Performance With and Without Runners On, and Hitter Valuation

The increased prevalence of defensive shifts, as well as recent stories touting certain players as “shift-proof,” got me thinking: Is it a good thing to be shift-proof?  Is it inherently better to be a player against whom defensive shifting is less effective, or is there room for different players with different make-ups?  A downstream effect of defensive shifts is that, because teams shift less often (and shifts are less exaggerated) with runners on base, we start to see differences in a hitter’s performance with runners on versus with the bases empty.  We also notice other effects of players performing differently based on the number of baserunners.  In this post we’ll take a look at how we observe significant changes offensive performance (often fueled by changes in BABIP) of a few sample players when there are runners on base, versus with the bases empty.

Let’s take 3 players with very high similarity scores to each other: David Ortiz, Jason Giambi, and Carlos Delgado.  First, a look at their career stats:

Player	G	PA	HR	ISO	BABIP	AVG	OBP	SLG	wOBA	wRC+	WAR
Delgado	2035	8657	473	0.266	0.303	0.280	0.383	0.546	0.391	135	43.5
Ortiz	2020	8467	443	0.261	0.304	0.286	0.381	0.548	0.392	138	41.7
Giambi	2242	8864	440	0.241	0.294	0.277	0.400	0.518	0.395	140	49.3

Pretty comparable overall.  Giambi has accumulated more WAR, primarily through having a few more plate appearances, but also from having a better walk rate, which drives up his OBP, wOBA, and wRC+ significantly as well.

Now let’s look at their splits with runners on vs. bases empty:

Player	Split	G	PA	HR	HR/PA	BB%	SO%	AVG	OBP	ISO	OPS	BABIP
Delgado	Bases Empty	1932	4430	255	5.8%	11.7%	21.4%	0.275	0.374	0.273	0.922	0.303
Delgado	Men On	1895	4227	218	5.2%	14.0%	18.9%	0.286	0.393	0.258	0.936	0.304
Ortiz	Bases Empty	1862	4193	262	6.2%	11.2%	19.1%	0.271	0.356	0.282	0.908	0.281
Ortiz	Men On	1851	4274	181	4.2%	15.2%	16.6%	0.302	0.406	0.240	0.948	0.327
Giambi	Bases Empty	1999	4513	224	5.0%	13.0%	18.1%	0.256	0.367	0.228	0.851	0.271
Giambi	Men On	2020	4351	216	5.0%	17.8%	17.1%	0.302	0.434	0.255	0.991	0.320

Here we start to see a lot of divergence.  With Ortiz and Giambi, we see a large increase in BABIP when there are runners on base (and corresponding increases to AVG and OPS).  With Delgado, there is only a trivial increase in BABIP, and a much smaller increase in OPS.

Here’s the difference in BABIP and OPS each player shows in the split between {bases empty} and {runners on}:

Player	BABIP(runners on) – BABIP(empty)	OPS(runners on) – OPS(empty)
Delgado	0.001	0.014
Ortiz	0.046	0.040
Giambi	0.049	0.140

Note that to some extent, all hitters tend to put up better numbers with runners on due to sampling bias – in an average “runners on” situation, a batter is more likely to be facing an inferior pitcher than in an average bases-empty situation.  Delgado’s splits are in line with the league-average splits for {bases empty} vs. {runners on}; in a given league season, the league-wide runners-on-vs.-bases-empty split in BABIP tends to range from 0.000-0.005; for OPS, the increase ranges from 0.010-0.030.  Ortiz and Giambi on the other hand show splits well outside this range that indicate there are other factors at play causing these effects.

Does this mean Ortiz and Giambi are tapping into some part of their psyche that allows them to suddenly transform into better players when runners are aboard?  Unlikely.  Ortiz and Giambi are pretty heavy pull hitters, especially looking at their ground ball spray charts, against whom defenses have often employed dramatic shifts to great effect.  However, with runners on base, these shifts tend to be less dramatic and less effective.  This is likely the primary reason for the large increases in BABIP with runners on (a 0.046 increase for Ortiz, 0.049 with Giambi).

Beyond this, although Ortiz and Giambi both show similar BABIP splits, they still differ greatly from each other in terms of their production with runners on.  Giambi’s OPS increases a whopping 140 points, while Ortiz’s only increases by 40 points.  This is largely due to Ortiz’s dramatic decrease in home run rate with runners on.  While Ortiz’s HR% drops by nearly 33%, Giambi has managed to continue hitting homers at the same rate when runners are aboard.  Do pitchers change their approach when facing Ortiz with runners on to “minimize the damage” and try to prevent him from hitting home runs?  Likewise Ortiz (based on the knowledge that pitchers will approach him differently) may change his approach at the plate as well.  The splits for other stats seem to bear this out, as Ortiz increases his walk rate and decreases his strikeout rate; this isn’t particularly revelatory, and in fact these trends are present for Giambi and even Delgado as well.

This has profound implications for player valuation.  Given 3 players who put up similar aggregate numbers over the course of the season, would you rather have the player who is going to produce at roughly the same level (similar AVG / BABIP / OPS) regardless of whether there are runners on base, or the player who is going to overproduce with runners on and underproduce with bases empty?  I’d go with the latter.  I’d prefer Ortiz to Delgado.  And then, since the decrease in Ortiz’s HR% with runners on is curious (and warrants further investigation), I’d prefer Giambi to Ortiz, Giambi being the even more extreme example of increased production with runners on.

As we start to see more and more defensive shifts (and if the assumption holds that shifts cannot be employed as effectively with runners on base), there will be more and more players who demonstrate these splits in performance.  WAR, for example, does not take this into account at all.  If a player is dramatically more productive (e.g. a 140-point increase in OPS!) with runners on, you would project his team to score more runs and win more games than if that player was replaced by a player who puts up equivalent full-season numbers (and hence, has the same WAR) but did not have the same splits.

It would be interesting to run some simulations (probably using Markov models) to more precisely quantify the impact a given player’s splits have on team run production.  Said impact would likely vary based on the team too (e.g. overall team OBP).  This could be similar to the analysis comparing how 2 players with similar wRC+ but different makeup (an OBP guy versus an ISO guy) can impact expected run totals for different teams in different ways.