|
Why are we so interested in statistics? When our favorite team is considering acquiring a player, fans like us get the urge to look at that players statistics from previous seasons, even though our team is going to get no benefit from those past goals. The answer is simple. What we really care about are his future contributions. His past statistics are only useful in helping us judge what his future will hold.
Let’s go ahead and do just that. Here are the statistics of a real player. Based on this information, how many points will this player score for our favorite team next season? For argument’s sake, scribble something down.
GP G A PTS PIM
Mystery Player 82 35 32 67 86
If you guessed fewer than 94 points, you got it wrong. This is Jarome Iginla of the Calgary Flames two seasons ago. Don’t feel badly because this just as easily could have been any of the following players, none of whom could have been expected to score that many points.
GP G A PTS PIM
Miroslav Satan 2005-06 82 35 31 66 54
Mike Knuble 2005-06 82 34 31 65 80
Bobby Schmautz 1973-74 76 33 32 65 89
Bill Goldsworthy 1970-71 77 34 31 65 85
Miroslav Satan 2006-07 81 27 32 59 46
Mike Knuble 2006-07 64 24 30 54 56
Bobby Schmautz 1974-75 56 21 30 51 63
Bill Goldsworthy 1971-72 78 31 31 62 59
It is quite understandable that you would be blowing the whistle right now! Jarome Iginla was having an off season. Everyone knows he’s not a 65-point man like Satan or Knuble. We need more information if we want to make predictions. If I had included the previous season’s information, you would have known if 67 points was a typical season or an outlier season. Let’s look at another mystery player, and this time I will give you that extra information about his previous season:
GP G A PTS PIM GP G A PTS PIM
Mystery Player #2 77 31 43 74 38 Before: 75 36 41 77 38
Does the extra information help us make a more accurate prediction of how many points he will score next season for our favorite team? Judging from his previous season, this player clearly was not having a particularly strong or weak season. Take a guess at what his next season will look like. Scribble down another number.
If you guessed anything less than 54 goals and 123 points you were wrong again. Mystery Player #2 is Jaromir Jagr of the New York Rangers during the 2003-04 season. Again, you really should not feel badly because your prediction would have been quite accurate had I actually been thinking of one of the following players instead.
GP G A PTS PIM GP G A PTS PIM
Jean Ratelle 1969-70 75 32 42 74 28 Before 75 32 46 78 26
Ivan Boldirev 1978-79 79 35 43 78 31 Before 80 35 45 80 34
Dave Poulin 1984-85 73 30 44 74 59 Before 73 31 45 76 47
Brian Bellows 1991-92 80 30 45 75 41 Before 80 35 40 75 43
That was probably frustrating because the league changed the rules in between seasons, opening up the ice. However, even if it had not, we still know that Jagr is in a different league than Poulin and Bellows. What additional information would make it easier for us to feel more confident with our predictions? Looking at the entire career history of a player would help us. With this information, it would be easier to determine if we were dealing with a young player on his way into his prime, or an established veteran playing out his twilight seasons.
GP G A PTS PIM
Jaromir Jagr 950 506 729 1235 212
Jean Ratelle 408 122 173 295 90
Ivan Boldirev 522 163 239 402 301
Dave Poulin 75 33 45 78 49
Brian Bellows 673 312 335 647 496
It would have helped to have studied this table in advance of our hastily scribbled guesses. With this information I’m sure it would have been much easier to predict that one of these players was more than your typical 75 point forward. The main reason we study statistics is to help us understand the past so that we can predict the future. We feel most confident when basing those predictions on a larger sample. Armed with current statistics, previous season’s statistics and career totals, we can predict the next season and the rest of the player’s career.
Stock market analysts do the same thing all the time. They examine various company statistics, and search their vast databases for similar companies in the past, and then use the future performance of those similar companies to help establish the future expectations for that company. There is no reason for not being able to do the same thing with hockey players. Here at Puck Prospectus, we're going to be doing this a lot, much in the vein of PECOTA, KUBIAK, or SCHOENE.
By applying a formula that creates a Similarity Score between players, we should be able to find the paths that all but the most unique players have traveled before, and then see where those paths led. They should help set reasonable expectations, along with both the best and worst case scenarios.I have chosen Jason Spezza of the Ottawa Senators for an example. First, we take a look at the readily available statistics we have discussed:
Jason Spezza GP G A PTS PIM
2007-08 76 34 58 92 66
2006-07 67 34 53 87 45
All Preceding 246 82 171 253 157
Then, we create a Similarity Score formula derived from these simple statistics, including points per game, penalty minutes per game, and playmaker ratio (assists per goal). Finally, we apply this formula to every player for every season throughout the NHL’s entire 90 year history to find the 10 most comparable players. Here is how they performed the following season, and the rest of their careers.
GP G A PTS PIM GP G A PTS PIM
Doug Gilmour 1988-89 72 26 59 85 44 1028 290 740 1030 1001
Jaromir Jagr 1994-95 48 32 38 70 37 962 521 759 1280 709
Andy Bathgate 1959-60 70 26 48 74 28 686 211 408 619 354
Andy Bathgate 1963-64 71 19 58 77 34 406 93 210 303 206
Bernie Federko 1982-83 75 24 60 84 24 592 208 473 681 290
Neal Broten 1987-88 54 9 30 39 32 662 123 320 443 310
Rene Robert 1976-77 80 33 40 73 46 381 135 205 340 271
Sergei Makarov 1992-93 71 18 39 57 40 194 58 91 149 158
Bobby Smith 1981-82 80 43 71 114 82 858 271 515 786 781
Sergei Fedorov 1993-94 82 56 64 120 34 966 375 519 894 579
Average 70 29 51 79 40 674 229 424 653 466
It’s that simple, assuming you have a basic understanding of stats, a database of statistics, and modern analysis tools! If our favorite team is considering trading for Jason Spezza, we can easily get an idea of what we can expect from him in the future using only basic statistics.
I will leave you with two questions to ponder until next time. First, if we went beyond the simple statistics and looked at shooting percentage, plus/minus, average time on the ice or power play point totals, would that help us make our predictions? Remember that these statistics weren’t always available, so we would be forfeiting several decades of NHL History. For instance, plus/minus and shots were not officially recorded until the 1967-68 season.
Secondly, notice Sergei Makarov on that list. He was 34 years old going into the 1992-93 season, hardly an appropriate match for the 25 year-old Spezza. What happened? Makarov didn’t join the NHL until he was 31 years old. But what if we could translate his history in the Russian league with CSKA Moscow? Stay tuned!
This article was originally published on March 2, 2009 at Puck Prospectus. During the off-season, we're working on new projects and also want to give some of you a chance to see this important work for the first time.
Robert Vollman is an author of Hockey Prospectus.
You can contact Robert by clicking here or click here to see Robert's other articles.
| 
0 comments have been left for this article. 
Conn Smythe Watch: Lundqvist Leads