We all come across urban legends from time to time -- things that get repeated over and over and many people believe to be true, but aren't. Well, skating is full of them. When I first came to skating, already well into my career as a research scientist, it struck me how many skating urban legends there are, how easy it is to quantitatively test them, and how no one in skating ever bothered to do it. Thus began my idle amusement of putting urban legends to numerical test.
One of the oldest urban legends in skating is that skating first is a disadvantage to a competitor. This legend has been repeated quite often in the past year, with advocates of CoP claiming it would eliminate this curse that is supposedly present in the current 6.0 system. It has been repeated most recently by the USFSA in a press release which endorses the use of CoP.
That skating first is a disadvantage, is myth that is easy to disprove. The numerical evidence was clear ten years ago, but the myth persists. So we look here again at the questions: is skating first in an event a disadvantage to a skater?
In a totally random draw, the order of finish should be completely uncorrelated with the start order. However, if skating first was a disadvantage, then competitors who skate early in the start order would place lower on the average than competitors who skate near the end of the start order.
To come up with a statistically meaningful answer one has to look at many events. In this report we look at the results from the novice and junior short programs and compulsory dances at the most recent U.S. National Championships. These ten event segments all had 12 or 13 competitors and are thus directly comparable. Among these ten event segments we then compare the results for the first three skaters in each segment and the last three skaters in each segment.
For the first three skaters in the segments under consideration there are 30 competitors. If there was no disadvantage to skating first then on the average half of the competitors should end up placing in the top half and the remainder in the bottom half. This was indeed the case with 14 of the 30 skaters placing first through sixth, and 16 placing seventh through last.
Further, on the average, 2.5 competitors should end up in each place from first through last, 7.5 of the 30 should end up placing first through third and 7.5 should end up placing tenth through last. Of course you can't have half a competitor, and there is also a statistical uncertainty determined by the number of cases which determines the margin of error in the statistics. What one finds is ten competitors placed first through third and eight placed tenth through last. That is, the competitors in this sample who were first to skate placed first through third more often than they placed tenth through last.
If skating early was a disadvantage then the competitors who skater later in the start order should do better, and end up with higher placements more often than lower placements. Among the competitors who skated tenth through last in the start order (a sample size of 34), 19 placed first through sixth and 15 placed seventh through last, eight placed first through third and seven placed tenth through last. Among the competitors last to skate, the occurrence of placements in the top three places was statistically equivalent to the occurrence of placements in the bottom three placements.
In examining the results of ten event segments at U.S. Nationals with totally random draws it is found that there is no statistical relationship between start order and placement! Skaters who skate early in the start order are just as likely to place well as competitors who skate last. This confirms similar studies done over the past ten years. The effect of start order on skater placement is not an issue relevant to the decision to use any particular scoring system since no such effect exists!
Copyright 2004 by George S. Rossano