|
A question that has nagged at me for years in fantasy sports is one that is especially pertinent to fantasy hockey: What number of prior fantasy weeks’ performance gives the best indication of next week's performance?
This is a fundamental piece of knowledge for increasing your chances of success, but many fantasy enthusiasts never dial into the importance of this issue. Especially for new participants, it is so easy to be swayed by a very good or very bad performance last week in making a lineup decision on a player. As a consequence, many fantasy players end up switching players out of their lineup just in time to miss a big week, or conversely, they end up putting in a hot player just in time for the inevitable clunker. While there is always uncertainty in the outcome of any real or fantasy game no matter how much you plan, like any good baseball manager, poker player or stock investor, you should make the best of playing the percentages so as to improve your chances to win more frequently in the long run.
There are players that are inherently streaky, giving you hit and miss performances. Vincent Lecavalier has been this way for my fantasy hockey team last year and this year. These feast or famine players that give you the occasional monster week probably make you want to pull out your hair. I imagine that is how they make their real life coaches feel as well! While there’s something to be said for performance upside, there’s something to be said for performance that you can count on.
With this in mind, as the owner of your fantasy team, remember that you’re the coach and the general manager. Paraphrasing Bill Parcells, the fantasy owner gets “to cook the dinner” and to “shop for the groceries” as well. (As owner, you also get to pay the bill, so to speak.) If you’ve consciously – I hope consciously – picked one of these bipolar performers as the “general manager” of your team, as the “coach” of your team you need to be committed to playing said player through the frequent lows to enjoy the benefits of those big highs. If you won’t follow through on that commitment, don’t bother drafting the guy!
What number of prior fantasy weeks’ performance gives the best indication of next week's performance? Behind this question is the tension between having a large enough sample size of data to be truly indicative of a player's skills while covering a short enough period of time to capture hot and cold streaks, changes in a player's role, nagging injuries, and other factors. The sabermetrically inclined among us might gravitate towards the importance of a larger sample size, while mainstream fans may tend to think of athletes in all sports in terms of hot and cold streaks, as demonstrated by Exhibit A, angry callers on sports talk radio.
A typical fantasy hockey website has stats sortable by performance over the last 7 days, last 14 days, last 21 days, last 28 days and year to date. Before doing the statistical legwork for this column, my intuitive feel was generally to look at the last 21 days’ or the last 28 days’ performance as the sweet spot between having enough data while having a good enough indication of current playing level. This column provided an excellent opportunity to put this assumption to the test, for your benefit as well as my own.
There are an average of 3 games in an NHL week. Many of you fantasy players will instinctively know this off the top of your heads, from eyeballing your lineups every week. This column will use “3 games" interchangeably with "1 fantasy week”, while keeping in mind that each specific fantasy week can have a team playing 1, 2, 3 or 4 times. This week, we will look at the current top 35 Goal scorers in the NHL to get an indication of what number of prior fantasy weeks correlates best to future performance in Goals.
For each game played by those 35 players, predicted performances were calculated by taking the average Goals scored for the previous x games, with x = 3 * number of previous weeks. Using the difference between predicted performance and actual performance,we calculated a Coefficient of Variation for all players at each time point. Coefficient of Variation (or CV) is the standard deviation divided by the average, a normalized statistic, allowing the players to be compared on a level playing field. What you see in the graph above is the average of the CV’s for all top 35 Goal scorers, for various numbers of previous weeks of performance.
Not surprisingly, the Last 1 week's performance had by far the least correlation to next week's success in Goal scoring (77% CV). Though significantly more accurate than looking at the Last 1 week, predictions based on the last 2 weeks were also not as accurate as the other sorts (62% CV). Predictions based on the last 3 weeks, last 4 weeks and year yo date were essentially equal (58% CV, 56% CV, 57% CV) and may be the best sorts that your fantasy website will provide. Unfortunately, your fantasy league's stats are unlikely to allow you to sort by last 8 weeks (51% CV), last 9 weeks (50% CV) or last 10 weeks (51% CV).
Let’s look at a few specific examples of well known players to see how they varied from the average:
Coefficient of Variation
Number of previous weeks
Player 1 2 3 4 5 6 7 8 9 10 YTD
Crosby 115% 87% 86% 87% 82% 84% 70% 54% 42% 43% 80%
Malkin 82% 67% 58% 51% 52% 53% 51% 53% 49% 46% 56%
Ovechkin 63% 45% 46% 47% 44% 41% 37% 38% 38% 39% 45%
These three players fit the same sort of pattern as the average shown in the graph above, though some players vary to some extent. Without any additional knowledge, it is suggested that the Last 9 weeks performance is used as a predictor for future Goals scored for all players, with last 4 weeks or YTD used if last 9 weeks is not available.
Putting what we have now learned into practice, let’s look at players that may be undervalued or overvalued with respect to Goals, by comparing the performances of the Last 1-2 weeks against the much better measuring stick of the last 27 games, or 9 weeks. We will also look at what performance to expect out of some recently traded players of note.
Undervalued – Better performance in previous 27 games than in previous 3-6 games
Number of Goals in Average Goals/fantasy week based on
Previous x games
Player Team Pos 3 6 12 27 3 6 12 27
Mike Green WAS D 1 1 7 15 1.0 0.5 1.8 1.7
Adrian Aucoin CAL D 0 1 2 5 0.0 0.5 0.5 0.6
Fedor Tyutin CLB D 0 1 2 5 0.0 0.5 0.5 0.6
Mike Green undervalued? Repeat after me: “I will not drop Mike Green out of my lineup just because he didn’t score in the last game or two”. The poor guy just set a record for defensemen for consecutive games with a goal scored just a couple of weeks ago! Don’t out-think yourself and find a reason not play to him even if he has unfavorable matchups or a lesser number of games per week.
Adrian Aucoin and Fedor Tyutin may be free agents in your league if plus/minus is a major factor in scoring. They are what they are: Slightly above average offensive defensemen that are going to give you slightly below average plus/minus. The 0.6 G per fantasy week clip is what you should expect.
Overvalued – Better performance in previous 3-6 games than in previous 27 games
Number of Goals in Average Goals/fantasy week based on
Previous x games
Player Team Pos 3 6 12 27 3 6 12 27
Jonathan Toews CHI F 4 6 10 17 4.0 3.0 2.5 1.9
Ryan Kesler VAN F 1 3 8 11 1.0 1.5 2.0 1.2
Jason Arnott NSH F 5 5 8 13 5.0 2.5 2.0 1.4
T. Plekanec MON F 2 6 7 11 2.0 3.0 1.8 1.2
Anton Babchuk CAR D 3 4 6 6 3.0 2.0 1.5 0.7
Pavel Kubina TOR D 3 3 4 5 3.0 1.5 1.0 0.6
Sheldon Souray EDM D 2 2 5 6 2.0 1.0 1.3 0.7
Many leagues will already be past their trading deadline and may perhaps even be ready to head into their fantasy playoffs this week. If trades are still available in your league, you may look to sell high on Jonathan Toews, as his recent pace of 3.0-4.0 G/week overvalues him, with 1.9 G/week as a more sustainable number. Of course, if you play in a keeper league, you probably want to stick with the young Blackhawks’ captain.
Ryan Kesler may be overvalued by some fantasy owners and undervalued by others. Kesler has received a boost recently by playing the last dozen or so games with talented veterans Pavol Demitra and Mats Sundin. While his 3-6 game performance has trailed off, he’s likely good for between 1.2 G/week, his Last 27 game performance level, and 2.0 G/week, his Last 12 game performance level.
Jason Arnott has been on fire as of late, culminating in his hat trick in the Predators’ unexpected 8-0 mauling of Detroit this weekend. While he’s been a quality player, don’t expect a 2.5 G/week or better pace out of the ex-Star and ex-Devil. His Last 27 game performance indicates a 1.4 G/week future output.
Tomas Plekanec has been toasty warm, putting up a 3.0 G/week pace over the last two weeks. We know to expect 1.2 G/week from him, so he is not worth adding unless you are particularly thin at Forward. If he’s on your roster, this is the time to sell high on him if you can find a buyer.
You should rarely believe in an offensive outburst from Defensemen, with Anton Babchuk, Pavel Kubina and Sheldon Souray as this week’s examples. You can pretty much bank on the fact that their next fantasy week will look much more like the sub-1.0 G/week that they have put up over the Last 27 games, rather than 2.0-3.0 G/week they have managed to put up recently. These blueliners may be free agents in your league. If so, proceed with caution in picking them up, especially if plus/minus is a major consideration for your league.
Players in recent trades
Number of Goals in Average Goals/fantasy week based on
Previous x games
Player Team Pos 3 6 12 27 3 6 12 27
Rich Peverly ATL F 4 4 4 9 4.0 2.0 1.0 1.0
M. Schneider MON D 1 3 3 4 1.0 1.5 0.8 0.4
Chris Kunitz PIT F 1 3 4 8 1.0 1.5 1.0 0.9
Ryan Whitney ANA D 0 1 1 2 0.0 0.5 0.3 0.2
Rich Peverly has been playing with Atlanta for 21 games now since his trade from Nashville and he has benefited from playing time with Ilya Kovalchuk. His recent hot streak may entice some to pick him up as a free agent in their leagues, but his Last 12 game and Last 27 game performances peg him as merely a 1.0 G/week scorer.
Mathieu Schneider, on the other hand, left Atlanta for his old stomping grounds of Montreal. Time will tell if his scoring will consistently increase from the anemic 0.4 G/week pace with the Thrashers to anything like the 1.0-1.5 G/week pace that he has put up since the trade. As mentioned in my Mucking and Grinding column last week, Schneider’s plus/minus took a huge tumble from last season by moving to the offensively and defensively challenged Thrashers. With moderate gains likely both in scoring stats and in plus/minus through his return to the Canadiens, Schneider should now be worthy of taking off of your bench and inserting into your Active lineup.
Last week’s big trade exchanged Chris Kunitz for Ryan Whitney. These players may be available as free agents in your league, as the 0.9 G/week and 0.2 G/week they have been putting up has not been much to write home about. Kunitz has a history of performing well when paired with star forwards. It is possible that he will get a boost from playing with one of the top lines in Pittsburgh, but he remains a marginal addition to your team with some slight chance of upside. Whitney has yet to really hit stride after coming back from injury and the trade to the Ducks will not provide him a boost in fantasy value. Assuming that Anaheim retools quickly and well, he will have some keeper value.
That's the story as far as predicting future performance in Goals scored, which hopefully should give you a leg up on your competition. Even if your fantasy league does not offer you better options than sorting by last 28 days or YTD, it should not take much time for you to look at the game logs of your players or free agents that you are interested in to get the best possible data for your own analyses. Sum the Goals for the last 27 games in each game log and divide by 9 to get the expected performance per fantasy week going forward.
For those of you heading into fantasy playoffs this week - Good luck!
Timo Seppa is an author of Hockey Prospectus.
You can contact Timo by clicking here or click here to see Timo's other articles.
| 
0 comments have been left for this article. 
Conn Smythe Watch: Lundqvist Leads