Tuesday, January 01, 2013

Predicting the outcome of the NFL playoffs

It's the time of the year to power up the simulations and attempt to predict the course of the NFL playoffs.  As in year's past I have used the Sagarin ratings and a standard deviation of about two touchdowns to build a monte carlo simulation of the playoff games.  The results are shown in the plot below.

On the AFC side, Denver and New England are the most likely teams to make it to the Superbowl at 44% and 41%, while the other teams are also-rans and make up the last 20% or so of simulations.  The AFC is more spread out.  With San Francisco, Atlanta at 37% and 28%, with Seattle and Green bay picking up some more of the possibilities at 17% and 11%.

Since the goal of this simulation is to try to win the RKB playoff fantasy football pool we also care about how many games each team plays and what players score the most points.  The plot below shows the number of games each team plays in the playoffs, teams with a bye can only play a maximum of three teams, the others can play four if they make it to the final game.  The top plot is the simulation, compared to the bottom plot where the win probabilities are strictly 50/50.


MIN, WAS, CIN and IND have a disproportionate chance of losing there first playoff game so players from those teams will only have a single game to earn points.  One team with a bye, Seattle, seems to hold its own and play through to the Superbowl a good number of times, and certainly high scoring players from that team will have two or three games worth of chances to score points.  As reflected in the pie charts above, NE and DEN play three games in many simulations.

Next step picking the right players given the expectations of which teams will play.