The Z Files: So Much To Think About

As many of you know, I do my own projections. Everyone has their own style and approach, but mine is mostly formulaic. This offseason they'll require more tweaking, with frequent instances where I'll have to override the system. The quirky nature of the 2020 season presents numerous challenges for the fantasy prognosticator. Here are some of some of the elements I've been thinking about.

How much weight should 2020 carry?

I don't want to turn this into a "how the sausage is made" piece, but a little background is needed. The player's baseline comes from a weighted average of previous performance, which in my process is distilled to skills. Different methods incorporate varying numbers of past seasons: I use three, with the most recent weighted more than a year ago, which is weighted more than two seasons ago.

In general, I don't adjust for playing time. Here is an example, using expected homers and a 5:3:1 weighted average. Expected homers translates the player's actual output to an expected amount based on the underlying metrics. Let's say a player's homers translated into the following number of expected homers:

2019: 30 xHR in 600 PA

2018: 27 xHR in 500 PA

2017: 15 xHR in 200 PA

((5 x 30) + (3 x 27) + (1 x 15))/((5 x 600) + (3 x 500) + (1 x 200)) = .052 xHR/PA

If this player was projected for 650 PA this season, his homer expectation would have been

.052 x 650 = 34 homers.

Now let's say in 2017, the player had 30 xHR in 400 PA, doubling output and playing time. The 2020 projection in 650 PA would have been 35 homers.

The issue with 2021 projections is this past season was a little more than a third of a normal campaign. Using the above math will penalize a player displaying true skills growth. The caveat is 60 games may not be sufficient to characterize that a change as real. Some may rely on stabilization points to help discern the extent of the change, but as I've discussed previously, this is a misapplication of stabilization numbers.

Something I've been pondering is prorating the 2020 season to 162 games (multiplying everything by 162/60) and using that to derive xSTATS and plug into my usual weighted average formula. The problem is that, as suggested, doing this takes for granted the skill levels would have been maintained over a regular number of games. The way to deal with it is to adjust the weighted average coefficient for the skills I'm not confident are real.

Another way to handle this is to leave the 2021 stats as is, but again adjust the weighted average of some skills to reflect how much influence I feel they exert.

My guess is both will yield close to the same result when distilled to per PA or per IP. Ultimately, it will come down to which pathway is easier to incorporate. Regardless, there will be more typing in a hard number over an Excel formula than ever before.

Skill changes to trust

As I mentioned, using stabilization points to strictly gauge the extent of a skill change is flawed. That said, the hierarchy of elapsed time for the various stabilization points is likely pertinent. A change in skill said to stabilize over a fewer number of events (PA, IP, batters faced, etc.) probably has a better chance to be real, just not as defined by the number of events as some contend. What this means is while I'll feel confident one skill change is more real than another, and adjust the weight accordingly, the extent is more subjective than I prefer. It's a guess, and the purpose of a formulaic-driven spreadsheet is to avoid guesses.

Another factor is that as of this writing, I'm more apt to pay heed to an improvement than penalize a decline. Anecdotally, there are many reasons (excuses?) a player might have had a disappointing 2020 campaign. Some have complained about the lack of in-game video access. Others miss the energy provided by live fans. There's talk of a different clubhouse feel, especially at home with social distancing and the like. The start-up of the regular season, after a long delay with just a few weeks of summer camp, may have messed with some players' abilities to optimize their readiness. Keep in mind players are human, with family back home dealing with the effects of COVID-19 in day to day life, and maybe this was too much of a mental grind. We'd like to believe professional athletes give it their all regardless of the scenario but, let's face it, they don't. Even with the expanded playoffs, a lot of players knew they were just playing out the string from the jump, without fans to drive their performance for a good portion of the season until the malaise settled in.

On the other hand, there's just one explanation for true skill improvement: the player got better. Focusing on skills leaves the luck aspect out of the equation.

If a player improved, this should be adequately reflected in the baseline. As stated, I'm leaning towards crediting a positive skill change more than docking a slide. This feeds into the subjective nature of massaging the weighting coefficients discussed above.

Games played in geographical zones

The season was contested under the umbrella of Major League Baseball. However, save for the instances of players released and picked up or traded, in essence there were three distinct leagues. Gameplay was confined to an East League, Central League and West League. A pitcher from each "league" may have posted a 23 percent strikeout rate, but they may not be equal, or reflect the same level of prowess since the pool of batters faced is different for each. Obviously this is always the case, but in a 162-game season with a normal schedule, quality of opposition is more level. This season, there are no doubt some batters and hurlers who faced significantly easier or harder competition.

Before embarking on 2021 projections, I'll measure the quality of competition for each region and likely make a global adjustment for each zone. I'll probably go more granular and investigate whether any player was affected more or less than others in the same region and adjust accordingly. Yup, more manual tweaking. Good times.

Park factors

This is going to be a total mess. In fact, my current plan is to all but dismiss 2020 indices.

First and foremost, my research shows a two-month slice of a standard season generates highly variable park factors. To put this in perspective, park factors are conventionally expressed as a three-year average, since one season isn't sufficient to flesh out all the bias. If three years are needed for a reliable index, a two-month sample is certainly untrustworthy.

At this point, I could drop the mic and move on. However, there is also a bunch of chaos within this particular two-month sample independent of the inherent variability.

Park factors are subject to the same geographical restriction as the players. That is, a park playing neutral for a given metric in the East likely played differently than a neutral venue in the Central or West. The best way to explain this is: say the regions were set up so one encompassed the 10 most homer-friendly parks, one the 10 least friendly, and the third region the 10 parks in the middle. Since the parks are compared within each region, the least homer friendly of the first group will rate as negative for homers, since it's compared to nine other places better for the long ball. Similarly, looking at the 10 worst parks, the friendliest of those will come out as positive for homers, since it is being ranked against nine parks even more of a detriment to long balls.

A way to help account for the regional nature of park factors would be to determine the averages for the three zones, and use those to adjust. I haven't undertaken this yet, but plan to and will report my findings.

There are a couple other nuances that can be dealt with but contribute to the "waste of time" nature of looking at 2020 factors. Due to some necessary scheduling, many clubs served as the home team in their opponent's yard. Not to mention, the Blue Jays played in Buffalo and will hopefully be back in Rogers Centre in 2021.

There is also a brand new venue (Globe Life Field) and two that underwent renovations before the season (Marlins Park and Oracle Park), requiring new factors. Not to mention, it was revealed Fenway Park, Citi Field and T-Mobile Park installed humidors and thus have factors in need of a refresh. Unfortunately, the 2020 stats are by no means suitable to make the proper adjustments.

Having so many seven-inning affairs is another issue for indices using games played as the denominator. The runs index is a prime example. Even if everything else wasn't a factor, the runs scored in all doubleheader games would need to be prorated to nine innings.

Will there be a universal designated hitter?

There are still a handful of games left, but the results are probably not going to change. The ERA in the National League is higher than in the American League. Similarly, batters in the senior circuit have been more productive than their cross-league counterparts. Crafting player baselines, especially for National League pitchers, depends on the presence or removal of the universal designated hitter. Some may hand-wave the difference away and say it doesn't matter. Trust me, it is HUGE. Reading the tea leaves, I suspect the universal designated hitter is here to stay, so I'll adjust my little black box accordingly.

Minor Leagues

This is another gigantic gray area. With only camp reports to go by, what should be used for a prospect's 2020 stats? Normally, an MLE (major league equivalency) is utilized to translate minor-league numbers into something applicable to MLB projections. We don't have them for this season. The result will be that the error bars for prospects will be even wider than normal.

Pitcher workloads

Regardless of how you feel about "The Verducci Effect", the innings count is obviously low across the board. It's already tough enough to project starts and innings with the manner teams are handling their rotation, but now a light 2020 workload needs to be factored in. In an ironic twist, starters working deep into the playoffs may have an advantage next year with more innings under their 2020 belts. Normally, the course of action is to be wary of a guy tacking playoff innings onto a regular-season total.

Summary

I'm no doubt leaving elements off while I'll unearth others as I dig into the process early next week. If you use projections to help guide your drafts, it is incumbent to find out how your source handles the described conundrums. If you disagree with the process, you'll need to make the required adjustments.

This is going to be the most challenging offseason ever. It's a good thing I welcome and embrace challenges.