Last month I introduced a new metric, earned home runs (eHR). For the background and origin of the metric, feel free to check it out here. I’ve made a few slight adjustments to the equation after determining that pulled fly balls and park factors had a more significant impact on home runs than I originally accounted for. Based on this information, I’ve adjusted the weights for each factor accordingly in my formula A more complete list of 2019 earned home runs is shown at the bottom of this article. The purpose of this article is to determine if the eHR metric has any predictive power.

I’ve received some requests about the year-to-year correlation of eHR. Unfortunately, I’m not proficient in r, not even a little bit, so pulling data takes time for a caveman, like myself. Additionally, the properties of the ball seem to change every year during the Statcast Era which began in 2015 making it difficult to accurately determine the year-to-year correlations. With this in mind, I only went back one year to 2018 (year 1), 2019  being year 2.

After calculating the 2018 eHR, I compared 2018 earned home run per plate appearance (eHR/PA) and 2018 earned home run per fly ball (eHR/FB) to 2019 HR/PA and 2019 HR/FB. Due to the large increase in HR/FB% from year 1 to year 2, I had to adjust the baseline from year 1 to year 2. I then compared the players that appear in both samples to determine the year 1 to year 2 correlation of HR/PA and HR/FB. Now, there is one caveat. I only utilized players within my sample which is much smaller than 100% of the player pool. I did this as an attempt to eliminate small samples because barrels are a major variable in my equation. Players with either a low number of plate appearances (under 200) or low barrel totals seemed to skew the results. So, they gone. It’s not perfect, but at least it utilizes the same sample of players from year 1 (2018) to year 2 (2019). Here are the results:

So, what does this mean? You may be wondering why are the correlations so low. I thought the same thing because in the past HR/PA and HR/FB year-to-year correlations were typically stronger. But, remember, I used a limited sample, so this data in no way mimic’s the league as a whole. Secondly, the change in HR/FB% from 2018 to 2019 was 2.6% (12.7% to 15.3%). Between 2015 and 2018, the year-to-year difference in HR/FB% was never more than 1.4%. The stark difference in home run rate between 2018 to 2019 partially explains the weaker correlation.

OK, back to the correlations in the table above. Let’s not get carried away. It’s just one year where the eHR metric performed slightly better than using actual home runs. It’s difficult to say at this point that the eHR metric has more predictability than strictly looking at HR/PA or HR/FB but it’s a start. What it can do with more confidence is help identify outliers. Based on this information we can expect a large portion of the players identified in my original article to regress towards the mean rather than continue to be extreme outliers.  Additionally, it can help establish breakouts that don’t register by just looking at a player’s Statcast page because eHR utilized home run park factors (home), pulled fly ball rate, etc.

Year 1 to Year 2 HR/FB Predictions based on eHR

Let’s start with the HR/PA metric first. Just over 56% of both 2018 over and under-performers regressed toward the mean based on my eHR metric in 2019. While these results don’t scream breakthrough, it’s a decent start. Based on the year 1 data, more than half of the players performed the way the metric expected them to perform in year 2, so that’s good. Moving over the HR/FB, we see the results are a little bit better. While once again, just under 56% of the under-performers regressed toward the mean and just over 66% of the over-performers regressed the following year! That’s two-thirds fam! I’ll take that all day but I can’t explain why the stark difference. More research to follow, I just need to either learn r or find some to scrape data for me. Any takers?

As I mentioned earlier, the most important way to digest this analysis is to focus on the outliers. After adjusting for the league-wide increase in HR/FB rate from 2018 to 2019, 15 of the top 25 players under-performed outliers from 2018 improved their HR/FB rate in 2019. That’s a modest 60%. On the other end, 18 of the top 25 over-performed outliers from 2018 decreased their HR/FB rate in 2019. That’s a more optimistic 72%. The year one to year two correlations are moderate-to-weak but they are not completely useless. I plan on interpreting Alex Chamberlain’s Deserved Barrel metric in conjunction with my eHR metric to determine a more concrete list of players I feel confident about going into 2020. 

If you’ve made it this far, thank you for reading and your patience is appreciated. Below is the complete list of earned home runs from players in my sample. I’ve also included a Google Sheet that has the 2018 data along with 2019. Remember, for reference, the league average home run per barrel rate (HR/BRL%) in 2019 was 59.59%.

Follow me @FreezeStats. Check out my work at FantasyPros and Pitcher List.

Thank you for checking out my content. If you enjoy this work and want to see more, please feel free to donate.