Oct 5, 2016
- I'm posting updates to the boys every other day as I clean up the match database.
- The girls U11-U15 should be up in another week.
- The first draft of the boys LP 2016-2017 rankings are up.
- Due to the birth year change, last year's data has been dropped. It'll take awhile for all the teams to show up since teams need 7 games before they appear on the tables.
- Last LP Ranking will be posted end of May. Then there is a 3 month gap as we collect summer data.
- How will we deal with the birth year change? Likely previous years' data will need to be dropped given that many teams have changed substantially.
- PNW Boys U11-U18 ratings updated biweekly. Focusing on WA, OR, ID.
- PNW Girls ratings posted. Focusing on WA, OR, ID. Western G01 tables cover all of the West though.
- PNW Girls ratings posted. Focusing on WA, OR, ID since adding CA is taking too much time. Western G01 tables cover all of the West though. Tried posting the team resolver files and getting volunteers to help, but in the end I had to enter the data myself. Need to come up with a different way to enter data.
Sept 2014
- PNW Boys ratings posted. Focusing on WA, OR, ID since adding CA is taking too much time.
Summer 2014
- Entering the summer data for WA-OR boys. Trying some different ideas to make it easier for others to help with the PNW girls; not much progress yet.
May 2014
- Summer 2014-2015 Tournament Guide posted (left nav column under LP Tools).
Nov 2013
- Working on adding the rest of the PNW girls ages.
- Revamping navigation on the site.
- Fixed a minor bug in the match predictor that messed up the menus.
- Match predictor has updated to 2013 model.
- Switched to glmnet() for parameter estimation, which can be forced to do non-penalized glm and is much much faster for large glms with 10000s of categorical variables
- Posted update of fbRanks to CRAN
- 2013 ratings are posted. Click links in the upper navbar.
- Note, match predictor has not been updated to 2013 model yet.
- I've switched to a total strength that looks more like a "chess" rating. Meaning is still the same through. It is the difference in strength that matters and that difference tells you the RELATIVE scoring rate of the two teams in a match up. Difference of 100 = stronger team scores on average 2x as fast. So for every goal by the weaker, you expect 2 goals by the stronger team. Difference of 200 = 4x faster scoring rate by stronger. Difference of 300 = 8x faster scoring rate. So it's rare for a team that is 300 points lower to win. Time to pull out the 9-1 formation...
- Now that the 2013 team are pretty much sorted out, I'm returning to work on scaling up the algorithm. Right now, I'm using a simultaneous calculation and even with speedglm, the size of the models is maxing out around 2300 teams. I'm working on various asynchronous algorithms that will be slow but ultimately scalable (I hope).
- Scraping and entering tournament data
- The 2013-2014 rankings will be posted hopefully mid-August after some of the bigger tournaments are done.
- West Coast G01 rankings updated in July 2013
- Finished an analysis of strength of A teams across all clubs in WA.
- Working on B95s and trying to get a good set of games for the west coast top 20. Goal is to get an idea of there the upper strength at age line lays.
- I released a R package called fbRanks for doing poisson regression ranking on the R CRAN site: fbRanks.
- I posted draft B96 rankings for WA, OR boys.
- I am slowly getting some BC ranks up. I started scrapping the MSL league in the Vancouver area and adding those teams to the database. I haven't decided whether to continue ranking the BC teams next year. The teams are awful about including age or team name (!) in leagues and tournaments, so it's very time consuming to figure out who is playing where. BC club webpages rarely list their teams, so it's takes a lot of research to figure who is playing where and what teams a club is fielding.
