Projections & Category Targets

On a daily basis, I receive a question from someone asking me – how many homers do you think Jim will hit this season? Can Harry hit .300 this season? Will Steve strike out 200 batters? Every day the questions come, and I get it. Growing up, one of my most favorite things every January was to get my grubby hands on that first fantasy baseball magazine, whip it open to the player capsule section, and start plowing through the projections for players. Alas, thirty years later, I now realize why this is a complete waste of time. Don’t agree? Let me prove it to you.

THEY AREN’T ACCURATE

Most legitimate sources will tell you projections are, at best, about 70-75 percent accurate.

There are many reasons for this fact; the most logical being playing time. Will a guy get hurt thereby throwing off the projections accuracy? If Player A is hurt, who is Player B who gains playing time? Will a player be traded and either lose or gain playing time? What team will call up some hot-shot rookie early, thereby taking playing time from an established player? All of these facts totally change a guy’s outlook.

Projection systems simply aren’t accurate. In fact, the Marcel the Monkey projection system, which basically just looks at the last three seasons, is about as accurate as anything anyone is putting together out there.

I will speak more to this below by giving a few examples of projection fallacies.

THEY ARE MADE UP

I have to be honest. I’ve been doing this for a long time. I’m one of the few out there who have been employed full-time in fantasy sports for the past 15 years. I’ve seen this industry grow from being a hobby to being part of the fabric of sports. I’ve also worked for companies large and small (at one point I worked for Fanball who was owned by Liberty Interactive who also happened to own QVC, SiriusXM and the Atlanta Braves). Point being is that I’ve been there grinding it out as well as sitting on a fluffy cloud. I bring this up, not to sound cool, but to let you know that what I’m about to say is based upon a ton of experience. Here it is:

Projections are made up.

I know you’re reading that and thinking, yeah, I know. But let me be clear. When I write “they are made up,” I am literally saying they are completely fabricated. Some of the places I’ve worked have someone crunching data, diving deep into the numbers, and creating algorithms on spreadsheets. At other times, frequently actually, someone is tasked with putting together projections for player profiles. There is no algorithm. There is no computer help. It is merely a fella sitting in a corner, late at night, suggesting projections for players based on their personal feelings about a guy. I’m being totally serious here. Sometimes folks base their projections on nothing more than a personal hunch. In other instances, the projections, the “computer ones,” spit out that Player A is better than Player B. However, the human analysts doesn’t believe that is accurate so they simply adjust the projections until Player B is ahead of Player A. Again, I’ve seen it happen... a lot.

If you want to base your thoughts on a player based on systems like that, be my guest. I think it’s a horrible ineffective direction to take if you’re asking me.

EXAMPLES TO EXPLAIN THE PROBLEM

Let’s go through an example.

I believe Franklin will produce a fantasy season of .275-25-75-75-15. Let’s go point by point.

Let’s say that Franklin hits .275.

Let’s say that Franklin has 25 homers.

Let’s say that Franklin has 75 RBI.

Let’s say that Franklin has 75 runs scored.

Let’s say that Franklin steals zero bases (hello Manny Machado).

So you tell me, how accurate is this projection?

I would be 100 percent correct on batting average.

I would be 100 percent correct on home runs.

I would be 100 percent correct on runs batted in.

I would be 100 percent correct on runs scored.

I would be zero percent correct on steals.

So again, I ask, how accurate is this projection?

I would be 100 percent accurate in four of five categories. That would be a “win” right? That’s an 80 percent success rate (4/5 categories at 100 percent). Would you consider this projection to be accurate? I certainly would. As I noted at the top, even the best systems struggle to even approach an 80 percent success rate. However, if you rostered Franklin hoping for 15 steals, and he stole zero, would you really consider this to be accurate? He hit all four batting numbers perfectly, but your loss of 15 steals might cost you four points in the steals category. The loss of 15 steals from the projection also costs you like $6 dollars in Franklin’s overall fantasy value. Therefore, is this an accurate or inaccurate projection?

What about this one?

What if I project Franklin to earn $20 based upon the numbers I’ve listed. Let’s say Franklin earns $20. That’s a dead on accurate projection, right? Now, let’s say that instead of producing $20 with a .275-25-75-75-15 he produces $20 of value with a .245-30-95-90-0 line. Would you consider that to be 100 percent accurate projection? You would have to if you were talking about the actual fantasy $ produced. However, real worlding it, you wouldn’t be happy with it. You would have lost .030 batting average points and 15 steals off the initial projection. How could you possibly be happy with that? It would be impossible to be happy about it, yet the projection would have to be considered accurate if we’re talking bottom line.

Realize as well the following issue that can make you look like a genius, or an idiot, even when the outcome is same relative to the projection.

As we’ve seen, Franklin has a 25-homer projection for 2017. Let’s play that out.

I say Franklin is a 25-homer bat.

He hits 30 I’m a genius.
He hits 20 I’m an idiot.

Pull back from that for a second.

I projected 25 homers, so 30 homers is a +5 while 20 homers is a (-5). In both instances, 20 and 30, Franklin was five off the homer projection. In both instances, my projection accuracy was exactly the same. In one instance you would love the success. In the other, you would hate it. But from an accuracy standpoint they are both identical. You can do this for every category. Follow me here:

Remember though, both the “plus” and the “poor” projections are equally correct – it’s just your perception that is off since one outcome would benefit you while the other would hurt you. Look at the massive difference in the total outcomes for each effort. Now realize both are equally “off” the projection.

How about another angle?

There are six months to a season. Let’s say a guy is projected to hit .290 over 500 at-bats (145 hits). Let’s say that guy has one more hit each month of the season, six of them, leading to 151 hits in 500 at-bats. In that instance he would hit .302. That’s one measly hit over an entire month. Let’s say that a guy has one less hit a month than expected, six less overall, so he ends the year with 139 hits in 500 at-bats. That guy hits .278. That’s a swing of .024 batting average points and we’re talking about one fewer, or one greater, hit total per 30 days. That’s a pretty accurate prediction model then, if I’m only off by one hit a month, but it takes a guy from being ‘just a guy’ in batting average to a borderline star. One measly hit over 30 days.

Facts are facts.

CATEGORY TARGETS

As you can tell from the above, I think shooting for certain targets in each category is a good idea but nearly impossible to attain with any accuracy. I might think a guy is a 30-home run hitter on Draft Day. What if he’s really a 25-homer guy? My projection, therefore, has him as a +5 homer bat. What if I have five players who I roster and are all +5 bats? In the end I “think” I have 250 homers on my team based upon my projections. But, if they are really only 225 homers worth of bats... you get the point. Knowing the category targets doesn’t hurt. Knowing what numbers your team likely needs to hit to win is a good thing. The problem is that we simply don’t know if the players you roster are going to reach those numbers.

Note that folks smarter than me have determined that only 65-85 percent of the totals that your team will produce over the course of the season are drafted meaning the rest of the production it picked up through waivers or trades. Therefore, the whole ‘I hit all my targets on draft day’ talk really isn’t accurate anyway. Regardless, for those of you that want some baselines, here you go (assuming nine pitchers, and 14 hitters):

12-Team Targets

15-Team Targets