The Z Files: Are In-Season Park Factors Reliable?

Written by

Updated on July 22, 2016 8:03AM EST

Fantasy enthusiasts, be it in seasonal or DFS formats, look for any edge they can garner. Buoyed by the growth of DFS, in-season park factors have become a popular means to get ahead of the curve. The theory is while most are relying on an average of the previous three seasons as the park factor, incorporating how the venue is playing this season is a more accurate representation of what to expect.

On paper, this is a great idea. But sometimes great ideas aren't so great after all. How actionable are in-season park factors? At what point of the season, if at all, are they reliable? Perhaps this assumed edge isn't really an advantage and is in fact hurting, not helping.

Before we embark on that study, below is a chart with the 2013-2015 three-year index for runs, left-handed homers and right-handed homers along with the 2016 index through games played on Wednesday, July 20.

	3-Yr R	2016 R	3-Yr HR LHB	2016 HR LHB	3-Yr HR RHB	2016 HR RHB
ARI	106	119	109	134	94	126
ATL	94	108	86	84	95	67
BAL	107	101	137	83	108	109
BOS	107	106	70	75	100	104
CHC	102	80	97	66	124	98
CHW	98	101	109	146	113	122
CIN	102	93	131	111	117	124
CLE	103	122	105	109	103	109
COL	140	162	115	154	121	125
DET	101	100	95	149	95	108
HOU	100	86	122	77	112	95
KC	104

	3-Yr R	2016 R	3-Yr HR LHB	2016 HR LHB	3-Yr HR RHB	2016 HR RHB
ARI	106	119	109	134	94	126
ATL	94	108	86	84	95	67
BAL	107	101	137	83	108	109
BOS	107	106	70	75	100	104
CHC	102	80	97	66	124	98
CHW	98	101	109	146	113	122
CIN	102	93	131	111	117	124
CLE	103	122	105	109	103	109
COL	140	162	115	154	121	125
DET	101	100	95	149	95	108
HOU	100	86	122	77	112	95
KC	104	123	88	95	82	80
LAA	92	99	85	109	90	109
LAD	90	78	115	138	101	83
MIA	100	88	64	128	78	75
MIL	107	103	133	134	123	130
MIN	104	105	90	77	96	102
NYM	87	104	124	105	82	138
NYY	102	99	129	135	126	158
OAK	95	86	75	85	90	74
PHI	102	76	110	124	140	95
PIT	94	106	92	119	73	65
SD	93	104	110	70	110	108
SEA	90	93	98	116	91	108
SF	88	94	59	41	76	74
STL	97	86	87	88	90	82
TB	96	83	109	85	83	84
TEX	106	114	101	99	95	106
TOR	102	104	114	109	121	8
WAS	103	94	80	82	88	91

Let's break things down in terms of the percent change of 2016 from the three-year averages.

There are seven venues playing at least 10 percent more pitcher friendly this season: Citizens Bank Park, Wrigley Field, Minute Maid Park, Tropicana Field, Dodger Stadium, Marlins Park and Busch Stadium. The initial pair are playing over 20 percent more pitcher friendly.

There are eight venues playing at least 10 percent more hitter friendly: Petco Park, Chase Field, PNC Park, Turner Field, Coors Field, Kauffmann Stadium, Progressive Field and Citi Field, the last of which is exactly 20 percent different.

Eight venues are 10 percent or more less homer friendly for righty swingers: PNC Park, Minute Maid Park, O.Co Coliseum, Dodger Stadium, Wrigley Field, Rogers Centre, Turner Field and Citizens Bank Park. The last four are at least 20 percent worse.

Seven venues are at least 10 percent more home run friendly for right-handed batters: Citi Field, Chase Field, Yankee Stadium, Angels Stadium, Safeco Park, Comerica Park and Globe Life Park. The first four are at least 20 percent better.

Nine venues are at least 10 percent less home run friendly for left-handed hitters with the last six over 20 percent: Target Field, Great American Ballpark, Citi Field, Tropicana Field, AT&T Park, Wrigley Field, Petco Park, Minute Maid Park and Camden Yards.

Eleven venues have increased left-handed power by at least 10 percent, the first eight checking in at least 20 percent: Marlins Park, Comerica Park, U.S. Cellular Field, Coors Field, PNC Park, Angels Stadium, Chase Field, Dodgers Stadium, Safeco Field, O.Co Coliseum and Citizens Bank Park.

Let's summarize as there's something very telling in this data:

	Runs	RHB HR	LHB HR
10 pct plus higher	7	7	11
20 pct plus higher	1	4	8
10 pct plus lower	8	8	9
20 pct plus lower	2	4	6

Note how the variance follows the size of each sample. Runs has the largest sample since it's not parsed into handedness with RHB HR having a larger sample than LHB HR. Runs and RHB HR have the same number of parks that are 10 percent different but there are more RHB HR over 20 percent. Runs and RHB HR each have half the total parks with a 10 percent difference while the LHB HR index shows a difference for 20 venues, the majority with factors differing by more than 20 percent. This suggests the variance could be sample-size related rather than a true reflection of any 2016 changes.

It's interesting to note the huge difference in Marlins Park, especially for left-handed power. Remember, they lowered and moved the right-center field fences in. Spoiler alert: while some of the jump to 128 from 64 is likely real, the sample is too small to comfortably call The Aquarium (that really should be the park's nickname) a very favorable venue for lefty pop.

OK, let's get into the meat of this discussion and address the question of in-season sample size reliability. I took each team's game logs from 2013-2015 and dumped them into a program that calculates first half and second half park indices along with a control set. Actually, the distinction isn't exactly first half and second half but first 41 games played in each venue versus final 40. The control set measured the park index of the odd numbered games and the even numbered, yielding the same 41 to 40 comparison.

The actual data goes beyond the scope of this discussion but the results were clear. The correlation between the first half/second half data to three-year average was the same as that of the odd/even data – essentially nothing. In other words, identifying the outlier parks above was mostly an exercise in futility.

Or was it? Knowing the factors others are using is always helpful, especially in DFS if you're looking for a contrarian angle. Let's circle back to the LHB HR index in Marlins Park. At some point, people are going to discover, if they haven't already, how much the index has increased. There really aren't many left-handed Fish hitters to take advantage (Christian Yelich, Derek Dietrich, Justin Bour when he returns) but I'll wager someone will highlight an opposing lefty swinger as a strong play based on the huge upgrade in lefty power. I'm fading that hitter unless the spot is too strong to ignore, like Bryce Harper facing Jose Urena. It's more not taking the bait on an average lefty – Cody Asche, for example. It's really easy to envision a scenario where Asche is on someone's value play list because Marlins Park is perceived to be more generous to lefties.

This study begs a follow-up which I'm in the process of undertaking. How many games are needed to show correlation? The plan is to keep increasing the sample by ten games to find the spot where correlation exists, supported by a control set of like size. I'll use the individual game logs from a full season and compare the respective factors after 90, 100, 110, 120 games etc, in the hopes of finding when the in-season index is actionable. Of course, we're still working blind until then, but at minimum we'll be ahead of the game for the most important part of the season: DFS championship tournament time and down the stretch in seasonal formats.

For what it's worth, this code can be tweaked to do a similar study on the in-season reliability of weighted on base average, strikeout percent or any of the other metrics commonly employed for in-season analysis.