RESEARCH: Home and Road pitchers


A few years ago, we introduced our Starting Pitcher Matchup Tool. We considered a lot of inputs in building the tool, and the result was a complex model that was more nuanced than what we had before. However, we are always in search of improvements. Today, we’ll consider what might be the biggest factor that the model neglected: home and road splits. In doing so, we will answer these questions:

  • In the current game, what are the broad effects of pitching at home vs. on the road?
  • What are the skill changes associated with those effects?
  • Are there pitchers who are repeatably better or worse than those averages either at home or on the road?
  • Would an in-season sample have any predictive value for the rest of that season?


We’ll review the impact of home and road on fantasy starters from 2017-2019.  Then, using data from 2003-2019, we'll look for systematic differences in pitching skills at home versus away. To minimize park effects and other sources of noise or bias, we’ll look at skills rather than outcomes. 

We’ll consider the obvious K%, bb%, SwK%,  Ball%, GB%, FB% along with less obvious candidates, BF/G, Hard%, Soft%, and defensive efficiency. Most of these metrics are inputs to the existing SP matchup tool. Ball%, Hard%, and Soft% currently are not, but they are candidates for later inclusion, so we’ll do the back-end work now.

For those metrics that show significant home/road splits, we’ll examine cases where individual pitchers have home/road splits that diverge from the league trends, and if so, determine if they are predictive from year to year. Then we’ll look at half seasons to see if they have promise for later incorporation into an updated SP matchup tool.

League-wide Data

Nearly any look at home vs. road performance will reveal a split that is favorable to the home team. We look here at the ratio of home-to-away results over three seasons. The first table shows the outcomes, i.e., what we typically care about in fantasy.

            Home/Away, by Season     3-year
Metric      2017    2018   2019      Average
======     ======  ======  ======    =======
ERA         .916    .921    .924       .920
WHIP        .947    .954    .948       .950
K/GS       1.062   1.087   1.054      1.067
W/GS       1.100   0.993   1.085      1.059

A number above 1.000 means that the league result in that metric is higher at home than on the road. Likewise, a number below 1.000 means that the league result in that metric is lower at home than on the road. For example, the .920 figure at top right means that the league-wide ERA is 8% lower at home than on the road.

Before moving on, let’s dig into wins per game started; that 2018 value isn’t consistent over these three years like the others are, so let’s look at a broader history:

Wins/GS by year

We see that this number is consistently around 1.08, except when it inexplicably drops below 1. The three year average from 2017-2019 is lower than what is seen historically, but, given uncertainty in the always changing offensive environment, not an unreasonable estimate.

This next table shows the changes in the leading indicators that logically feed those outcomes, as well as parameters that are current inputs (or potential future inputs) into the SP matchup tool.

            Home/Away, by Season     3-year
Metric      2017    2018   2019      Average
======     ======  ======  ======    =======
K%          1.042   1.064   1.041      1.049
bb%          .960    .926    .951       .945
SwK%        1.042   1.056   1.062      1.054
Ball%        .992    .984    .988       .988
BF/GS       1.019   1.021   1.013      1.017
GB%         1.006    .995   1.013      1.005
FB%          .988   1.009   1.000       .999
Soft%       1.007   1.033   1.031      1.024
Hard%        .982    .998    .991       .990
GB_DE%      1.019   1.012   1.009      1.013
FB_DE%      1.007   1.001   1.011      1.006
LD_DE%      1.019    .972    .989       .993

Note: The last three are measurements of defensive efficiency, outs per ball-in-play (non-HR) by batted ball type.


  • The plate discipline metric ratios support the differences we observed in the ERA, WHIP and K/GS metrics above
  • Home pitchers pitch deeper into games, perhaps because they are pitching better
  • Both GB Rate and FB rate bunch around 1.0. A look at a longer history reveals:
    • From 2003-2019, mean H/A ratio for GB rate is 1.004 ± 0.008
    • From 2003-2019, mean H/A ratio for GB rate is 0.999 ± 0.010
    • These ratios are so close to 1.0 that we can neglect both of these effects, at least on a league-wide level
  • Quality of contact against is slightly more favorable when a starter is at home
  • Teams are more efficient turning GB and FB into outs at home
  • It appears a near certainty that including Home/Road impacts will improve our SP matchup tool. We will address this in future work.

Individual Pitcher-Seasons

In this section we examine whether individual pitchers show a tendency to excel at home (or away) more than the typical amount found above. To gauge this, we will look at a pitcher’s home/road skills in one season and see how well-correlated they are to the next year. Note: when we have them we will use historical park factors to park-adjust the results.

We filter for individual pitcher seasons with at least 150 batters faced as a starter both at home and on the road.  This yields 2,525 pitcher seasons.  We further filter out any seasons where he did not meet those batters-faced thresholds in the previous year. This leaves 1,703 pitcher-season pairs.

In each of these pairs, we calculate the ratio of home performance to road performance, as we did earlier. Then we adjust them based on a moving seasonal average, because we are looking for starters that exceed our newly found home/road expectations.

For example, in 2018 Shane Bieber struck out 24.0% of batters at home and 26.2% of batters away.  Accounting for park factors for each game he pitched, those rates adjust slightly to 24.0% and 26.1%.  Thus, the park-adjusted H/A K% ratio is 0.92.  Once we account for league trends, that ratio is adjusted to 0.88. I.e., he struck out batters at home at 88% of the rate he did away, compared to the league average for starters.

We’ll plot the current year result against the previous year result, weighted by the geometric mean of the batters faced in the two years.  Then we’ll fit a line to try to find a relationship.  The results aren’t promising. Here is the result for park- and league-adjusted K% Home/Away ratio:

K percentage home-away ratio

The line of fit is CY = 1.023 - 0.003 * PY, with and R2 value of essentially zero.  The correlation is so small that there is no predictive value contained in the prior year’s Home/Away trend.

Most of the fits looks similar, with shallow slopes and low R2 values. To spare the reader the tedium, we present them in a table:

Metric        Line of Fit        R2 value
======     ==================   ===========
K%^         1.023 - 0.003*PY       0.000
bb%^        1.058 - 0.019*PY       0.000
SwK%        0.870 + 0.128*PY       0.015
Ball%       0.992 + 0.010*PY       0.000
BF/GS       0.972 + 0.029*PY       0.001
GB%^        0.984 + 0.027*PY       0.001
FB%^        0.946 + 0.058*PY       0.004
Soft%       0.931 + 0.097*PY       0.007
Hard%       0.925 + 0.088*PY       0.008
GB_DE%      0.929 + 0.083*PY       0.007
FB_DE%      0.518 + 0.484*PY       0.222
LD_DE%      0.909 + 0.102*PY       0.011
^ park adjusted

Aside from FB_DE%, these are miniscule correlations. While this doesn’t rule out the idea that there could be pitchers out there who consistently excel at home, it does tell us that we can’t find them.

Regarding the fly balls, this finding reveals that while league-wide there is no home/road benefit, certain teams’ ballparks do enable them to convert more or fewer non-HR fly balls into outs on a consistent basis. Here is the plot of fly ball defensive efficiency versus the prior year result.

Flyball Defensive Efficiency

Thus, in the future, we should validate that our model adequately accounts for the differences in a pitcher’s defense when at home or on the road.

Impact within a season

Given that there were no significant season-to-season correlations for pitchers, it seems highly unlikely that there are within season correlations.  However, we have previously found some surprises in that regard, notably that hard% and soft% allowed are better correlated within a year than they are year-to-year. So, we’ll look anyway for correlations between first half Home/Road splits and second half. There aren’t any.  The best correlation was Soft%, and it had an R2 value of less than 0.01. If they exist, we can’t see them.


We found that:

  • Home/road splits are consistent across the league for many parameters, and we should endeavor to incorporate them into the SP Matchup Tool.
  • In general pitchers do not exhibit Home/Road skill tendencies beyond these that persist from year to year.
  • Teams have significant repeatability in converting non-HR fly balls into outs, and this should be incorporated into the SP Matchup Tool in the future.

It’s worth reminding the reader that we looked at pitcher skills, not outcomes. How those skills play out is certainly influenced by ballparks. This shows up in our park factors, and the park factors are included in the SP Matchup Tool.

And finally, note that our inability to find correlations in home/road skills splits from year to year doesn’t mean that there aren’t any pitchers who thrive at home and suffer on the road, and will do so every year.  But, it does mean that among the thousands of pitcher-seasons, they are so few as to be invisible.  If a pitcher does have an extreme split one year, you are far better off betting on that split vanishing the next year, no matter how many times they’ve shown the split.

Click here to subscribe

  For more information about the terms used in this article, see our Glossary Primer.