	33	35:55:00	17:58
78	1999-04-10	78	38-103	BOS		TBL	W	0	1	1	0	0	3	0	21	34:40:00	17:20

My best guess at the rule they used in this case is:

If the game was 60 minutes long, cut TOI values of 32 or greater in half.
If the game went into OT, cut TOI values of 35 or greater in half.

This box score has Bourque playing 19:40.

Washington Capitals vs. Boston Bruins Box Score: November 21, 1998 | Hockey-Reference.com

Get the latest box score updates on the Washington Capitals vs. Boston Bruins game played on November 21, 1998 on Hockey-Reference.com

www.hockey-reference.com

Back in 2014 it showed Bourque playing 39:20, twice as much. Both showed him with 30 shifts, leading Boston.

Washington Capitals at Boston Bruins Box Score, November 21, 1998 | Hockey-Reference.com

web.archive.org

 

[overpass]

From 1997-98 through 2007-08, defencemen were capped at 35:00 for a regular season game. Presumably any higher values were cut in half, as with the forwards over 30 TOI for a game. You can see there are many regular season TOI totals approaching or equalling 35, but none exceeding 35.

NHL Stats

The official source for NHL Stats including skaters, goalies, teams stats and more.

www.nhl.com

Then from 2008-09 to the present, there have been 21 regular season games where a defenceman exceeded 35 TOI, topping out at 38:54 by Cam Fowler at the end of the 22-23 season. So it looks like they stopped editing those high TOI totals in more recent seasons.

 

[BraveCanadian]

So strange to cut ice time practically in half, I can’t see the reasoning.

 

[ContrarianGoaltender]

I've sent feedback to hockey-reference to see if they're interested in fixing this issue. If anyone has a contact at nhl.com, would you mind bringing this issue to their attention?

Interesting find. I used to work with the NHL's Sr. Manager of Hockey Analytics, I can send him a message to see if he knows anything.

 

[morehockeystats]

I mean, TOI of all skaters and goalies should add up to a rather fixed number...

 

[PenguinSuitedUp]

So strange to cut ice time practically in half, I can’t see the reasoning.

It makes the player look better and hides the fact that NHL coaches knew so little about health safety and overused some guys.

 

[overpass]

I mean, TOI of all skaters and goalies should add up to a rather fixed number...

Indeed.

I said that the NHL stats team may have applied a data validation process to deal with outlier numbers. But they should have enough confidence in their numbers to at least check to see if the player TOI adds up before making a major change.

Cutting high totals in half is a step you might take if you had very little confidence in the data collection and no way to check it. For example, if ice time was recorded by having each player write their estimated minutes played on a paper after the game, and then by scanning those papers.

 

[BraveCanadian]

It makes the player look better and hides the fact that NHL coaches knew so little about health safety and overused some guys.

What? Playing 30 minutes over a few hours span with breaks isn't putting anyone's safety at risk.

This idea that we only invented fire after the 2000s really needs to die.

 

[PenguinSuitedUp]

What? Playing 30 minutes over a few hours span with breaks isn't putting anyone's safety at risk.

This idea that we only invented fire after the 2000s really needs to die.

At that level, it’s been scientifically proven that player performance typically drops because the body gets tired. When your body gets tired, you’re more prone to injury. Very simple concept, scientifically backed.

 

[BraveCanadian]

At that level, it’s been scientifically proven that player performance typically drops because the body gets tired. When your body gets tired, you’re more prone to injury. Very simple concept, scientifically backed.

People get tired? I'm glad we had scientists figure this out.. c'mon..

 

[PenguinSuitedUp]

People get tired? I'm glad we had scientists figure this out.. c'mon..

So you agree that we don’t need a scientist to tell us it’s not good to overplay athletes? Cool.

 

[morehockeystats]

Indeed.

I said that the NHL stats team may have applied a data validation process to deal with outlier numbers. But they should have enough confidence in their numbers to at least check to see if the player TOI adds up before making a major change.

Cutting high totals in half is a step you might take if you had very little confidence in the data collection and no way to check it. For example, if ice time was recorded by having each player write their estimated minutes played on a paper after the game, and then by scanning those papers.

I noticed NHL does very little data validation. Especially at the easiest stage - the input.

 

[Mubiki]

So you agree that we don’t need a scientist to tell us it’s not good to overplay athletes? Cool.

You know what's even safer? Not playing hockey at all. Why should anyone risk any sort of bodily injury just for the sake of entertainment? In fact, why even leave the house? It's dangerous outside bro.

You choose your tiny hill for the sole purpose of being annoying. Knock it off.

 

[Gorskyontario]

Thanks for the hard work in this thread.

 

[Regal]

I knew this was the case. I remember a Bure discussion of that season on here last year and brought up he had high ice time and was told it wasn’t that outrageous relative to the league, citing these numbers, but I knew it used to be over 26. Nice to see the work on figuring this out

 

[Jumptheshark]

Edit: I found that all games where a forward recorded over 30:00 TOI had that TOI edited down, usually cut in half. This change was made on both nhl.com and hockey-reference sometime in 2023.

As a result, any forward who legitimately played more than 30 minutes in a regular season game from 1997-98 to 2022-23 now has incorrect TOI stats. Each game of 30+ TOI that was edited subtracted at least 15 minutes from the player's TOI.

I'm pretty sure the ice time stats for Pavel Bure's 2000-01 have changed, on both nhl.com and hockey-reference, at some point in the past few years. And I think the current stats are wrong.

Has anyone else noticed this issue with any players?

Recently I saw that nhl.com and hockey-reference both have Bure with an average time on ice of 23:14 for the 2000-01 season. But I swear I remember his average TOI for that season was displayed as 26:52 in the past. Here's a link to an old HF post that gives the same number, so I'm not just imagining things.

Here's Bure's 2000-01 game log from hockey-reference. Just looking through here, I see his median TOI was 24:32, well above the average of 23:14. Here's a breakdown of listed his TOI/game, so you can see how it's a skewed distribution.

29-30 TOI: 3
28-29 TOI: 7
27-28 TOI: 6
26-27 TOI: 10
25-26 TOI: 12
24-25 TOI: 6
23-24 TOI: 8
22-23 TOI: 5
21-22 TOI: 2
20-21 TOI: 2
19-20 TOI: 1
18-19 TOI: 5
17-18 TOI: 1
16-17 TOI: 6
15-16 TOI: 8

The part in green looks like a normal distribution centred around 25 or 26 TOI per game, tapering off around 30 TOI at the top and 20-22 TOI on the bottom. But then there's a weird second distribution where Bure had a bunch of games with under 20 TOI, including 14 with 15-17 TOI.

Let's look at a couple of these games where Bure played under 20 minutes.

March 3, 2001, 2-2 vs ATL

• The game was 65 minutes long, so TOI for skaters would be 325 if there were 5 skaters on at all times. Florida's skaters only sum to 288 TOI.
• Florida took a few penalties and played a fair bit of 4-on-5 so their skater TOI should be about 310. 22 minutes missing.
• Bure is listed with 18:09 in ice time, 8th on the team, behind D-men Svehla, Hedican, Simpson, Barrie and forwards Kozlov, Nilson, Sillinger.
• Bure was 3rd in shifts with 35, behind only Svehla and Hedican. If he really played only 18:09, he averaged 31 seconds per shift
• Bure took 13 shots and scored 0 goals.

Feb 14, 2001, 4-2 vs PHX

• Game was 60 minutes long. Florida took 4 minors and allowed 1 PPG, so team TOI should sum to about 293
• Team TOI sums to only 271.
• Bure is listed with 16:35 ice time, 7th among FLA skaters and 4th among FLA forwards (behind Kozlov, Nilson, Niedermayer)
• Bure is listed with 29 shifts, #2 on FLA behind Svehla's 30, and #1 among FLA forwards. If he played 16:35, his shifts averaged 34 seconds.
• Bure took 3 shots and scored 3 goals.

There's over 20 minutes of missing TOI for FLA in both games. And Bure's TOI is weirdly low, even though he played a lot of shifts. I'm pretty sure Bure's TOI is not correct for either of these games, or for many others.

I would say probably any game where Bure had an average TOI/shift of under 45 seconds is wrong.


What about Bure's 1999-00 season? There are a few weird TOI totals in there as well, where he played a lot of shifts but not a lot of TOI. Including one game where he is listed with 29 shifts and 0 TOI. Overall I think the 1999-00 numbers have fewer errors than the 2000-01 numbers, but I think Bure's listed TOI is low for that season too.

Do you guys agree that something is wrong here? Any thoughts on how/why it happened, and have you noticed anything similar for other players?

Becarefull of any TOI stat prior to 2002. That is when it was more widely accurate