There are three kinds of lies: lies, damned lies and statistics.
Mark Twain wasn't referring to fantasy baseball when he discussed the precept in his autobiography, but if that was his philosophy, he'd be behind the curve competing in today's climate. Fantasy baseball is obviously a game of numbers with more information available today than ever, and we're only beginning to scratch the surface of the next level of statistical analysis generated by Statcast data and the like.
On the other hand, Twain has a point. It may not be the one intended, but if statistics aren't completely understood and applied correctly, the result could be a misleading or erroneous conclusion. This discussion will elucidate some of the misconceptions and faulty utilization of common theories, principles and metrics pertaining to statistical analysis.
Regression
Admittedly, this is a semantical argument, but it's a shame regression has become synonymous with playing worse and not connoting a more hardcore statistical meaning. When an analyst states a player will regress, it would be helpful to comprehend the exact implication.
Given, true regression can go in either direction but most of the time it refers to declining performance. Technically, skills can regress, but often the phrase "regression to the mean" is used, which should refer specifically to elements of performance out of the player's control. Of course, this is just a fancy way of labeling luck.
It shouldn't be a slight on a player to contend an element of his game will regress. That is, he can maintain the same basal skill level, but outcomes previously positively influenced by luck should regress to the mean. He's not playing worse; the outcomes are no longer embellished via good fortune.
This is an important distinction when attempting to forecast future performance. Does the analyst believe the player's skills will drop, or will his luck run out? Sometimes it's deficient to discern when regression is bandied about with no explanation.
In a perfect world, metrics such as BABIP (batting average on balls in play), HR/FB (home run per fly ball) and LOB% (left on base percent) would regress to the mean. Whereas, a pitcher's velocity and batter's hard-hit rate or batted-ball distribution would revert to previous levels. Again semantics, but regress and revert should have discrete meaning, shaping the ensuing player evaluation.
If afforded the opportunity, always request the person decreeing regression clarify the nature of their contention. This way, you know exactly what to look for when doing your own inspection of the player.
Gambler's fallacy
Luck doesn't always even out. The classic example of gambler's fallacy is expecting a coin flip landing as heads be followed by tails to even things out. Every flip is independent of the previous, with a 50% expectation it is heads. Going back to regression, good or bad luck shouldn't be expected to reverse. The proper expectation is neutral luck when projecting future performance, regardless of what's transpired to date.
Obviously, ballplayers are humans -- not flat, round, minted metal -- so there are more factors than just luck. Skills are thought to be static, but they're fluid. Or better explained, the result of a player's skill set can be influenced by factors such as quality of opponent, venue, etc. A pitcher facing potent lineups in hitters' parks for a stretch are likely to see their number suffer. However, they'll soon draw lesser lineups in pitching venues with much better results. The basal skills (K%, BB%) will be different for each series of outings.
In this regard, a pitcher who is expected to post a 3.50 ERA but is sitting at 3.75 after a rough stretch will often come back to 3.50 after the string of softer outings. This is different than luck evening out.
The classic example is looking at a career .300 hitter halfway through the season. If he's hitting .275, some will say to expect .325 the rest of the way, while others will argue .300 is correct. In fact, neither is right. The proper projection separates the skill from happenstance. Unfortunately, this is easier said than done.
Delineating luck from skill when putting a player under the microscope is a daunting task. One of the chief purposes of the Statcast data is aiding to distinguish the two. As of now, the bulk of the data is descriptive as opposed to predictive. As more is collected, more actionable conclusions will emanate.
Park factors
In general, venues are considered pitcher or hitter friendly. Usually, the perception dovetails into whether the park helps or hurts power hitters. In fact, there are several instances of a venue aiding power while suppressing runs and vice versa. Past and current park indices can be found HERE.
A common mistake when incorporating park factors into analysis is double-dipping. That is, the favorable home run index for Globe Life Park already bakes in the high temperatures usually found in Arlington. While it may be hotter than usual, be careful not to overcompensate for the heat, since the park factor already includes similar days from previous summers.
BABIP
The formula for BABIP excludes homers. Some deem this a flaw. It's not, but it is necessary to keep BABIP in context. In general, BABIP is looked at as a metric measuring luck. Over time, batters generate their baseline with anything over being labeled lucky and under assumed unlucky. The same is true for pitchers, but they possess a much narrower expected range, clustering around the league norm.
The pitfall with BABIP is the omission of homers can be misleading in certain instances, especially a low or unlucky BABIP. This is amplified this season with the explosion of homers. As suggested, a low BABIP for a hitter is anticipated to regress to the mean, which is the batter's career level. If the player is hitting more homers, some batted balls previously incorporated into the formula will no longer be accounted for, as they left the yard. In past seasons, these homers were doubles and fly ball outs. Obviously, it depends on how many of each occurred, but since a hit affects the numerator more than the out influences the denominator, doubles and outs becoming homers usually torpedo the player's BABIP.
The mistake is expecting a player enjoying a power surge with a low BABIP to have their performance get even better as their BABIP regresses to its career mark. In some cases, the batter may be a victim of misfortune. Even then, the landing point after regression won't be the career mark but lower, since hard-hit doubles and fly outs don't contribute to the current BABIP.
wOBA
Weighted on-base average (wOBA) has become a catch-all metric to describe a player's fantasy potential. Granted, it's superior to something like OPS, but there are shortcomings that need to be kept in context.
For those unaware, wOBA is a souped-up version of OBP with coefficients for each component of the formula. These coefficients emanate from the run-scoring matrix, weighting each in proportion to how it contributes to a run. It happens that wOBA correlates pretty well to fantasy production. However, it's not perfect.
The primary shortcoming is wOBA doesn't reflect a hitter's stolen-base acumen, something obviously paramount. That said, this isn't sufficient to categorically dismiss utilizing wOBA when making pickup and lineup decisions, only to keep in mind that stolen-base production is shortchanged.
In addition, wOBA isn't park corrected, That's OK, neither is batting average, on-base average, etc. However, they're not used specifically to rank or rate players against each other in a global fantasy baseball sense. As an example, the wOBA from a Miami Marlins hitter is more impressive than the same level from a Colorado Rockies batter, since Marlins Park deflates offense while Coors Field embellishes it. If each is playing on the road, the Marlins' wOBA is probably higher then the Rockies' players.
This isn't the only determinant as opponent and venue are factors. The key is the players aren't starting on equal footing since their road wOBA are not the same. Similarly, if both are at home, the Colorado batter is favored, since his home wOBA likely exceeds that of the Miami batter in South Beach.
Ground ball rate
The ground ball rate for a pitcher is often referenced as a skill, implying higher is preferred. While more grounders result in fewer homers, they also lead to more base hits. Double plays help mitigate the added base-runners, but not totally. In a fantasy sense, an extreme ground ball pitcher has a low ERA relative to his WHIP, while a fly ball pitcher sports a high ERA compared to his WHIP. It isn't always the case the ground ball pitcher is better (higher skilled).
The more apropos approach is viewing ground ball rate as a trait to be woven in with other factors. For instance, a ground-ball pitcher is more useful in a homer-friendly venue. It may be more beneficial to be a fly ball guy in a park with an expansive outfield since there's less of a chance a fly ball will land safely than a ground ball getting through the infield. Defense also matters, as do the strikeout and walk rates of the pitcher.
When making lineup decisions for pitching, it's necessary to know whether the pitcher is more of a ground ball or fly ball guy. That's not the end of the story. Determine if that's a benefit or detriment when accounting for the associated aspects.
Exit velocity
The key is the exit velocity often relayed as average, encompassing all types of batted balls. Sure, individual hits have their specific exit velocity reported, but leader boards rank average exit velocity.
Your mechanic doesn't ask for your average miles per hour to decide if it's time to change your brakes. Some do most of their driving in suburbs with lower speed limits while others take the freeway, combining cruising at higher speeds with bumper to bumper traffic. Both may average 38 mph, but one is harder on their brakes. A such, more information is needed to drive conclusions with respect to exit velocity.
Granted, the data isn't readily available, at least not yet to the general public, but realize the average velocity of a ground ball is slower than that of a fly ball, which in turn is below a line drive. Thus, looking at a player's average exit velocity and relating that to BABIP could be misleading. The higher exit velocity may not correlate to a higher BABIP, it also depends on the hit distribution, not to mention the launch angle.
An example of a player with interesting component exit velocities is Derek Dietrich, as his ground ball exit velocity is below league average, while that on fly balls and line drives is above league norm. The assumption is if a player hits more grounders, their power will wane but their BABIP, hence average, will climb. This may not be the case for Dietrich, as there's something about his swing mechanics resulting in slower hit ground balls than average.
Another player with an intriguing skill set is Christian Yelich. Just as it seemed there aren't any more accolades possible describing the Brewers' fly-chaser, looking at his exit velocity on the different batted balls, he strikes grounders with more force than some exert on line drives. That's just not fair.
Spin rate
Like exit velocity, a pitcher's average spin rate is often the means by which they're ranked. However, different pitches have varying average spin rates. Here's the breakdown: curve > slider > cutter > fastball > sinker > change > splitter. In addition, pitches thrown up in the zone spin less than those lower.
With the offerings at the front of the list, more spin is better. However, ranking by average spin and assuming effectiveness dovetails is a mistake since repertoire matters. The best example is Hyun-Jin Ryu, as he doesn't rank high in terms of average spin rate, as he relied on a higher proportion of changeups and sinkers as compared to curves and sliders.
Synopsis
While it's obligatory to keep up with the advanced metrics and next level analysis, it's necessary to make sure you're doing it properly. Before incorporating a new principle or metric into your evaluation process, make sure you thoroughly comprehend when and how to use it, as well as when not to utilize it.
Unfortunately, many individuals responsible for disseminating information don't have a true grasp of the concepts, As such, the onus is on you to conduct the needed due diligence so your application doesn't render more harm than good.