What is cMax?
The results rank teams, and more importantly, provide a relative estimate of how fast a team is. The rankings are intended for entertainment purposes, and are not intended to replace official league and coaches polls.
The "Secs Back" figures predict how much each ranked team will be behind the #1 ranked boat in an average "standardized" race; approximately six minutes for men's eights, seven minutes for women's eights. So, if Team A is 12 "Secs Back", then they are predicted to be 12 seconds behind the #1 team in a standardized race. To compare two teams, just look at the difference in their "Secs Back" numbers. If A is 12 seconds back and B is 16 seconds back, then the model predicts that, on average for the standardized races, A will beat B by 4 seconds.
(Note that since each new ranking includes the full aggregate results of the season, sometimes crews that have placed ahead of another crew might be ranked behind the same crew. If the crews remain consistent in relative speed, this tends to correct itself over time as more racing occurs.)
You'll see that there is much more information in the "Secs Back" than in the rankings. In some cases, closely ranked teams have very different "Secs Back" and are thus not likely to be very competitive... and in other cases, 10 teams are within 1 second of one another.
The model uses a well-known statistical technique called OLS (ordinary least squares). It is based on the premise that although individual race results may vary, if Team A is on average 2 seconds better than Team B, and if B is on average 2 seconds better than Team C, then, on average, Team A will beat Team C by 4 seconds.
In addition to head-to-head race times, the model considers 1) general race conditions (distance, current, wind, etc.), 2) date of the race (more recent races count more than races earlier in the season), and 3) whether the team was "pressed" in the race (was there another boat close by at the finish or did they win by open water?).
Heat results, which are given less weight than Finals results, are included in the analysis only to the extent that an outcome was "pivotal." So, for instance, if the top three boats in a six-team race go to the Finals, and the bottom three go to the Petites, only the heat race times for the third and fourth place boats that battled to get into the Finals are included in the model. The model does not consider "home-court" or lane advantages, but does consider equipment failure to the extent that is reported at www.row2k.com .
The statistical model is based solely on the current season Varsity 8's race results as reported at www.row2k.com , beginning with races in February and continuing through last weekend. Some effort is made to clean up the data, as there can be errors in the posted results. Teams are included in the final rankings only if they have raced at least twice against other teams that have raced at least twice.
cMax collegiate crew ratings were developed by Chris Maxwell. Chris currently teaches Sports Economics and Game Theory at Boston College. He has a BA in Mathematics and Economics from Penn and a Ph.D. in Economics from Harvard.
Andrew Kirk does an amazing job pulling the race results database together on a weekly basis. Chris has been producing cMax collegiate crew ratings since 2002. He has fallen out of a single into the Charles many times and does not pretend to know much about rowing or sculling. Fortunately, others have helped in the development of the model -- most notably, Tom Murray with rowing and Chris Cavanagh with statistics.
Any comments or questions can be sent to Chris by writing row2k.
You will see tabs at the top of the page; clicking through the tabs will show you this week's results in different presentations, including by rank, alpha by team, and by division for women's eights.