From the Data to the Super Bowl - End of Season Power Rankings and Week 17 Recap

Week 17 Recap

Week 17 was rather straight forward. The model was able to predict  13 out of the 16 games performing better than most expert picks. The two games incorrectly predicted were the CHI/GB game (low confidence prediction) and the MIA/NYJ game (self-destruction by MIA).  Lots to learn as I move forward to apply this next year!


Posted below are the power rankings for the end of the season.  I will be posting post-season predictions later this week. Should be exciting.


Week 17 Power Rankings:
 Note: Disagree is denoted by a disagreement of >5 spots between ranking systems.  

Rankings	Team	Disagree	ESPN Ranking
1	SEA		SEA
2	SF		DEN
3	ARI		CAR
4	DEN		SF
5	STL	***	NE
6	CIN		NO
7	CAR		CIN
8	PHI		IND
9	MIA	***	ARI
10	NO		KC
11	PIT		PHI
12	BAL		SD
13	SD		GB
14	TB	***	PIT
15	DET	***	CHI
16	IND	***	BAL
17	KC	***	DAL
18	MIN	***	MIA
19	NE	***	NYJ
20	TEN		STL
21	JAC	***	DET
22	CHI	***	TEN
23	ATL		NYG
24	BUF		MIN
25	CLE		BUF
26	GB	***	ATL
27	HOU		TB
28	NYG		CLE
29	DAL	***	JAC
30	WAS		OAK
31	NYJ	***	WAS
32	OAK		HOU

From the Data to the Super Bowl - Week 17 Picks and Games to Watch

Week 17 Predictions: 

The last week of the season is likely one of the most difficult to pick. With some teams calling it quits, others protecting their stars for playoffs, and other still with lots on the line. This year is not to be outdone with star players making returns and others injured.

For those new to reading my analysis  go back and read the early entries on the method.

Notes:

1. Yes I love numbering things. Stop mocking me. Didn't the fact I run stats on sports give away my love of numbers.

2. I decided to spice it up and compare to a Canadian sports writer - Bruce Arthur and his picks for week 17.

3. Games to watch: DET/MIN, WAS/NYG, GB/CHI (Obviously), PHI/DAL (Yes i know...no brainer). 

4. Some teams have lost some major players or called it quits for the year/saving for the playoffs. There is still lots on the line so I doubt any team is really that willing to lose. 

5. Let the Playoffs begin!!

 

Game	Predicted Winner	Level of Certainty	Vegas Odds	Adam From ESPN	Bruce Arthur
CAR @ ATL	CAR	High	CAR	CAR	CAR
BAL @ CIN	CIN	Medium	CIN	CIN	BAL
HOU @ TEN	TEN	Medium	TEN	TEN	TEN
JAC @ IND	IND	High	IND	IND	IND
NYJ @ MIA	MIA	Very High	MIA	MIA	NYJ
DET @ MIN	DET	Very Low	MIN	MIN	MIN
WAS @ NYG	NYG	Very low	NYG	NYG	WAS
CLE @ PIT	PIT	Very High	PIT	PIT	PIT
GB @ CHI	CHI	Low	GB	GB	CHI
DEN @ OAK	DEN	Very High	DEN	DEN	DEN
BUF @ NE	NE	Medium	NE	NE	BUF
TB @ NO	NO	Medium	NO	NO	TB
SF @ ARI	SF	Low	SF	ARI	SF
KC @ SD	SD	Medium	SD	SD	SD
STL @ SEA	SEA	Medium	SEA	SEA	SEA
PHI @ DAL	PHI	Very High	PHI	PHI	PHI

From the Data to the Super Bowl - Week 17 Power Rankings and Week 16 Recap

Week 16 Recap



Week 16 was a tricky one with upsets and heart-breakers. Last week's predictions were 9 of 16. Vegas was 10 for 16 and the expert was also 9 for 16. Adjustments for recent performance and home field advantage may have been able to improve the predictions.


As we enter the last week of the NFL season the picture becomes even more murky. Posted below are the power rankings for week 17. A few things from this weeks analysis became evident:

Week 17 Notes:

1. This analysis takes the whole season's performance into account. The team that played week 1 is considered the same as the team that played week 16. We all know this is not true based on team dynamics and injuries. The ESPN power rankings are based on recent performance - a lesson that can be learned from (Although I do think the power rankings bias towards larger market teams....I'll prove it). Perhaps for next year I will take into account the prior performance and create a moving average algorithm as the season moves along.

2. Scores matter. You need to win by large margins throughout the season.If you want to be ranked higher than someone you need to beat who they beat by MORE! More is better.

3. Denver being so low makes sense only based on the recent loss to SD and the margins decreasing.

4. One big win by NE does not erase the close games they have lost/won throughout the season to some bad teams (see point 1).

5. SEA still on top even after the loss.

6. I have added last weeks ranking based on feedback I received. Keep the feedback coming!

This weeks predictions to be posted on Saturday!

Week 17 Power Rankings:
 Note: Disagree is denoted by a disagreement of >5 spots between ranking systems.

Last Week	This week		Disagree	ESPN Ranking
1	1	SEA		SEA
2	2	SF		DEN
5	3	ARI	***	CAR
3	4	STL	***	SF
12	5	CIN		NE
7	6	CAR		NO
15	7	PHI		CIN
6	8	NO		IND
11	9	PIT	***	ARI
8	10	MIA	***	KC
4	11	DEN	***	PHI
13	12	BAL		SD
10	13	SD		CHI
16	14	TB	***	BAL
14	15	DET		DAL
21	16	IND	***	MIA
9	17	KC	***	GB
18	18	MIN	***	PIT
19	19	NE	***	STL
20	20	TEN		DET
23	21	JAC	***	TEN
17	22	CHI	***	NYJ
22	23	ATL		BUF
24	24	BUF		NYG
25	25	CLE		MIN
30	26	GB	***	TB
31	27	HOU		ATL
26	28	NYG		CLE
27	29	DAL	***	JAC
28	30	WAS		OAK
32	31	NYJ	***	WAS
29	32	OAK		HOU

From the Data to the Super Bowl - Season Network Visualization

 Season Network Visualization 


This is the network map visualization of the analysis that is ongoing. This map represents all the games of the 2013 season. These are all the comparisons that are placed into the model and analyzed in the simulations. More importantly it looks pretty cool .

Notes when looking at the network map:
 
1. Every node (point) is a team.
2. The nodes are clustered by division. 
3. Every line represents a game played between the two teams.

This specific network has no weighting to it yet. This will be future fun!



Enjoy!

Special Thanks to Cody for creating this!


















 Network Map of all games in the 2013 NFL Season:

From the Data to the Super Bowl - Week 16 Predictions

Week 16 Predictions:

Game	Predicted Winner	Level of Certainty	Vegas Odds	Adam From ESPN
MIA @ BUF	MIA	Very High	MIA	MIA
NO @ CAR	NO	Very Low	CAR	CAR
MIN @ CIN	CIN	Medium	CIN	CIN
DEN @ HOU	DEN	Very High	DEN	DEN
TEN @ JAC	TEN	Medium	TEN	TEN
IND @ KC	KC	Low	KC	KC
DAL @ WAS	DAL	Very Low	DAL	WAS
CLE @ NYJ	CLE	Low	NYJ	NYJ
TB @ STL	STL	Very High	STL	STL
ARI @ SEA	SEA	Very High	SEA	SEA
NYG @ DET	DET	Medium	DET	DET
NE @ BAL	BAL	Low	BAL	BAL
OAK @ SD	SD	High	SD	SD
PIT @ GB	PIT	High	GB	GB
CHI @ PHI	PHI	Low	PHI	PHI
ATL @ SF	SF	Very High	SF	SF

Notes:

1. Some games to watch: NO/CAR, CLE/NYJ, PIT/GB.
2. Still no home-field adjustment or  any adjustment for injuries and crazy coaches (Yes that is a skins jokes)
3. If you lose money using this- it is not my fault. On the other hand - if you win I want a cut!
4. Work in progress. I would love some feedback.

 Lets do this!

From the Data to the Super Bowl - Week 16 Power Rankings and week 15 results

Week 15 Results and lessons learned:

1. The predictions were able to correctly select 10 out of 16 games. Not great but in my defense the experts and the Vegas picks were also 10 out of 16. It was a week of upsets ( this season has been crazy).

2. The results raise the question of what would be the bar needed to show that this network analysis works. There will always be upsets but the hope is to be able to improve the ability to ranks teams and predict with greater certainty than the current system of prediction. There is lots of work still to be done to improve the predictability. See past discussion of limitations.

3. There needs to be some sort of home field adjustment. Not all teams are good home teams but some seem to be unable to leave home (*ahem* New Orleans). This will be a work in progress. 

4. I have adjusted the way I do the Power rankings and now take into account multiple rankings to compute the final ranking. This will give a more realistic placement. 

Week 16 Power Rankings:

Note: Way off is noted for >5 spots.

	Power Ranking	Disagree	ESPN Rankings
1	SEA		SEA
2	SF		DEN
3	STL	***	SF
4	DEN		CAR
5	ARI		NO
6	NO		KC
7	CAR		NE
8	MIA	***	ARI
9	KC		CIN
10	SD		IND
11	PIT	***	PHI
12	CIN		BAL
13	BAL		CHI
14	DET		MIA
15	PHI		SD
16	TB	***	GB
17	CHI		DET
18	MIN		DAL
19	NE	***	PIT
20	TEN		STL
21	IND	***	TEN
22	ATL	***	NYJ
23	JAC	***	MIN
24	BUF		TB
25	CLE		BUF
26	NYG		NYG
27	DAL	***	CLE
28	WAS		ATL
29	OAK		JAC
30	GB	***	OAK
31	HOU		WAS
32	NYJ	***	HOU

Some comments:

1. St.  Louis in 3rd place really makes me question the mode. I  went back and looked closer at the results and found why-They like to beat good teams....by a ton! They also go on to lose by a little to other teams. Their 6-8 record may actually be misleading. I agree not 3rd place misleading but again beating New Orleans will help.

2. ESPN loves New England.....the data not so much. They dont win by a lot and lose to bad teams. Enough said.

3. Dallas has a lower ranking in the network model. I believe that this is the ESPN bias at play, placing bigger market teams higher up. They just don't win as much as other teams. 

4. The new way I have compiled the rankings has shifted things a little bit as noted. I am surprised by the drop for Green Bay even after a big win this week (against a low Dallas Team)

I will post my week 16 picks later this week. 

From the Data to the Super Bowl - Week 15 predictions

Firstly- this is a work in progress but as promised here are the picks for this week. This is a pretty straight forward week. The two games to watch will be the Hou/Ind and NE/MIA. The data points to a possible upset occurring.

Some quick notes:

1. As you can see it was wrong about the Thursday night game. In my defense- everyone was wrong about Thursday game. 

2. The level of certainty is based on the probability of one team winning. Each estimate has a range around it.  The higher that range is the more certain we are. Very few predictions come out with 95% certainty most of them have varying levels.  I tried to categorize them to help better predict in the future.

3. Just like I mentioned in the last post- I did not take into account home field and injuries. These are future works.

4. Thanks for all the feedback into better completing the rankings and now the predictions. This is a work in progress.

5. I also picked one of the more popular analysts online to see how we match up against them. This week is not that much fun since everything aligns pretty well.

Game	Predicted Winner	Level of Certainty	Vegas Odds	Adam From ESPN
SD @ DEN	DEN	Low	DEN	DEN
WSH @ ATL	ATL	High	ATL	ATL
CHI @ CLE	CHI	Medium	CHI	CHI
HOU @ IND	HOU	Very Low	IND	IND
NE @ MIA	MIA	Very Low	NE	MIA
PHI @ MIN	PHI	Very low	PHI	PHI
SEA @ NYG	SEA	Very High	SEA	SEA
SF @ TB	SF	High	SF	SF
BUF @ JAX	BUF	Low	BUF	JAX
KC @ OAK	KC	Very High	KC	KC
NYJ @ CAR	CAR	Very High	CAR	CAR
GB @ DAL	DAL	Low	DAL	DAL
ARI @ TEN	ARI	Very High	ARI	ARI
NO @ STL	NO	Low	NO	NO
CIN @ PIT	CIN	Low	CIN	PIT
BAL @ DET	BAL	Very Low	DET	DET

A Network of Glory:From the Data to the Super Bowl! - Entry 1

"If the Lions beat the Packers and the Packers beat the Bears then that must mean the Lions can beat the Bears" - As kids this was the constant assessment we were  making of our favorite teams (more often than not I was making this about the Leafs in lost playoff heartbreaks).

Background

Analytically  what we were doing was not too far from an indirect comparison. But faced with questions of match-ups this becomes more complex. In reality, more often than not these assessments may hold to be true. As we watch more games and match-ups we are better able to assess who we think will win - that's why sometimes predictions in the start of the season are so much more difficult to make. What we are doing is collecting data. Each team in a season is a new unit and all games a network of comparisons.

What if we could analyze that network to try to predict who will win? Could we simulate who is really the current best team? So that is what I set out to do with the NFL.

My recent research (yes I have a day job!) applied a rather novel means of analyzing data called a network meta-analysis (Click here to see my recent work and proof I did actually do this). This type of analysis attempts to use all the data to indirectly compare drugs to each other in how well they work or how safe they are. It applies Bayesian statistics which doesn't assume everything can randomly happen but rather that past information and inform our future predictions. Picture this - we have on the market drugs A and B but both A and B have no comparisons to each other because to gain access to the market they had to only compare to placebo (drug c). This area of work sets out to compare A and B indirectly by using the studies compared to placebo.

                                                           Source of Image : BMJ

This type of analysis can be directly paralleled with a season full of games. As we get more information of results we are able to indirectly compare teams and then can run simulations to try to predict possible outcomes and rankings.

What I did and hope to do (Analysis):

I took all the games of this 2013 season and recorded all the scores. The analysis I completed takes into account not only who wins the games but what the scores are (more importantly the difference in the score). I then ran the simulations and I am able to determine the probability for each team of them being in first to last place and the head to head odds of the win. I plan to compare my analysis (predictions) for the last 3 weeks of the season to random chance (50%) and the Vegas picks. I probably wont do as well as the experts but its worth a shot. Ill start today by posting my power rankings compared to the experts power rankings. Ill post predictions Sunday. The overall goal is to use the season data to predict the playoffs.

NOTE: I did not take into account if the games were played at home and injuries and  other major changes. These are obvious limitations to this analysis. I know the purists will argue this and I agree this is just a tool to help make those types of analysis. 

Results:

Simulated Power Rankings:

Rankings Simulation Way off Expert (ESPN) 1 Seattle (76%) Seattle 2 Denver (18%) Denver 3 San Francisco (15%) New Orleans 4 New Orleans (12%) New England 5 Kansas City (10%) San Francisco 6 Arizona (9%) *** Carolina 7 Carolina (8%) Kansas City 8 Miami (7%) *** Cincinnati 9 Philadelphia (7%) Philadelphia 10 Cincinnati (7%) Arizona 11 San Diego (7%) *** Indianapolis 12 Baltimore (6%) Detroit 13 Pittsburgh (6%) *** Chicago 14 Detroit (6%) Baltimore 15 St. Louis (6%) Dallas 16 Tampa Bay (6%) *** Miami 17 Chicago (6%) San Diego 18 Atlanta (6%) *** Green Bay 19 Minnesota (6%) *** Pittsburgh 20 New England (7%) *** St. Louis 21 Buffalo (6%) *** New York Jets 22 Dallas (6%) *** Tennessee 23 Houston (6%) *** New York Giants 24 Green Bay (7%) *** Tampa Bay 25 Tennessee (7%) Cleveland 26 New York Giants (8%) Buffalo 27 Indianapolis (9%) *** Minnesota 28 Cleveland (8%) Oakland 29 New York Jets (9%) *** Jacksonville 30 Jacksonville(6%) Atlanta 31 Oakland (16%) Washington 32 Washington (22%) Houston
NOTE: Way off is judged by greater than +/-5 spots from ESPN placement. Some interesting notes: 1. This analysis takes into account the season as a whole. So recent improvements and trends are not taken into account. So you see teams in some ways higher or lower based on the most recent games in the expert opinion. 2. The analysis weighs blowout difference heavily. So a team like Houston that had 2 wins but many close loses will perform better than expected. On the other hand a team such as New England that has had many close wins (and an upset to bad teams) will not perform as well as expected. Also getting blown out will hurt your ranking. 3. Some surprises in the rankings are : Arizona. New England, Indianapolis, and St. Louis. These moves are easily explained by point 2. 4. The top and bottom of the packs are easy to predict. The center portions are much more clumped. I will need to work on a better algorithm to trudge through the data for the ranking. Perhaps some sort of weighing. This hurts some teams that get upset by low ranked teams and makes their values across the board flat and hard to predict. A great example is Detroit who was hard to place due to their loses to low teams and wins against higher ranked teams (no bias here).