Friday, September 14, 2007

USCL Predictions Week 4

Using the power rankings I generated, and developing a predictive model that takes into account the difference between the two power ratings, applies it to a lookup table, the model has generated the following predictions for Week 4 of the United States Chess League.

UPDATE: The goal of these predictions is to do this without knowing the lineups -- the power rankings take into account team performance only (not individuals) and so, therefore, does the model. It is risky, for sure, but it is a one parameter model, and I'm curious more than anything on how well it can do.

So, here they are for week 4.

New York Knights vs. Boston Blitz - Boston wins 3-1
Queens Pioneers vs New Jersey Knockouts - Tied Match 2-2
Philadelphia Inventors vs Baltimore Kingfishers - Tied Match 2-2
Tennessee Tempo vs Miami Sharks - Tied Match 2-2
Dallas Destiny vs Carolina Cobras - Tied Match 2-2
Seattle Sluggers vs San Francisco Mechanics - Tied Match 2-2

What a dull set of predictions!

Ok, yes, it is dull, but look at the power rankings...

NY (11) vs. BOS (1) -- wide disparity
QUE (9) vs. NJK (8) -- next to each other in the list
PHI (6) vs. BAL (7) -- also next to each other
TEN (10) vs. MIA (12) -- only one team separates
DAL (2) vs. CAR (5) -- two teams separate
SEA (3) vs. SFM (4) -- next to each other

So, the only one that seems like it is might be a decisive prediction is Dallas - Carolina, and indeed, a slight tweak in the model would make it so that it would predict 2.5 - 1.5 for Dallas, but there is no human intervention in the model output.

Lots of teams close to one another, and it should be an interesting week!

No comments: