2020 Fantasy Football - How to Make a Projection

I was going to publish a new set of rankings but with the opt outs coming hot and heavy and no set date for the cut off alongside camps starting I’m going to hold off on a finalized 2^nd set of rankings. Which means you get something I’ve been thinking about doing but wasn’t quite sure if it was too inside baseball. This is a deep dive/explainer of how I get to my projections, or as some of you might call it… “How a Guesser, Guesses!”

 

The first step just seems logical, look at where each leader of an offense has worked over the past three seasons. For most teams this will be their offensive coordinator except in places where the head coach is well established as an offensive minded coach. Think, Reid in KC.

OC or HC	19Tm	18Tm	17Tm
Kingsbury	ARI	-	-
D. Koetter	ATL	TB	TB
G. Roman	BAL	BUF'16	BUF'15
Daboll	BUF	BUF	NE'16
J. Brady	-	NO	NO
Nagy	CHI	CHI	KC
Taylor/Callahan	CIN	LAR	LAR
Van Pelt	MIN	MIN	MIN
K. Moore	DAL	DAL	DAL
P. Shurmur	NYG	NYG	MIN
D. Bevell	DET	SEA'17	SEA'16
LaFleur	GB	TEN	LAR
B. O'Brien	HOU	HOU	HOU
Reich	IND	IND	PHI
Marrone/DeFilippo	JAC	MIN	JAC
A. Reid	KC	KC	KC
A. Lynn	LAC	LAC	LAC
S. McVay	LAR	LAR	LAR
C. Gailey	NYJ '16	NYJ '15	-
G. Kubiak	MIN	DEN	DEN
J. McDaniels	NE	NE	NE
S. Payton	NO	NO	NO
J. Garrett	DAL	DAL	DAL
A. Gase	NYJ	MIA	MIA
G. Olson	LVR	LAR	JAC
D. Pederson	PHI	PHI	PHI
Fichtner	PIT	PIT	PIT
B. Schottenheimer	SEA	SEA	LAR'14
K. Shanahan	SF	SF	SF
Arians	TB	ARI'17	ARI'16
A. Smith	TEN	TEN	TEN
Turner	CAR	CAR	MIN'16

 

That’s pretty straight forward, except where there are gaps in their CVs or young coaches or guys coming from college. In those cases, the next step gets slightly more messy but pretty straight forward. The next step is to look at three pieces of data: Passes, rushes, and Pythagorean win% for each of the three previous seasons. Passes and rushes are pretty simple they’re just the number of each type of play the team ran that season. Pythagorean expected wins is a bit more complex. This is a concept popularized by Bill James’ work in baseball. He found that if you take a team runs scored and square them and divide by runs scored squared and runs allowed squared. That it was a solid predictor of wins expected for that team. On a simple level if a team scores as much as they allow, you’d expect them to be right around .500. The coefficient can be adjusted for each sport so for football the best fit has been found at 2.37. So, the Py columns are what we would expect the teams winning percentages to be based on points scored and points allowed to give us an idea of how good each team was. Something that will get factored in in our next table.

 

Note: Where there isn’t a history or data I just put in the league averages for passes and rushes.

 

OC or HC	Tm	19Pa	19Ru	19Py	18Pa	18Ru	18Py	17Pa	17Ru	17Py
Kingsbury	ARI	554	396	0.382	548	430	0.555	548	448	0.608
D. Koetter	ATL	684	362	0.473	625	389	0.407	605	390	0.423
G. Roman	BAL	440	596	0.818	474	492	0.532	465	509	0.532
Daboll	BUF	513	465	0.612	499	468	0.314	550	482	0.793
J. Brady	CAR	559	427	0.521	519	471	0.699	536	444	0.680
Nagy	CHI	580	395	0.463	512	468	0.719	543	405	0.618
Taylor/Callahan	CIN	616	385	0.275	568	459	0.679	518	454	0.708
Van Pelt	CLE	466	476	0.668	606	357	0.532	527	501	0.728
K. Moore	DAL	597	449	0.671	527	439	0.527	493	480	0.538
P. Shurmur	DEN	607	362	0.340	583	354	0.435	527	501	0.728
D. Bevell	DET	571	407	0.375	555	409	0.558	567	403	0.789
LaFleur	GB	573	411	0.607	437	456	0.514	518	454	0.708
B. O'Brien	HOU	534	434	0.489	506	472	0.639	525	448	0.354
Reich	IND	513	471	0.481	644	408	0.633	564	473	0.738
Marrone/DeFilippo	JAC	589	389	0.340	606	357	0.532	527	527	0.740
A. Reid	KC	576	375	0.712	565	421	0.668	543	450	0.618
A. Lynn	LAC	597	366	0.486	512	399	0.651	583	419	0.653
S. McVay	LAR	632	401	0.547	568	459	0.679	518	454	0.708
C. Gailey	MIA	550	418	0.281	604	448	0.621	548	448	0.608
G. Kubiak	MIN	466	476	0.668	588	393	0.465	566	457	0.340
J. McDaniels	NE	620	447	0.814	574	478	0.667	587	448	0.738
S. Payton	NO	581	405	0.668	519	471	0.699	536	444	0.680
J. Garrett	NYG	597	449	0.671	527	439	0.527	493	480	0.538
A. Gase	NYJ	521	383	0.349	455	371	0.326	602	360	0.311
G. Olson	LVR	523	437	0.334	568	459	0.679	527	527	0.740
D. Pederson	PHI	613	454	0.550	599	398	0.531	564	473	0.738
Fichtner	PIT	510	395	0.472	689	345	0.601	590	437	0.658
B. Schottenheimer	SEA	517	481	0.510	427	534	0.622	515	395	0.448
K. Shanahan	SF	478	498	0.737	532	423	0.361	607	408	0.414
Arians	TB	630	409	0.512	598	410	0.383	646	399	0.584
A. Smith	TEN	448	445	0.613	437	456	0.514	496	443	0.462
Turner	WAS	633	386	0.238	563	416	0.491	588	380	0.537
		59.334	48.151	0.521	60.786	44.427	0.555	38.978	41.042	0.608
				0.156			0.113			0.136

 

So now we have a stack of raw data for coaching “tendencies”, I scare quote that because players, and situations affect these numbers to some extent so we can see general trends but there is still more work to do. My next step is to adjust for the season. Did a team have a lot of rushes or passes, why was that? Did they run a lot because they grossly outscored opponents? Did they pass a lot because they were always behind? Or did they go against trend and pass a lot while winning a lot (looking at you Josh McDaniels). So that’s what this next step is trying to normalize the numbers for each team to add context about who is actually pass happy or run heavy outside of situation.

 

OC or HC	Tm	Plays Ave	STDev19	Pa	Ru	STDev18	Pa	Ru	STDev17	Pa	Ru
Kingsbury	ARI	974	-0.893	501	439	0.000	548	430	0.000	548	448
D. Koetter	ATL	1018	-0.312	666	377	-1.311	545	447	-1.364	552	446
G. Roman	BAL	992	1.905	553	504	-0.204	462	501	-0.560	443	532
Daboll	BUF	992	0.585	548	437	-2.137	369	563	1.362	603	426
J. Brady	CAR	985	0.000	559	427	1.280	597	414	0.528	557	422
Nagy	CHI	968	-0.373	558	413	1.458	601	403	0.069	546	402
Taylor/Callahan	CIN	1000	-1.582	522	461	1.102	635	410	0.734	547	424
Van Pelt	CLE	978	0.944	522	431	-0.204	594	366	0.883	561	465
K. Moore	DAL	995	0.966	654	402	-0.250	512	450	-0.517	473	501
P. Shurmur	DEN	978	-1.164	538	418	-1.064	518	401	0.883	561	465
D. Bevell	DET	971	-0.939	515	452	0.022	556	408	1.331	619	348
LaFleur	GB	950	0.552	606	384	-0.368	415	472	0.734	547	424
B. O'Brien	HOU	973	-0.206	522	444	0.744	551	439	-1.873	452	525
Reich	IND	1024	-0.260	498	484	0.692	686	377	0.957	601	434
Marrone/DeFilippo	JAC	998	-1.166	520	445	-0.204	594	366	0.972	565	487
A. Reid	KC	977	1.225	649	316	0.998	626	377	0.069	546	447
A. Lynn	LAC	959	-0.225	584	377	0.851	564	361	0.328	596	406
S. McVay	LAR	1011	0.165	642	393	1.102	635	410	0.734	547	424
C. Gailey	MIA	1005	-1.545	458	492	0.588	640	422	0.000	548	448
G. Kubiak	MIN	982	0.944	522	431	-0.798	540	428	-1.970	489	538
J. McDaniels	NE	1051	1.885	732	356	0.996	635	434	0.953	624	409
S. Payton	NO	985	0.944	637	360	1.280	597	414	0.528	557	422
J. Garrett	NYG	995	0.966	654	402	-0.250	512	450	-0.517	473	501
A. Gase	NYJ	897	-1.106	455	436	-2.027	332	461	-2.186	517	450
G. Olson	LVR	1014	-1.205	452	495	1.102	635	410	0.972	565	487
D. Pederson	PHI	1034	0.183	624	445	-0.209	586	407	0.957	601	434
Fichtner	PIT	989	-0.316	491	410	0.409	714	327	0.367	604	422
B. Schottenheimer	SEA	956	-0.070	513	484	0.592	463	508	-1.181	469	443
K. Shanahan	SF	982	1.389	560	431	-1.719	428	499	-1.426	551	467
Arians	TB	1031	-0.060	626	412	-1.529	505	478	-0.175	639	406
A. Smith	TEN	908	0.591	483	417	-0.368	415	472	-1.074	454	487
Turner	WAS	989	-1.823	525	474	-0.571	528	441	-0.521	568	401

 

Which brings us to the part that actually ends up in my projections. The team specific totals. I start off with 2019 being weight 50% more than 2018, which is weighted 100% more than 2017 to weight towards more recent data for each coach. And then adjust rushing and passing totals based on the Vegas win totals to try and best represent situations for context. Then this becomes much more art then science. Any adjustments based on standard deviations (which I prefer) are going to adjust too far at times because we’re dealing with a sample size of 32 teams, and as I’ve said specific situations, injuries, and talent are going to skew year by year so I take those raw numbers and work them a bit more based both on my own opinions of teams and rosters and as I make my full blown projections based on what I’m seeing with individual players and position groups. For example, I had Arizona with a greater run/pass split but with that split I couldn’t get Drake to the number of rushes I wanted so I skewed them a little more towards the run, which Kingsbury has actually skewed a little more heavily towards both in his first year in Arizona, but also at Texas Tech where he wasn’t quite as Pass happy as his mentor Mike Leach.

 

OC or HC	Tm	20Pa	20Ru	20Tot	VegW	VegL
Kingsbury	ARI	587	424	1011	7.5	8.5
D. Koetter	ATL	638	392	1030	7.5	8.5
G. Roman	BAL	523	487	1010	11.5	4.5
Daboll	BUF	529	455	984	9	7
J. Brady	CAR	648	383	1031	5.5	10.5
Nagy	CHI	558	416	973	8.5	7.5
Taylor/Callahan	CIN	612	398	1010	5.5	10.5
Van Pelt	CLE	557	419	976	8.5	7.5
K. Moore	DAL	572	451	1023	9.5	6.5
P. Shurmur	DEN	579	421	1000	7.5	8.5
D. Bevell	DET	592	401	993	6.5	9.5
LaFleur	GB	548	462	1010	9.5	6.5
B. O'Brien	HOU	530	450	979	8	8
Reich	IND	565	448	1013	8.5	7.5
Marrone/DeFilippo	JAC	609	366	975	5.5	10.5
A. Reid	KC	571	418	989	11.5	4.5
A. Lynn	LAC	599	364	963	7.5	8.5
S. McVay	LAR	597	421	1018	9	7
C. Gailey	MIA	587	425	1012	6	10
G. Kubiak	MIN	536	421	957	9	7
J. McDaniels	NE	588	451	1039	9	7
S. Payton	NO	548	434	982	10	6
J. Garrett	NYG	559	417	976	6.5	9.5
A. Gase	NYJ	501	402	903	7	9
G. Olson	LVR	563	445	1009	7.5	8.5
D. Pederson	PHI	597	436	1033	9.5	6.5
Fichtner	PIT	622	405	1027	9.5	6.5
B. Schottenheimer	SEA	507	471	978	9	7
K. Shanahan	SF	512	463	975	10.5	5.5
Arians	TB	592	425	1017	9	7
A. Smith	TEN	468	430	898	8.5	7.5
Turner	WAS	582	418	1000	5.5	10.5

 

Then it’s a relatively straight forward process I look at rush and target shares from last season, try and adjust for injuries, players that came in or left the team and give each role on  each team a share of the offense. Below I have the QB breakdowns for each team as an example.

 

Tm	QB1%	QB2%	QB3%	QB1Ru%	QB2Ru%	QB3Ru%
ARI	0.975	0.020	0.005	0.235	0.010	0.005
ATL	0.950	0.045	0.005	0.090	0.010	0.005
BAL	0.925	0.070	0.005	0.295	0.035	0.005
BUF	0.915	0.080	0.005	0.225	0.005	0.005
CAR	0.950	0.045	0.005	0.100	0.010	0.005
CHI	0.700	0.295	0.005	0.095	0.035	0.005
CIN	0.900	0.085	0.015	0.100	0.010	0.015
CLE	0.990	0.005	0.005	0.070	0.005	0.005
DAL	0.995	0.005	-	0.115	0.005	-
DEN	0.980	0.015	0.005	0.125	0.010	0.005
DET	0.980	0.015	0.005	0.100	0.015	0.005
GB	0.990	0.005	0.005	0.105	0.015	0.005
HOU	0.960	0.035	0.005	0.200	0.010	0.005
IND	0.975	0.020	0.005	0.035	0.015	0.005
JAC	0.950	0.045	0.005	0.175	0.010	0.005
KC	0.950	0.045	0.005	0.135	0.015	0.005
LAC	0.645	0.350	0.005	0.105	0.060	0.005
LAR	0.990	0.005	0.005	0.075	0.005	0.005
MIA	0.600	0.395	0.005	0.115	0.085	0.005
MIN	0.955	0.045	-	0.065	0.005	-
NE	0.900	0.095	0.005	0.120	0.010	0.005
NO	0.920	0.070	0.010	0.030	0.055	0.065
NYG	0.950	0.045	0.005	0.150	0.010	0.005
NYJ	0.925	0.070	0.005	0.100	0.010	0.005
LVR	0.845	0.150	0.005	0.050	0.015	0.005
PHI	0.975	0.020	0.005	0.130	0.010	0.005
PIT	0.975	0.020	0.005	0.075	0.005	0.005
SEA	0.995	0.005	-	0.150	0.005	-
SF	0.990	0.005	0.005	0.090	0.005	0.005
TB	0.995	0.005	-	0.025	0.005	-
TEN	0.950	0.045	0.005	0.125	0.015	0.005
WAS	0.950	0.045	0.005	0.115	0.025	0.005

 

From there I compile INT:TD ratios for QBs, Yards per rush attempt data for QBs/RBs/WRs (with enough data), Yards per target data for RBs/WRs/TEs and crudely adjusted college data for rookies and young players with no meaningful pro data. I go through each team tweaking and adjusting yards per a bit, looking for touchdown outliers, and anything that may be a bit off, and finally… You got yourselves a set of projections. This current set of which was done before players started opting out.

Hope you've enjoyed this look into my damaged fantasy football psyche. Who are we kidding just  a look into the way my brain works. I'm sorry, and have fun picking this apart.