MLB Gameday 7-26: Feelings Over Facts
Updated: Sep 8
"Captains Log: The algorithm is being a real bitch right now"
Because there really aren't that many data points in the 2020 season, I am assuming that is why the model only likes one team today: the Astro's. Granted picking a 3-1 favorite isn't the boldest prediction in the world, it's all that I have. So instead, let's look at the pitchers notes and see if the sentiment scores can tell us anything.
To do this, I am going to plug pitching notes into a library that will spit out a score on how positive, negative, or neutral the statement is. For example this quote:
Chris is great.
Should yield a positive score while this quote:
Chris is a lying bitch.
Should yield a negative score.
Most of the discrepancies that exist are due to the notes being neutral on one pitcher and positive/negative towards another. Which makes it interesting to see why neutrality exists, for example:
Greinke, who carried a shutout into the seventh inning of Game 7 of the World Series, went 8-1 with a 3.02 ERA in 10 starts with Houston after being acquired in July.
Pretty by-the-books statement there. His opponent?
The 29-year-old southpaw is seeking his first win over the Astros after going 0-2 with a 5.40 ERA in four starts last year as a rookie, including 0-1 with a 6.43 ERA in three outings at Minute Maid Park.
A lot more hopeful than Grienke's. I am guessing words like "seeking" and "win" carry a much more positive value. The issue here, which ultimately is an issue for the algorithm, is that this sentiment is one of hope and desire rather than compliment, and I don't think the model can decipher.
A negative score example:
Lopez feels as if he's in a good place, mentally and physically. He adjusted his focus to avoid one mistake turning into the issue of a big inning, which was a problem plaguing him in 2019.
The words "avoid", "mistake", and "issue" seem to be driving this score. It will be interesting to see what happens here. If a negative sentiment score is more honest than a positive sentiment score, then it may be better to do away with positive scores altogether.
Data provided by www.MLB.com