Projecting the Leafs Top Six: A Different Approach


After reading Robert Vollman’s post at The Leafs Nation about projected Leafs Scoring I thought that it was finally time to revisit my attempts at forecasting what the Leafs could reasonably expect for scoring in the upcoming season (assuming of course there is one.)

While I don’t think there is anything wrong with the current metrics that are currently out there, I am interested in seeing how my own attempt holds up and would like to apply my own methodology to the projection game. Where Snepsts67 and VUKOTA both rely on comparables, my own attempt (for consistency I’ll call it MIRONOV) looks at each players recent history in conjunction with trends associated with players aging and overall offensive output changes.

Like Snepsts67 I chose to go with a range. The high value is based off the best season the player has had in the past three seasons, the low value is based on the worst season in the past three seasons, and the average is the average over the past three seasons. If the younger players do not have three seasons with 20 more or games played in the NHL than CHL, AHL, or NCAA equivalencies from Behind the Net were used.

The next piece that is incorporated is the effect of aging, experience, etc. It’s an attempt to quantify if whether a player should be trending upwards, plateau, or be on the decline, and to what extent.

So I took the top 180 scoring forwards for each season since the lockout (180 players divided by 30 teams gives you the average top six forwards) and looked at sorted them by age to determine the total number of games played by each age of player and the average goals or points per game for each age.

Age
GP Total
Average PPG
18
464
0.765
19
1143
0.751333333
20
2793
0.72027027
21
3662
0.750851064
22
5161
0.73641791
23
7160
0.708645833
24
7061
0.732
25
7853
0.736407767
26
7070
0.742795699
27
7758
0.739417476
28
6659
0.732873563
29
6114
0.754390244
30
5257
0.746666667
31
5377
0.717183099
32
4730
0.727580645
33
4314
0.756034483
34
3368
0.779555556
35
3159
0.714761905
36
2178
0.738275862
37
1640
0.678181818
38
1042
0.688
39
630
0.754444444
40
291
0.7475
41
163
0.665
42
81
0.59

From there I took the age with the highest number of games played (Age 25 with 7853 games played) and divided each of the other age totals from the max number to create a ratio of how often this age group will fill a scoring role on the team.

The next step was to multiply that age ratio by the points per game. From there it’s taking the higher aged number and dividing it by the previous age value (Age 19 value divided by Age 18 value, Age 20 value divided by Age 19 value and so on.) This finally gives you the value that you can multiply against the players point per game values.

The last step is looking at the change in scoring around the league. Since the year after the lockout the number of points per season from the top six has been on a modest decline. While not a huge impact the players point per game value and the age value and multiplied by the change in scoring value which is 98.6%.
All of the values are listed as point per game, but for the purpose of showing predicted seasons I have multiplied out what the totals would be over 82 games.

Points over 82 games
Points Per Game
High
Low
Avg
High
Low
Avg
Kessel
90
70
77
1.10
0.86
0.94
Lupul
78
43
56
0.95
0.53
0.68
Grabovski
55
45
51
0.67
0.55
0.62
Kulemin
62
35
46
0.75
0.43
0.56
MacArthur
52
29
41
0.64
0.36
0.50
Bozak
53
29
43
0.65
0.35
0.52
Connolly
70
40
53
0.86
0.49
0.65
van Riemsdyk
60
48
55
0.73
0.59
0.67
Frattin
50
24
35
0.62
0.30
0.42
Kadri
54
37
45
0.65
0.45
0.55

Goals over 82 Games
Goals per Game
High
Low
Avg
High
Low
Avg
Kessel
45
36
41
0.55
0.44
0.50
Lupul
33
20
27
0.40
0.24
0.33
Grabovski
27
13
21
0.33
0.16
0.26
Kulemin
32
9
19
0.38
0.10
0.24
MacArthur
18
13
16
0.22
0.16
0.20
Bozak
18
13
16
0.22
0.16
0.19
Connolly
18
15
16
0.22
0.18
0.20
van Riemsdyk
32
22
28
0.39
0.26
0.34
Frattin
30
13
20
0.37
0.16
0.24
Kadri
26
11
19
0.32
0.13
0.23
(what I consider the most likely result in bold.)

I’m sure I’m overstating Tim Connolly’s numbers to some extent, but in a healthy season with him playing in the top six it may be a plausible range. The fact is that given that the numbers are based off of top six forwards it seems that any of these forwards who do not receive top six ice time will not likely make their average numbers, and possibly not their low numbers either. 

Below is a comparison to the other predictions for some of the Leafs:

VUKOTA
Snepsts67 Average
MIRONOV Average

Player
G
Pt
G
Pt
G
Pt

Kessel
33.5
72.5
30.4
63.2
40.6
77.2

Lupul
19.6
50.1
20.1
54.6
27.3
55.9

Grabovski
21.4
47.5
25.5
53.8
21.3
50.8

MacArthur
18.3
44
18
49.3
16.4
40.7

van Riemsdyk
16.6
17.1
23
48
27.8
55.2

Bozak
14.7
37.2
12.6
36.6
15.8
42.9

Connolly
11.2
30.8
12.3
33.3
16.1
53.3

*Last Year Numbers include actual games played. VUKOTA numbers are based on predicted number of games played and Snepts67 and MIRONOV are based on 82 game schedules.

Certainly I seem to have inflated numbers for Connolly in comparison, but if he returns to a top six role I think these numbers are plausible. My Bozak, van Riemsdyk, and Kessel numbers are also well above the other metrics, which I can pretty much guarantee I’ll either be right on Bozak or right on Connolly, not both. Lupul and Grabovski are right in the same range, although if the VUKOTA metric was applied to 82 games for Lupul it would be significantly higher than Snepsts67 or MIRONOV. Only on Clarke MacArthur was there a significantly lower number than the other two metrics.

My approach is based on the following assumptions:  
  • The three most recent seasons in a player’s career provide a more baseline for projections than looking at their entire career.
  •  Players should generally see their career numbers improve until the age of 27 and decline after that with an accelerated decline by the age of 34.
  •   A steep downward slope is not accurately captured in this forecast as experienced with the Tim Connolly projections. Based on the historical decline/incline, role of the player (top six/powerplay) and total ice time it’s likely that a player will be closer to the high or low rather than the average.
  • Since the previous numbers from some these players like Kadri (OHL), Frattin (NCAA), and Connolly (Buffalo) reflect them getting top six ice time in their previous role their numbers will likely not be achievable in bottom six minutes.
I am aware that when you total the average number of goals for all ten predicted players you‘ll get 223 goals, which we would be happy to have as a team total. When you factor in the range of projections and injury impact it makes the numbers more plausible.

Now that I’ve floated my predictions I’m looking forward to seeing how out to lunch these may be.  Any feedback on my first attempt at this would be appreciated. 

No comments: