The Z Files: What Is a Projection?

Before heading to the Fall Stars Game at the recent First Pitch Arizona (FPAZ) conference, a couple of guys approached me in the hotel lobby where I was asked about importance of projections. They wanted an expert's opinion, but since they couldn't find one, they settled for me.

Their conversation stemmed from each drafting a team the previous evening. With no ballgames on the Friday night preceding Saturday's Fall Stars game, FPAZ sponsors several draft and auction leagues of various formats. These two participated in one of the National Fantasy Baseball Championship (NFBC) draft-and-hold leagues. One remarked to the other, "I hate my team, drafting without projections sucks." The other countered with, "I love my team, but I never use projections anyway." From there, the two began arguing about the importance of projections before asking me to mediate.

The first thing I did was explain while some may not use spreadsheet-driven, formulaic projections, everyone has player expectations, so it's more semantics. One person's projection is another's Zen expectation. Both nodded in agreement.

I was then asked, "Which is better?" I responded with a question, "What's a projection?"

Paul (not his real name), the guy that uses them replied, "They're how the person feels the player will perform." Norm (also made up), the one that doesn't use projections chuckled, "Aren't they just guesses anyway?"

I suggested, "Projections are more like educated guesses, based on history. Some use a Zen-like approach, doing each by hand, looking at the player's history and simply deriving what they expect from that. Those likely look beyond the surface stats, incorporating underlying skills, but the genesis is mostly gut. Others begin with a three-year weighted average of some sort, perhaps including an aging curve, but not necessarily. They then massage the average up or down, again mostly feel."

"What do you do?" Paul inquired.

"My method begins with skills and builds out from there." I didn't get into the gory details but did explain, "I start with a three-year weighted average neutralized for age and park effects, then distilled to per plate appearance. Multiplying by the number of projected plate appearances yields the final projection."

I noted, "I use my own hitting formulas for expected BABIP (batting average on balls in play) and home runs per fly ball, as well as runs, RBI and steals, regressing against the player's actual stats. Similarly, I've developed some pitching algorithms to account for the aspects of a projection requiring regression."

The guys again nodded, but this time with a somewhat glazed look in their eyes. Wanting to turn the conversation from theoretical to practical, I asked, "Regardless of where they come from, what do you think a projection actually is?"

Now I'm getting a couple of blank stares. "OK, here's the deal," I went on, looking to put things in better perspective.

"There are several different views on what a projection truly represents, beyond how a player will perform. One of the most common ideas is pretending the season is played a gazillion times. Projections are then the average of all the seasons. The notion being, in a large enough sample, all possible outcomes would be covered."

The blank stares morphed back to approving nods, now we're getting somewhere.

Norm opined, "The thing I like least about projections is no one takes a chance. Someone coming off a good season is always projected to regress while a guy is always projected to improve after a bad year."

"Yeah, I hear this a lot, but that's not what a projection is supposed to be. Think about what I just said. A projection encompasses all plausible outcomes. Yes, there's a chance the average of these is better or worse the following season, but you have gravity, and history working against you. I've done a study for the past 15 or so seasons. Between 55 and 70 percent of players perform at a lower level than the previous year. This limits the probability of a better season. It doesn't mean it can't happen. In fact, one of the studies I want to do is finding a common thread between the 30 and 45 percent of improving players. Plus, luck generally normalizes so numbers inflated or deflated from fortune or misfortune regress to the mean. There's nothing wrong with a bold prediction. It's one of the most common topics among fantasy pundits. Heck, we've had sessions on the subject at the conference. We used to devote time to identifying "Black Swans". One year, one of the speakers asked for everyone to submit their "NFW" predictions and he'd review them the following year. There's a big difference between a projection, rooted in research and a prediction, more an off-the-cuff feeling. Norm, what you're looking for are bold predictions. There's a place for that, but I'd be much more interested in the reasoning than the prediction itself."

I followed up with, "Perhaps the biggest mistake made in drafts, one that yours truly makes too, is drafting the static projection, ignoring the bold predictions on either end."

"What do you mean?" Norm eagerly inquired.

"Projections aren't static. Yeah, that's how they're displayed, an expected stat line, but they're best thought of as a range of plausible outcomes in either direction. Even skills are static, they're a range. One year, a guy can fan 20 percent of the time, the next year it could be 24 percent. Maybe his contact skills declined, but more likely one season he performed to the upper end of the range, the next it was towards the lower end. Same guy, just slightly different results. This is the main reason a weighted average is used as the foundation. Using multiple years averages the different parts of the range, while weighing recent years more heavily accounts for the chance of a change in skills. The same goes for pitchers. One year, he can walk hitters at an 8.5 percent rate, the next it drops to 7.3 percent. Maybe he improved his control, but especially for a veteran, it was probably just variance around the same skill level. Again, weighted averages flesh this out to an expectation for the following campaign."

"Isn't this just upside and downside?" Norm asked. "Or floor and ceiling," Paul added.

"Yes, exactly. Most bold predictions are a hand-picked plausible outcome. The thing is, especially under the duress of the clock ticking down, or the pressure of having to decide to bid another dollar, we fixate on the stat line in front of us, failing to consider the range of outcomes and how that influences our roster construction. Keep in mind, irrespective of the process, projections aren't perfect, else we wouldn't even be having this conversation, right? Sure, we're looking at the most likely outcome or whatever you want to call it. A proper roster is all about balance, not just power versus speed or strikeouts versus ratios, but the level of risk incurred and the exit strategies built-in to alleviate the risk if things turn out poorly."

"Can you give an example?" posed Paul. "I set a target for the different categories and use projections to meet each one."

"Don't get me started on target drafting, I don't believe in it, but that's a discussion for another day. Let's look at some batters projected for about 25 homers. Let me think, how about Joey Votto, David Dahl and Anthony Rendon. They play different positions but let's say you can pick any spot. I'll bet all three are projected for low to mid-20 homers and ranked closely. This could be too simplistic, but Votto and Dahl both have downside and upside, but from different angles. Votto is the aging vet. He was a top-15 fantasy player in 2016 and 2017. Did he suddenly lose power? Was he hurt? Dahl is the up-and-comer, somewhere between shiny new toy and post-hype prospect. He had a strong second half and looks to have a job heading into 2019. Rendon is the stable one, perhaps underappreciated but with less likely variance around his projection. An argument can be made for rostering all three, though cost will obviously matter. Still, the point is, we're looking at similarly ranked players, with close to the same number of projected homers, but completely different profiles. Which is better, having a projection for each, or spending time researching them, so you have a better feel how they may influence your roster, good and bad? Shoot, call it coming up with a bold prediction for each player if you want."

"See, I told you projections are worthless," chuckled Norm as he elbowed Paul.

"Hold on," I interrupted, "I never said that. We all need to start somewhere. Some are fine with relying on their gut. Others prefer a more pragmatic approach. Count me in with those wanting to make the theoretical more practical, which is much easier said than done. As I alluded to earlier, it's not about the projection, but how you use it. Those eschewing a formulaic projection can still have great success because they have an outstanding sense of what to do with their rougher expectations. Others may indeed have a more accurate projection but lack the ability to put it to maximum use. That said, my money will always be on the drafter armed with projections, but also the sixth sense to optimally build the puzzle that is a fantasy roster."

"See, I told you projections are needed," retorted Paul slyly as he returned the playful nudge to Norm's ribs.

"Let me ask you this" I said, recognizing Paul was kidding. "If a genie appeared and offered you the choice of knowing the true skill level of every player or how much they'll play in the upcoming season, which would you prefer?"

Paul and Norm looked at each other, shrugging their shoulders with another blank look on their face.

"Personally, I want playing time. If you take a bunch of different projections and break each down to the skill level, I bet you find they're remarkably similar. Sure, there will be differences, but remember, we're talking about a range, so even when there are differences, there's huge overlap within the range of plausible outcomes. Tell me who will get hurt, who will emerge with more playing time, etc. and I'll crush the league. If you opt for skills, maybe you pick up a break out or two while avoiding a landmine here and there, but can't you figure that out from playing time? A breakout player will be rewarded with more at-bats/innings while a struggling player will hit the pine. No doubt about it, give me the playing time."

"The problem is, we don't have that information," lamented Norm.

"No, we don't," I agreed, "But if you're going to spend time in the offseason preparing for your leagues, which is better: projecting every player or relying on an average skill level and deciding how much you feel everyone will play? I've said this for years, if you incorporate your personal playing time, you do your own projections. You don't need to write everything down to the individual plate appearance or inning pitched. Going through and generating your own depth chart and likely batting order/pitching rotation for each team perfectly suffices, along with thinking about likely in-season callups. Constructing a depth chart elucidates potential platoon partners as well as the quality of a team's reserves. The better the bench players, the more action they'll take from the regulars, stuff like that. I run a business, a great part of which is doing projections. Yeah, I research every player as much as I can. But trust me, sometimes I wish I wasn't in the business, so I could get some average projections and spend my time really digging into the players and conjuring some of my own bold predictions instead of having to make spreadsheet-driven adjustments for every single transaction coming across the wire. Don't get me wrong, I love what I do, but you know, there's never enough time to do everything, so I do what I do so you guys can take it to the next level."

"Hey Norm, looks like our ride is here," Paul exclaimed. "Thanks man, see you at the game?" he asked me.

"I'll be there," I affirmed.

A couple hours later, I overheard a conversation between Norm, Paul and who I assumed was their driver.

"It's not about projections," Paul emphatically insisted.

"Yeah, it's about what you do with them," Norm asserted.

That's why I love what I do.