paper: Stop the Stop Build

[Jim Zondag]

Originally Posted by Greg Woelki

Great read, thanks for posting!

Minor stats comment: Fig (5) is a little concerning because each curve represents a different population of teams, so it isn't a very clear way to show the trend of increasing performance as teams have attended more and more events. It is unclear to what degree teams from the left-hand side are moving rightwards as they attend more events or if most of those teams simply aren't included in the next curve. While it can be surmised by looking at the right-hand bounds of the distributions there are some performance increases, the graph would more directly support your point if a single population of teams (perhaps the 304 that competed at 4 events or the 765 that competed at 3) were tracked across their multiple events instead.

To answer: Fig(5) is this graph:

Each population of teams is a subset of the previous group.
3114 teams played this year, they all played at least one event (Blue)
Of these teams, 1928 teams played at least 2 events (Red)
Of these teams, 765 teams played at least 3 events (Green)
Of these teams, 304 teams played at least 4 events (Orange)
Of these teams, 58 teams played 5 or more events (Black)
The chart shows the progression of skill improvement by the population with each consecutive event played.

This trend is basically the same every year, regardless of the game, the only change is the magnitude of the vertical axis, which is a function of the annual game design and how many points are available to be scored.

To see the trend more clearly, the dotted black line in Fig (6) shows how the averages of each of these group subsets increases through the season.


So, in a nutshell, if you choose to play late, odds are there are more experienced teams in the house who have progressed in skill while you have been waiting.