Predicting the success of prospects into the future can be fool’s gold. But in baseball, there are a few parallel attempts in trying. Throughout the years, we’ve advanced in these attempts. Major league teams now run their own projections. Player development can be more accurately measured with new forms of technology. Everything is tracked and we’ve never been more accurate.

The truth is, there is no gospel to projecting minor league performance to the major leagues. We try to find one, but there are just too many immeasurable factors that we can’t see.

With all this being said, there is good use in the few ventures used to project the performance of minor leaguers. Scouts can pick up things computers can’t when it comes to available minor league data. How much of his success was against non-major league competition? How well can he hit high-velocities? Will smarter pitchers at higher levels be able to expose his approach at the plate?

On the flip, projections can have their helpful qualities too. How have comparable players hit at the major league level? How have batters at a certain age in a certain league fared? Both types of evaluations can’t be compared against each other. They have to be taken for what they are. Understanding the strength and limitations of both is probably the best way to do this thing. With all that said, I formally come here to tie into this, announcing a project I’ve been working on for the past few months.

---

Quickly after becoming a baseball fan, my biggest interests were two aspects of looking at the game for some unexplained reason. The first being minor league baseball. It’s a guessing game, but there’s just something about watching players improve and model their game pre-majors that I love. The second aspect was analytical. That should be obvious, as I’ve been writing at Beyond the Box Score for almost a year now. I’ve always had an analytical mind, and with that being an excelling point for baseball, I guess that’s why it always jumped ahead in front of other sports for me.

With how my passion for baseball was modeled through analytics and prospects, seeing a projection system like KATOH pop up was a dream. And then seeing it go away wasn’t so fun. But without KATOH and some free time on my hands, I figured... why don’t I try making one? This would originally be something not-so serious and something I only kept to myself, but ideas sprung upon ideas, and I just kept making it bigger and more detailed. A few days ago is when I finally finished this projection system, coming after months and months of sorting through hundreds of Excel pages, moving data around, coming up with formulas, and then a whole lot of refining. Thus, I bring to you the KILA Projection System (yes, it’s named after former Quad-A great Kila Ka’aihue).

---

The Method

The blueprint to KILA is honestly quite easy to understand, yet very detailed at the same time. The main goal of this projection system is to project a career wRC+ for any prospect in baseball (sample size pending). Correlations for minor league wRC+ to major league wRC+ are somewhat helpful, but not very great (r=~0.3). My thinking was to improve this correlation by developing a score (KILA Score) that weighted a player’s performance in certain statistical areas.

It wasn’t as simple as grabbing every minor league player season since 2007 and performing this all together. If you did that, there would be huge disparities between players in things such as league environment and competition. That’s where the biggest part of this KILA comes into play. Each player is compared to a group that contains their league and age peers. I took six statistics that I deemed noteworthy for minor league hitters through some side research (wRC+, K%, BB%, ISO, BABIP, GB%).

I originally went in with the plan that I would weigh each statistic the same at every level. Deeper research showed that wasn’t wise, the level of usefulness between stats like K-rate and ISO in the lower minors and upper minors is basically night and day. This table below will show I had to weigh each set differently.

As you can see, a lot of the metrics in short-season ball just aren’t predictable. Strikeout and walk-rate are the most important, while with the subpar fielding down there, ISO and BABIP have almost no predictability. This flips on full-season ball, as contradicting an old narrative, strikeout-rate is almost useless in predicting the success of a major league hitter. At the higher levels, power and walks are more important, while BABIP and GB-rate can have their use. I also scrapped all directional batted ball stats after initially exporting them, as their correlations are inconsistent and pretty useless.

Each player was given a z-score relative their age and league counterparts. These z-scores set an even scale for these weights to be distributed.

Lastly, after all the weighted z-scores were summed for each league and age group, I added one last age adjustment in, accounting for difference in age compared to the rest of league. This was weighted and finally the KILA Scores were complete. Once I had the scores, I used simple linear regression with ‘X’ being the KILA Scores and ‘Y’ being the results of past major league hitters.

The Results

One of the main goals of this was to find minor leaguers that had somewhat fallen off the spectrum of what you would consider a “prospect,” yet were still putting up numbers that would indicate future major league hitting success. I did find a couple of those guys, such as Roberto Ramos, Jared Walker, Zach Green, and Brandon Wagner. But anyway, here were the top 50 KILA Scores from the 2018 season (sample sizes at each level eliminate a few guys, example being Juan Soto, though I’m 100 percent sure he’d be near the top if qualified).

Disclaimer: Remember this is 100% stats based. There will be obvious misses with this, while hopefully some good under-the-radar pickups. Also, there could be some sample size issues here.

And to set more of a guideline, here are the top 50 KILA hitting scores dating back to when extensive minor league data became available (2007-). And now you see why it’s named after Kila Ka’aihue.

The Predictability

I can’t sit here and rave about KILA’s predictability as much as I want to, but what I can say is it is more predictable than any standard minor league metric you could go off of. That’s the product of added context and focused importance.

Out of a sample size that adds up to nearly 3,000 batters from since 2007, the correlation coefficient between your standard minor league wRC+ and major league wRC+ comes out to r = 0.39.

Now for the correlation between the KILA projected wRC+ and the actual major wRC+, which comes out higher, standing at r = .423.

While only a slightly better predictor, it is still better than any other minor league number you could use. I’m almost confident it could be improved in the future too, as it is a goal of mine to make these numbers ballpark adjusted, an issue often faced when analyzing minor league numbers.

---

Projecting how a player will perform against higher competition is a tough task. There are just too many mental and physical factors that we tend to miss. But hey, making an attempt at it is fun. If it wasn’t, then I wouldn’t put all this work into doing it. And I plan to keep working on it, as I mentioned adding in the ballpark factors above, along with adding in an age curve, making pitcher projections, and doing a whole lot more refining to the system. Expect more content on this in the future.

---

Patrick Brennan loves to research pitchers and minor leaguers with data. You can find additional work of his at Royals Review and Royals Farm Report. You can also find him on Twitter @paintingcorner.