Momentum in Baseball: An Analytical Perspective

Article written by Ari Goldberg

1. Introduction

Momentum remains one of baseball’s most persistent in-game narratives: the belief that a team can parlay a “shutdown inning” into offense, that a “big inning” boosts run prevention next half, or that a “hot offense” persists across innings. This study tests those claims using data from the Prospect League and the Kernels Collegiate League at the half-inning level using indicators for clean defensive halves, poor pitching halves, consecutive good defensive halves, and offensive success thresholds (e.g., ≥2 or ≥3 runs) in previous innings. We evaluate impacts on average runs and the probability of scoring at least one run in the immediately subsequent half-inning.

Some models interpret coefficients by exponentiation: for a coefficient β, the incidence rate ratio (IRR) is e^β and the implied percent change vs. reference is (e^β−1) × 100%. For example, β=0.10⇒ ~+10.5%. Where “probability of scoring at least one run” is modeled, we still use the same exponentiation to interpret effects as percent changes relative to baseline. Throughout, we include contextual controls (e.g., inning/side) to avoid spurious momentum from lineup position and inning run environments.

2. Variables, Outcomes, and Interpretation (Academic framing)

Momentum indicators (previous half unless noted):

• Clean/quick defensive half. Coding is typically 0 = not clean/quick, 1 = clean/quick
  • 3 up 3 down flag explicitly: 0 = not 3 up 3 down, 1 = 3 up 3 down
• Bad pitching half: 0 = not bad, 1 = bad (runs / multiple baserunners allowed)
• Two consecutive good defensive halves: 0 = not good, 1 = good
• Offensive run thresholds for the prior inning:
  • ≥2 runs: 0 = <2, 1 = ≥2
  • ≥3 runs: 0 = <3, 1 = ≥3

Outcomes (next half-inning):

• Average runs
• Probability of ≥1 run
• Average batters faced

3. Results

3.1 Intra-inning Offensive Momentum

Much like an avalanche needs a trigger, momentum can sometimes be kick-started by a singular play or event. The most obvious spot for the batting team would be in the shape of a leadoff hitter. As would be expected, all outcomes that end with a player on base have a higher probability of a run scoring and a higher average runs in the inning. The unexpected result was that the average runs scored was higher when the leadoff result was a triple as compared to a home run. Also, the probability of any run scoring in the inning was roughly the same for a triple vs a double, which makes sense considering that both outcomes require another batter to drive them home. It was also surprising that the probability of scoring a run was noticeably higher when drawing a walk than when hitting a single. This is likely explained by the fact that a walk results from the pitcher missing his spots, which carries over into the next at-bats, while a hit is due to the hitter’s ability, which varies from hitter to hitter.

Similar to the surprising average run discrepancy between a leadoff triple and a leadoff home run, we see there is an extra batter faced on average for the leadoff triple. In fact, the average batter faced when there was a home run was less than any of the other outcomes that allowed a batter to reach base safely. 

When we run a logit regression on the probability of scoring a run based on the lead-off result, we see some consistencies with the previous model. The statistically significant results are found when the leadoff hitter strikes out or bats into an out. We prove again that leadoff outs are killers. Both BIP outs and strikeouts dramatically reduce the chance of scoring, with strikeouts showing the largest drop, likely due to the presence of a pitcher good enough to throw consistent strikes. A different outcome from the previous model is that leadoff hits don’t guarantee runs. Although directionally positive, they aren’t statistically significant in this dataset, suggesting lineup order and depth matter as much as just getting that first hit. 

There is some evidence that the leadoff outcome does influence the probability of scoring a run and the number of runs scored, which could be labeled as momentum for the offense. Still, it can also be confused with the simple fact that a runner on 3rd with no outs is much more likely to score a run than a situation with no runners on and 1 out.

3.2 Defensive Carryover Momentum

We see some indications of momentum when there is a positive leadoff play, but can a well-pitched inning also lead to more runs by the offense? First, we look at innings where the pitcher retires all 3 batters that he faces. When a pitcher can get through an inning spotless, we see that the probability of a run scoring in the next half-inning increases by 2% and the average runs scored increases by 0.15. While these are only marginal improvements, they still show better performance by the offense on average.

However, when I widened the range for what was considered a good defensive inning to include innings defined as at most 4 batters faced, at least 1 strikeout, and at most one walk, we see the opposite effect. Average runs decrease by 0.04, and the probability of a run scoring decreases by 4%. Looking at these results together with the previous results, it is hard to conclude the effect of a well-pitched inning.

So we cannot say that a good defensive inning leads to better success on the offensive end, but maybe the inverse situation holds, which could be described as negative momentum. If a pitcher/defense has a bad inning (in this project, that is, an inning with at least 1 run scored and at least 6 batters faced), does that poor performance lose momentum and stall the offense out, or does the offense not feel the consequences of momentum against them? We see that after a bad defensive inning, average runs increase by 0.0,2, and the probability of a run scoring increases by 2%. This again disproves the notion of momentum across half innings, because if momentum were to be felt, we would see a decrease in runs scored after a bad half inning.

Finally, I tested if stacking good defensive innings (those quick innings with fewer than 4 batters faced and no runs) could lead to better offensive performance. However, one hypothesis is that the effects of momentum are not felt after just one inning, but require a buildup throughout the game. As we see in the data, the average runs scored barely increases, only 0.006 more runs scored, and the probability of a run scoring decreases.

I then wanted to see if really bad defensive innings, where the opponent scored multiple runs, would affect the offensive production. We see a small uptick in production when at least 2 runs are scored in the previous inning, but when 3 or more runs are scored, there is a decrease in the amount of runs scored. However, the probability of run scoring in both of these situations is slightly higher when the defense performs poorly. However, across all these data points, there is little to no effect of momentum carrying over from defense to offense.

Now, when I run a generalized linear model regression, I see none of the results from this section are statistically significant; none of the variables that I tested have an indication one way or another. We cannot say that momentum is built from any of these defensive indicators to positively or negatively influence offensive production consistently. 

3.3 Offensive Carryover Momentum

Here we are trying to see if a strong offensive performance can build up momentum that then inspires a strong defensive performance. However, we see that not a single variable here is statistically significant, meaning there is no indication that a better offensive half inning would lead to a good defensive half inning or that a bad offensive half inning leads to a worse defensive performance. While the batters and fielders are the same players, just because a batter is doing well does not mean he will produce the same level of results in the field. A great performance at the plate will not suddenly transform a 40-grade defender into an 80-grade defender.

I then tried to see if a strong offensive performance could increase the odds of allowing zero runs in the next half inning. Similarly to the previous results, we see that there is no indication that any of the variables tested would actually lead to better overall defensive performance, much like the previous analysis.

3.4 Inter-inning Offensive Momentum

Since I found no connection between defensive and offensive performances, I wanted to test whether good offensive production can carry over between innings. Does a strong rally in one inning mean that there will be continued strong performance in the following innings due to built-up momentum? No, there is no detectable same-game offensive persistence after accounting for inning and side. In other words, there is no indication that momentum is the cause of a hot streak by a team. If they were to be successful across multiple innings, it is likely due to poor pitching that the batters take advantage of, or it is because of random chance that all the hitters managed to consecutively and consistently get on base/get hits. 

4. Conclusion

This project set out to evaluate whether momentum holds up under statistical scrutiny. Across four dimensions of analysis, intra-inning, defensive carryover, offensive carryover, and inter-inning persistence, the evidence consistently shows that momentum is not a reliable or significant force. While leadoff outcomes meaningfully affect run expectancy within an inning, these effects can likely be explained by the structure of base-out states and pitcher stress rather than contagious performance spreading through a lineup.

When looking across half-innings, neither strong defensive frames nor poor ones translate into consistent offensive outcomes, and big offensive innings do not reliably improve defensive performance in the next half. Even inter-inning offensive persistence vanishes once inning and side controls are introduced. The numbers point to independence: each half-inning resets the strategic board rather than carrying forward psychological energy or momentum.

This leads us to an important conclusion about baseball’s character as a sport. Unlike soccer or basketball, where continuous flow allows for emotional swings and momentum to directly alter possession quality, baseball’s structure more closely resembles a series of discrete events. Still, certain variables are not independent, such as pitch count, base-out states, or previous plate appearances, but when it comes to momentum specifically, it is better to think about baseball as independent events. Each plate appearance represents a reset, shaped more by matchups, the base-out state, context, and talent than by the events that immediately preceded it. What feels like momentum in baseball is usually the result of sequencing luck, pitcher degradation, or mathematical changes in run expectancy, not a contagious force.

For front offices and coaching staffs, the implication is clear: strategy should be anchored in controllables, pitcher usage, matchup leverage, and lineup optimization, rather than chasing momentum narratives. Momentum may remain useful as a motivational tool, but from a data-driven perspective, baseball outcomes behave more like independent trials than a continuous cascade. Understanding this independence is critical to sharpening decision-making and resisting the temptation to lean on myths unsupported by evidence.

Since only data from the KCL and Prospect League were used, we can not technically extrapolate these results to the professional levels of baseball, but some of the insights can still inform future projects and decisions.