NFL Survivor Pool Pre Week 2

After going to for two on week one, I have now incorporated further information (most notably Fivethirtyeight's ELO rankings and Jeff Sagarin's USA Today rankings) and came up with this week's two picks of the week: Baltimore (Home and 80% chance of winning over Oakland) and the New Orleans Saints (who have a whopping 89% chance of victory at Tampa Bay).

This week, due to the results from last week, I decided to weigh heavily on both ESPN's FPI rankings and Fivethirtyeight's ELO rankings. I waited them heavily with a score of nine apiece while using the JSR rankings were only given a weighting of one.

New Orleans' 89% chance of winning weeks to a game at Tampa Bay is not surpassed, presently, for the rest of the season. The only two weeks of rate of 70% chance of winning are both weeks 14 and 16 against Tampa Bay and Jacksonville respectively.

Coming in at 80% versus Oakland, Baltimore has a greater chance of winning in week 5 against Tennessee and week 10 against Jacksonville.

Other notable this week, Miami and Jacksonville) is the only team to break 70% chance of winning their game this week.

Team	Odds of Winning
ARI	65%
ATL	42%
BAL	80%
BUF	41%
CAR	63%
CHI	35%
CIN	63%
CLE	49%
DAL	45%
DEN	52%
DET	55%
GB	62%
HOU	37%
IND	69%
JAX	30%
KC	48%
MIA	70%
MIN	45%
NE	59%
NO	89%
NYG	58%
NYJ	31%
OAK	20%
PHI	55%
PIT	56%
SD	37%
SF	44%
SEA	38%
STL	63%
TB	11%
TEN	51%
WSH	37%

NFL Survivor Pool Post Week 1

So after week 1, I got a little lucky.  With my teams winning a sure fired combined score of 53-22 to start me off with a 2-0 record.

In the mean time, I have discovered two other ranking systems: Fivethirtyeight's NFL Forecasts (ELO) using the simple yet elegant ELO system and Jeff Sagarin's Ratings (JSR) on USA Today. I discovered these rating systems on Friday after the start of week one games.  To shuffle in these predictions into the prediction model, I applied a weighed average to each (on Friday) of 9:5:2 FPI:ELO:JSR.  It should be noted that New England had a greater than a 50% chance of winning from all three systems.

On Tuesday, I did my own post game analysis of the three rating systems plus my combo weight.  Although I am using probabilities to determine risk and to defer risk,  I still must ultimately make a deterministic choice.  So my initial analysis is quite deterministic.  If a team had a greater than 50% chance of winning and actually won the match, the score of 1 was awarded.  Otherwise a score of 0 was awarded.  The highest score is what is desired here.

Information	Week 1	Post Week 1
FPI Average Correctness	75%	88%
ELO Average Correctness	75%	88%
JSR Average Correctness	69%	88%
COMBO Average Correctness	69%	88%

While it is an interesting look, it doesn't tell the full picture.  The other factor I am noting is the 'average incorrectness'.  That is if a call was incorrect, what is the difference between the call and 50%?  Were they close games or blowout mistakes?

Information	Week 1	Post Week 1
FPI Average Delta Incorrect	17%	8%
ELO Average Delta Incorrect	10%	7%
JSR Average Delta Incorrect	(2.14)	(1.35)
COMBO Average Delta Incorrect	16%	7%

FPI seemed to yield slightly worse results than ELO. Both ELO and FPI got the Atlanta, St. Louis, and Tennessee games incorrect.  The ELO also got the Kansas City game incorrect, while FPI got the Buffalo game incorrect.

Looking deep into the games that were incorrect, one game stands out: the Tennessee game in the FPI system. Tennessee had just a 12% chance of winning the game. However they won the game, and they won the game in a relatively decisive fashion. However, upon further review Tennessee's FPI was rated at a -5.9 while Tampa Bay was rated at a -5.6. Except for Oakland and Jacksonville these two are the two weakest teams in the NFL. This is definitely a lesson in a wake-up call to myself, be wary if two-week teams are going at each other with low FPI's.

The only other game worth noting, would probably be the St. Louis versus Seattle game. It was probably extremely upset by any stretch of the imagination. Seattle came in with approximately a 70% chance of winning and ended up losing the game. Outside of these two outliers, I am pretty happy with the results.

My dad emailed me with the following question:

You may need to rethink strategy, Eagles made it look close but were not really in it, ELO appears better than FPI

to that I responded:

You are partly correct, ELO is marginally better than FPI.  They both got correct 12 of 16 games.  If we look at the "Average Incorrectness" - which I define as the average distance between the probability of a wrong pick and 50% of all incorrect picks (A pick is wrong if a team had less than a 50% chance of winning and actually won the game) - FPI had an average of 17% wrong and ELO was 10% wrong.  

However if we look closely, FPI really screwed up the Tampa-Tennessee Game.  Tampa was given an 88% chance of winning and lost.  However, if we look closely at the raw FPI of the squad, Tampa had an FPI of 0.5 and Tennessee had 0.2.  Both are absolute garbage.  If we remove this outlier, the average incorrectness drops to a palatable 12%.  

I'm actually pretty pleased with the results too.  If we look at games with odds close to 50-50 (I went as high as 65-35, i.e. to exclude outliers), the number of incorrect choices was about 15% (depending on scheme), much better than the expected number of incorrect of... well... 35% as if I was deciding a game by 100 coin flips.  

FPI, JSR, and ELO were all updated so I have tentative picks of Baltimore and New Orleans.

NFL Survivor Pool - Mathematical Model

The mathematical model developed to solve the decision making process is as follows:

Objective Function

Subject to

Choose 1 team in the first 16 weeks

Choose 1 team in the 17th week

Choose 2 teams per week

Bye week limiter

Definition of variables

NFL Survivor Pool Pre Week 1

Recently I got an offer to join an NFL Survivor pool.  The rules are simple:

1. Pick 2 teams to win each week.  Straight win, no spreads.  
2. In order to be eliminated, you must get 2 selections incorrect (in aggregate) during the duration of the pool
• Not to be confused with only being eliminated by getting 2 incorrect selections in same week
3. You cannot pick the same team twice in a season until all teams have been picked twice.  
4. Winner take all (all participants must be paid up before week #1 begins)

The problem I built a MILP to solve this and used ESPN's FPI as a sort of ranking method. 

I determined the 'odds of winning' for team A vs B as 

where HFA is a home field advantage if team a is the home team (HFA'=0 if the team is the away team).  FPI was also zeroed because there was some teams with a negative FPI.  Also because Oakland would have a zero FPI, I added HFA and a factor to 'bump up the score'

A mathematical integer program was developed to solve the conundrum surrounding rule 3: you cannot pick a team twice in basically a season.

The model is set up as a way to defer riskier picks (i.e. lower probabilities) to later weeks using the same idea of a discounted cash flow.  The model I came up with dictated I should pick CAR (@JAX) and CIN (@OAK) in week 1 versus a sure thing such as NE or GB.

The following table is the the overall picks I would make if I had to pick all 17 weeks in week 1.  

Week	Team 1 Pick	Opponent	Team 2 Pick	Opponent
1	CAR (60%)	@JAX	CIN (61%)	@OAK
2	BAL	@OAK	NO	TB
3	NE	JAX	SF	CHI
4	CHI	OAK	IND	JAX
5	BUF	@TEN	DEN	@OAK
6	GB	SD	MIA	@TEN
7	SD	OAK	WSH	TB
8	ATL	TB	HOU	TEN
9	NYJ	JAX	PIT	OAK
10	DAL	@TB	MIN	@OAK
11	DET	OAK	PHI	TB
12	NYG	@WSH	TEN	OAK
13	JAX	@TEN	KC	@OAK
14	ARI	MIN	CLE	SF
15	SEA	CIN	STL	TB
16	OAK	SD	TB	CHI
17	IND	TEN	KC	OAK

Of course as more information becomes available and the season progresses, we will get more information (updated FPIs and/or the possibility of new rating schemes) and be able to refine my choices for week 2 and beyond.

Note: this blog post was written before yet posted after week  1's games.  My bad.