We all have our addictions and my current time sink is fantasy basketball. I got myself involved in a winner's league in yahoo! and the game shifted form. Whereas last season I played in an accumulate the totals on the season league - this year the game is head to head on a weekly basis. As the season winds down, I have found that my numbers based approach has secured my berth in the playoffs with three weeks of the regular season left.
What makes this all the more interesting is that I do not have a television and do not watch the games themselves. I have a feel for who the players are, but other than looking at the numbers and reading a daily article or two - the only team i follow is the Portland Trailblazers. That is a very good team to root for this year as a 40-18 record provides the first reasonable team since the jail blazers got broken up due to owner embarrassment.
The reason that i am writing this is to introduce the value of derivative statistics. Everybody has a sense of derivatives in banking - they are marginal tricks of second generation mathematics that allow the realization of meta-trends - things that wouldn't necessarily sift themselves out unless you were looking broader than the initial numbers. For this game - i use the yahoo! standard 9-category algorithm as my initial metric.
Each player in the game is ranked on a daily basis, based upon metrics that are gathered from the box score of every NBA game. If a player plays on a particular day, he affects all 9 metrics that include both accumulated offense and defensive statistics and scoring percentage. Each of 12 teams carries 13 players - there is a limit of ten players in the line-up on any given day. Most days, most team's players can play every game - once a week there may be a conflict.
The game requires maybe 20 minutes a day worth of attention - to change line-ups, move players into positions to keep everybody playing the maximum amount when they are rostered and to replace injured or slumping players. There is a lot of hype in the game of basketball and the 'name' players do not necessarily come across as the statistical best. This year - nobody is close to Kevin Durant.
I spend another half hour per day collating and digesting the player statistics. I keep track by hand in a journal - that has been my learning method of choice for many decades - it is a memory enhancer and a calendar substitute. I never fail to follow up on things that i have written down, except when i lose track of the current journal. I generally run two concurrently, for that very reason. I plan to mine the journals for ideas, eventually, but toting around nearly 50 volumes of daily grind is not a convenience that i can afford. I do have the last one, then next one, one almost finished and two new blanks - to choose one from.
I plot the change in Yahoo! rank each day and note whether the player actually plays a game that day. If a player is injured, his rank goes down marginally for each game that he doesn't play. I note the variance of each player's ability for the continuity of only games played and the sum of the entire season. By the end of the season, the 82 game schedule will provide statistical significance based on player performance.
Thus this one rank will give me a summary of the nine other measures. I am only keeping track of my own team; i suspect that with a decent algorithm, i could generate this measure for the entire league. It might be interesting to plot each of the starters on each of the teams. In retrospect - perhaps i can, now that the idea is registered - all it would take is the time to do the research.
Now - if we can do this with athletic performance, why can't we do it with middle school and high school performance measures; such that our kids can have real feedback with what they know and don't know, relative to standards and other kids? In fact, if the entire workforce had valid measures of how their performance stacked up to where it mattered in pay grade, don't you think things might get ship-shape rather quickly? If we scored on actual performance, using proper metrics, then ...
Namaste' ... doc