Quick: What is the biggest hindrance to the usefulness of statistics in flat track roller derby?
Give up? The answer is sample size.
Let's look at a sport known for it's use of statistics: baseball. In baseball, an everyday player will come up to bat about 600 times over a 6-month season. A regular starting pitcher will face a lot more hitters than that: 800 is common, and several pitchers reach 900 each year. With such gigantic samples, it's easy to draw definable conclusions from even a single month's statistics. Once you get around 100 plate appearances, you can start evaluating and make adjustments throughout the season.
So how long does it take for a jammer to rack up 100 jams? One year, and that's if she's lucky. Unfortunately, you need to evaluate your team a lot more often than once a year. So most of your mid-year stat analysis is going to be based on insufficient sample sizes. Some of the data is pointing at a trend, but anomalous jams are throwing off the math. How do you tell a good differential built on consistency from a bad differential skewed by 1 or 2 big jams?
The answer is so simple that I can't imagine why I didn't think of tracking it earlier: Look at the jam win-loss record. In nearly every jam, both teams take the track with one goal in mind. No, that goal is not to get lead jammer; that's just a means to an end. Any team would let the opponent ultimately get lead jammer if it means that they can score 5 points before that happens. The goal is to win the jam. Tracking jam wins and jam losses is a simple way of seeing how often you are successful at achieving that goal.
Time to play make believe: You're evaluating a new jammer's performance in last Saturday's bout. Do you want her to be a regular part of the 3-jammer rotation in the upcoming tournament, or a secondary jammer who you can put in here and there to fill holes? Do you want to rely on her in a key situation to come through with a good jam?
Let's say the jammer in question had these 8 jam results last weekend: 0-4, 0-2, LJ 3-0, 0-0, LJ 24-0, 0-4, 0-9, LJ 0-0.
Her line would look like this: 3 for 8, 27 points, +8 differential. Pretty solid, eh? Not the best lead jam percentage, but still pretty good overall. A jammer you can count on in an important jam?
What about her jam wins and losses? 2 Jam Wins, 4 Jam Losses.
Well now, he have a red flag here. The win-loss record points to inconsistency. She scored the vast majority of her points in a single jam where she only had to skate laps while the opposing 2-pack was helpless. Apart from that jam, she struggled. Is this a jammer you want on the line in a tie game with 2:45 to play? The result of this jam could mean the difference between winning and losing the bout, and she appears to be twice as likely to lose the jam as she is to win it.
Obviously, I tailored the hypothetical situation here to make my point, but it still remains valid. A jammer with a positive differential who has won more jams than she has lost has been consistently good. A positive differential jammer who has lost jams more often than winning has probably gotten lucky. Of course, the other side of that is true as well. You could have a jammer with a negative differential who has more jam wins than jam losses and is being undervalued due to a couple bad jams.
Yes, there are several factors you should take into consideration before you choose who's in your main jammer rotation. (Which skaters are improving the fastest? Who is still recovering from an injury?) There are even more things you have think about when you decide who gets the star in a critical game situation. (How have each of your jammers been skating today? Who is rested and ready to go?) Still, I think consistency statistics like jam wins and jam losses can be helpful in evaluating skaters, lines, and teams as a whole.
No comments:
Post a Comment