Melvin's 2021-22 NHL Team Projections

[Melvin]

This is the one I read (and mostly don't understand) - https://fivethirtyeight.com/features/what-our-new-nhl-forecast-says-about-this-season/

Thanks.

That is actually very disappointing. As you say, they are just looking at it at the team level and not considering the underlying rosters. So team's are projected based on their play last year irrespective of any roster changes that occurred in the off-season. For Seattle, they are making a "guess" at an ELO based on the play of previous expansion teams. That's disappointing and, if I'm being blunt, honestly pretty lazy of them.

My post above could be therefore read as which teams I'm projecting to do better/worse than last year pretty much.

 

[Melvin]

Okay.

I said I would post a final update before the season kicks off, so here we go. Rosters were finalized last night, and I've updated to include all the waiver claims and other deck chairs and so forth.

As I've said before, I did not build a mechanism for dealing with partial-season injuries, so players are either in or out. My method was basically if they are injured but expected to be back by December or so, I included them. If they are expected to miss more time than that (e.g. Colin White in Ottawa) then I took them out. In some cases I am just guessing. I took Brandon Sutter out; it's just a hunch but I don't expect him to play this year. I dunno, your guess is as good as mine. . Boeser and Motte on the other hand are in. I am sure I made mistakes but it's not possible to do this perfectly and also following the training camps of 32 teams is exhausting.

First up, I will post the same goal differential rankings as I did in the OP, minus the little anecdotes. I am also providing other info on where the rank in the specific categories. "NonGoalieRank" is where the team ranks just on the strength of offense-defense, not taking goaltending into account. Last column notes the change from the original ranking. There are a couple big changes, I'll get to that:

Rank	TEAM	Goal Diff	ES Offense	ES Defense	PP	PK	Goaltending	NonGoalieRank	CHANGE FROM OP
1	VGK	36	1	25	14	1	4	2	-
2	COL	36	6	2	1	8	18	1	-
3	TOR	32	2	19	15	7	2	3	-
4	SEA	21	31	5	4	17	3	8	-
5	T.B	14	15	6	13	16	13	6	+1
6	NYI	13	21	4	11	24	7	9	+2
7	BOS	13	32	1	3	30	14	7	+4
8	CAR	13	3	18	8	22	19	4	-3
9	FLA	7	22	3	2	31	24	5	+4
10	WPG	6	8	32	25	2	1	24	+7
11	NYR	2	19	16	9	28	6	21	+1
12	NSH	2	14	15	26	12	11	18	-3
13	DAL	2	10	10	23	15	20	11	-3
14	EDM	1	4	30	20	14	17	14	+1
15	CHI	-1	11	31	21	3	9	22	+9
16	PIT	-1	23	9	7	21	27	10	-9
17	N.J	-2	12	12	10	25	21	13	-1
18	WSH	-4	7	24	28	5	22	16	+5
19	L.A	-4	30	21	5	27	10	28	+7
20	VAN	-4	29	20	6	29	8	29	+2
21	S.J	-5	16	17	12	32	12	26	-1
22	MTL	-6	13	8	22	26	23	19	-3
23	STL	-6	27	14	17	18	16	23	-2
24	CGY	-7	17	11	19	23	28	15	-10
25	OTT	-7	9	27	24	4	25	17	-
26	DET	-10	20	29	32	9	5	31	+3
27	MIN	-13	18	7	27	11	31	12	-9
28	BUF	-15	28	23	18	13	26	27	+4
29	ARI	-19	26	13	16	19	30	25	+2
30	PHI	-20	5	22	30	6	32	20	-3
31	CBJ	-21	25	26	29	10	29	30	-3
32	ANA	-22	24	28	31	20	15	32	-2

[TBODY] [/TBODY]

There are basically 4 different things that causes change in rankings:

1) There were a few players that I had flat out missing from my database, and some were some pretty big names. For example, I was completely missing David Pastrnak AND Brad Marchand which is part of the reason Boston was in 11th. I have no idea how I was missing those players and am pretty annoyed about it, but I believe I have all significant players now.

2) Actual roster changes, obviously. Injuries, trades, etc. Also some tweaks to usage, when trying to guess what % of starts each goalie is going to get. This is again basically an educated guess on my part. I still don't know what Dallas is doing.

3) Bugs in code. There were a few just flat-out bugs I had where I fat-fingered something or wasn't handling things properly, leading to dumb results like Arizona having the best PK of all time, or weighting 2019 heavier than I meant to, etc.

4) Improvements to methodology. Maybe I should put "improvements" in scare quotes, but basically I've spent a lot of time back-testing with previous seasons worth of data to find the right coefficients and such, and I finally implemented a few things that I didn't have time for before (properly allocating ice time between forwards and defensemen being the biggest one.)

I am still finding basically no correlation in year-to-year goaltender performance so I take the goaltending number with a huge grain of salt, which is why I provide the NonGoalieRank column.

OK, at last, I have run the simulations. To explain these once more, I have run the season 10,000 times taking the schedule into account, the goal-scoring and allowing rates of each team and giving a small home ice advantage. The "Point Range" represents the 25th-to-75th percentile result for each team, in other words the point range in which the majority of simulations land. I have also provided the floor and the ceiling, along with a bunch of other data points.

The last few columns represents the % of time the team achieves a certain result, so in order, % of time they finished 1st in division, then 2nd in division, and so forth, culminating in the % of the simulations in which they made the playoffs.


Atlantic Division

Team	MinPts	Point Range	MaxPts	AvgGF	MinGF	MaxGF	AvgGA	MinGA	MaxGA	D1%	D2%	D3%	WC1%	WC2%	Playoff%
TOR	77	106 - 117	140	304	228	370	227	167	288	75%	17%	5%	2%	0%	100%
T.B	64	93 - 104	129	276	213	342	244	188	306	10%	31%	28%	13%	6%	89%
BOS	66	92 - 104	126	285	212	352	253	190	310	10%	31%	27%	13%	7%	88%
FLA	59	87 - 99	127	300	232	375	284	225	354	3%	15%	23%	19%	12%	73%
MTL	51	77 - 89	126	264	206	327	282	219	348	0%	2%	6%	8%	11%	27%
OTT	49	76 - 88	114	264	209	329	286	231	346	0%	2%	5%	7%	9%	23%
DET	46	75 - 86	112	236	176	303	262	206	323	0%	1%	4%	5%	8%	19%
BUF	47	70 - 82	104	249	189	306	290	232	351	0%	0%	2%	2%	4%	8%

[TBODY] [/TBODY]

Metro Division

Team	MinPts	Point Range	MaxPts	AvgGF	MinGF	MaxGF	AvgGA	MinGA	MaxGA	D1%	D2%	D3%	WC1%	WC2%	Playoff%
NYI	64	93 - 104	129	274	215	336	242	180	303	40%	27%	16%	3%	4%	90%
CAR	64	92 - 103	129	317	248	383	284	216	354	36%	28%	17%	4%	4%	88%
NYR	54	84 - 95	122	277	215	343	274	212	334	10%	15%	19%	7%	8%	59%
PIT	55	82 - 93	116	279	217	343	283	226	350	6%	11%	17%	6%	9%	48%
N.J	53	81 - 92	120	280	223	343	286	227	362	5%	10%	15%	6%	9%	44%
WSH	51	79 - 91	119	260	205	328	272	196	341	4%	8%	13%	5%	7%	37%
PHI	42	67 - 78	104	254	194	326	307	243	382	0%	1%	2%	1%	2%	4%
CBJ	42	66 - 77	105	239	181	301	294	226	354	0%	0%	1%	1%	1%	3%

[TBODY] [/TBODY]

Central Division

Team	MinPts	Point Range	MaxPts	AvgGF	MinGF	MaxGF	AvgGA	MinGA	MaxGA	D1%	D2%	D3%	WC1%	WC2%	Playoff%
COL	80	108 - 119	144	317	248	386	230	170	289	92%	6%	1%	0%	0%	100%
WPG	59	87 - 98	121	262	205	328	249	195	311	3%	31%	23%	10%	8%	75%
DAL	60	84 - 96	124	260	203	326	257	192	321	2%	21%	21%	11%	10%	63%
NSH	58	84 - 95	121	255	199	322	252	204	314	2%	20%	21%	10%	10%	62%
CHI	50	81 - 93	120	262	207	328	268	207	327	1%	13%	17%	9%	10%	49%
STL	51	78 - 90	123	256	195	332	272	216	334	0%	7%	11%	8%	9%	34%
MIN	45	72 - 84	110	246	179	320	280	212	342	0%	2%	4%	3%	5%	15%
ARI	40	67 - 79	104	253	195	313	304	235	379	0%	1%	2%	2%	2%	6%

[TBODY] [/TBODY]

Pacific Division

Team	MinPts	Point Range	MaxPts	AvgGF	MinGF	MaxGF	AvgGA	MinGA	MaxGA	D1%	D2%	D3%	WC1%	WC2%	Playoff%
VGK	84	108 - 119	139	320	261	387	233	183	295	80%	18%	2%	0%	0%	100%
SEA	73	98 - 109	132	273	204	330	226	170	287	18%	61%	14%	3%	1%	97%
EDM	58	82 - 94	123	289	229	352	290	235	353	1%	8%	26%	11%	9%	54%
L.A	53	79 - 90	118	268	214	324	281	222	359	0%	4%	17%	9%	9%	39%
VAN	50	78 - 90	119	266	206	338	280	219	355	0%	4%	15%	9%	9%	37%
S.J	47	78 - 89	113	273	211	336	290	226	353	0%	3%	14%	8%	9%	34%
CGY	51	77 - 88	116	267	210	329	287	228	352	0%	3%	12%	7%	8%	30%
ANA	44	65 - 77	102	236	159	304	293	232	357	0%	0%	1%	1%	2%	4%

[TBODY] [/TBODY]

So, for Vancouver:

* They finish with between 78-90 points the majority of the time, worst sim was 50 points and best was 119 points
* They score on average 266 goals, allow on average 280 goals.
* They win the division less than 1% of the time, and come in 2nd 4% of the time, 3rd 15%, win the 1st WC 9% of the time and win the 2nd WC 9% of the time, all adding up to making the playoffs about 37% of the time.

Again this is somewhat in line with what I've seen from other projection systems, which have Vancouver making the playoffs 30-40% of the time in simulations.

If you see anything that seems off (except for Seattle being too high ,) let me know! Otherwise, basically going to put this to bed and re-visit it in 6 months.

 

[J Corso]

Looks great. Except for Seattle being too low.

 

[Melvin]

Looks great. Except for Seattle being too low.

I'm gonna have to deal with so much trolling if Seattle shits the bed, god damn it. And now half the team has COVID.

 

[AppleHoneySauce]

I'm gonna have to deal with so much trolling if Seattle shits the bed, god damn it. And now half the team has COVID.

on the other hand you get to make fun of them if Seattle makes it into the playoffs at all seeing as most are acting like they are bottom 10.

 

[J Corso]

The numbers look solid and the logic is reasonable. Only thing that might sink things is if the coach goes nuts or massive injuries which you can't control for.

 

[Goatlender]

This is a fun bump.

It's amusing to see some people who scoffed at Melvin's predictions (despite them being based on an algorithm and not personal bias), especially for ranking the Canucks at 22nd overall. Well, here we are over halfway through the season, and the Canucks are *checks notes* tied for 20th overall with Detroit. Lol.

 

[FilHronek Orchestra]

This is a fun bump.

It's amusing to see some people who scoffed at Melvin's predictions (despite them being based on an algorithm and not personal bias), especially for ranking the Canucks at 22nd overall. Well, here we are over halfway through the season, and the Canucks are *checks notes* tied for 20th overall with Detroit. Lol.

Seattle has let Melvin down.

 

[Nona Di Giuseppe]

Seattle has let Melvin down.

In b4 “but if they had goal tending”

 

[Melvin]

In b4 “but if they had goal tending”

Well, yeah. I mean, it's the truth. I don't know what else I can say.

I had them with the 3rd best goaltending in the NHL and instead they are 32nd. The other metrics have been pretty close. Below average offense, good 5v5 play and bad PP.

I will do a deeper dive on all 32 teams when the season concludes but the other big misses were Calgary and Minnesota. Minnesota is goaltending again, had them with 30th place goaltending and instead it's been pretty good. Calgary ... I don't know. Darryl Sutter is a wizard, man.

I do enjoy that for all the outrage about where I had Vancouver, they are basically right where I had them.

 

[Peen]

i mean, the seattle one is obvious

it's like 20-21 holtby except every game

flames made a coaching change so the model he camp with was probably just pulling from the old data under the prev coach

but, uh, why so off on minnesota?

edit: okay melvin responded 0.5 seconds before i posted that

 

[AppleHoneySauce]

Well, yeah. I mean, it's the truth. I don't know what else I can say.

I had them with the 3rd best goaltending in the NHL and instead they are 32nd. The other metrics have been pretty close. Below average offense, good 5v5 play and bad PP.

I will do a deeper dive on all 32 teams when the season concludes but the other big misses were Calgary and Minnesota. Minnesota is goaltending again, had them with 30th place goaltending and instead it's been pretty good. Calgary ... I don't know. Darryl Sutter is a wizard, man.

I do enjoy that for all the outrage about where I had Vancouver, they are basically right where I had them.

I mean i assume Marky and Sutter implementing better PP and PK probably ups them. I think your still pretty close and its pretty clear what happened to make them better.

 

[Melvin]

i mean, the seattle one is obvious

it's like 20-21 holtby except every game

flames made a coaching change so the model he camp with was probably just pulling from the old data under the prev coach

but, uh, why so off on minnesota?

edit: okay melvin responded 0.5 seconds before i posted that

I think

1) The goalie stats *hated* Kahkkonen, and didn't love Talbot either.
2) The model really had no way of knowing how good Kaprizov is, with just 55 games of data to work with.

These were the two weakest spots of the model that I knew about before the season even started. Rookies and goalies. Similar situation with Anaheim where Terry/Zegras were much better than the model could have ever foreseen.

 

[M2Beezy]

This is a fun bump.

It's amusing to see some people who scoffed at Melvin's predictions (despite them being based on an algorithm and not personal bias), especially for ranking the Canucks at 22nd overall. Well, here we are over halfway through the season, and the Canucks are *checks notes* tied for 20th overall with Detroit. Lol.

Actually we are 22nd in point percentage

But that doesnt change the fact that Melvin hates the Canucks

 

[Goatlender]

Actually we are 22nd in point percentage

but Green

 

[M2Beezy]

but Green

 

[Vector]

I think

1) The goalie stats *hated* Kahkkonen, and didn't love Talbot either.
2) The model really had no way of knowing how good Kaprizov is, with just 55 games of data to work with.

These were the two weakest spots of the model that I knew about before the season even started. Rookies and goalies. Similar situation with Anaheim where Terry/Zegras were much better than the model could have ever foreseen.

How do you even adjust for goalies being spectres that float between realms? Obviously some goalies are consistent while others (stares long and hard at Marc-Andre Fleury) are not.

 

[arttk]

I think

1) The goalie stats *hated* Kahkkonen, and didn't love Talbot either.
2) The model really had no way of knowing how good Kaprizov is, with just 55 games of data to work with.

These were the two weakest spots of the model that I knew about before the season even started. Rookies and goalies. Similar situation with Anaheim where Terry/Zegras were much better than the model could have ever foreseen.

At some point the Canucks are going to give you a job

 

[Just A Bit Outside]

At some point the Canucks are going to give you a job

Hopefully it’s blow or hand.

 

[Canadian Canuck]

Love you Melvin

 

[arttk]

Hopefully it’s blow or hand.

Want to make a joke about Swedish twins..

 

[FilHronek Orchestra]

but heinze (the other) 57 (games) !!!

 

[PuckMunchkin]

Looks pretty damn close so far.

A couple of surprises notwithstanding; Biggest ones being the Islander's freefall, Seattle's goaltending woes and Florida becoming, well Florida.

 

[Reverend Mayhem]

I do enjoy that for all the outrage about where I had Vancouver, they are basically right where I had them.

This happens literally every year and I'm so sick of it.

 

[StreetHawk]

Looks pretty damn close so far.

A couple of surprises notwithstanding; Biggest ones being the Islander's freefall, Seattle's goaltending woes and Florida becoming, well Florida.

Unlike the NFL, because of fewer playoff teams, most of the 16 NHL playoff spots are kind of expected. In the NFL around half of the prior year's playoff teams miss.

Of the 8 playoff teams in the East, only the Isles are the one that is a surprise to me to not be there. But, considering they started with a 13 game road trip due to their arena and they were among the first teams with a Covid outbreak, not shocked that they are in the playoffs, but didn't expect them to be that far out. I thought one of NYR/Phi would take a PO spot if one of the top 8 stumbled. Once a young goalie like Hart struggles, you kind of wonder if they can get it back the next season.

In the West, expected Col, STL, and it was a coin flip between 2 of Min/Dal/Nas to make it in the Central. Pacific, Vegas, Edm, Cal, and any of Van/SJ/LA or 5th place in the Central. I didn't foresee Anaheim's hot start/Terry or Seattle (just the odds of an expansion team again competing for the spot).