Under Pressure
Honors Thesis for Jonathan Huntley
- 15 -
The above ANOVA plots display the mean game score for each group that contains outings
flagged as following positive, negative, or neutral sentiment and good, bad, or average
previous performance. Twitter sentiment is meant to be a proxy for pressure and previous
performance indicates whether the player is entering the game on a “cold streak”, “hot
streak”, or neither. Viewing these plots, one can not only learn about the individual
relationships between future performance and past performance or pressure, but one can also
visualize the interaction effect that could be occurring between the two independent variables.
This analysis yielded several key insights that were consistent with the literature review. First,
as seen in Figure 8, both pitchers had their highest mean game scores after a period of neutral
sentiment. In other words, the pitchers performed better when there was limited pressure from
fans. Additionally, performance was generally the worst when there was negative sentiment
from fans. This is consistent with the conclusion of Harris et al. that higher pressure was
associated with worse performance.
Secondly, analysis of Chris Sale shows remarkable consistency with the findings of Gray et
al. When viewing Sale’s ANOVA plot in Figure 8, the spread between average previous GSc,
and good or bad previous GSc increases substantially under negative sentiment. This indicates
that when Sale experiences large amounts of external pressure, quantified by large volumes of
negative tweets, he performs better when on a “cold streak” or a “hot streak” compared to
starts following an average performance. This phenomenon is exactly what Gray et al
reported.
Finally, viewing David Price’s results supports the conclusions of Golding et al. Although
there are some similarities between trends in Price’s data and that of Sale, such as both
pitchers performing relatively well after neutral sentiment, overall, it appears Price reacts
quite differently than Sale does in similar conditions. For example, Price’s highest average
game scores were following bad performances. Furthermore, Price’s worst average game
score is following a good performance and period of negative sentiment. These findings
support the idea that players react differently to pressure and psychological momentum. It is