Seattle Mariners at New York Yankees, 1:05 p.m. ET
- Mariners (James Paxton) +183
- Yankees (Luis Severino) -205
- Total: 7.5
Bet to Watch
It's easy to look at Paxton vs. Severino this season and think a 1-0 game is looming. Severino not only has an ERA just over 2, but his numbers in day games this year are absurd: 6-0 with a 1.07 ERA and 46/10 K/BB ratio.
Meanwhile, despite a lackluster previous start, Paxton has developed a reputation as Seattle's ace, and his underlying numbers and predictive metrics support that case. He should have a very good rest of the season.
This was the most interesting game I handicapped Wednesday night for Thursday, for a few reasons. First off, my model spit out a slightly surprising run total of 8.31. That's never the end for me (it won't just trigger a play automatically), but it does generally tell me where I should be spending my time. I think this is a situation where the reputations of the starting pitchers have overwhelmed what an "average" game between these teams would be like, and the number is set artificially low by a small margin to balance the action a bit more.
There might be a little variance in lineups with the day game to end the series, but I proceeded with the "standard" Yankees lineup vs. left-handers and Mariners lineup vs. righties. I'm assuming all bullpen pitchers for both teams are available except for Chasen Shreve of the Yankees (who does not merit a significant adjustment either way).
Let's circle back to those Severino numbers I mentioned earlier. His ERA is a hair over 2, but his xFIP is actually three-quarters of a run higher than that, and even if it wasn't, the likelihood he holds his current numbers for the year is very low. Regression seems to be on the horizon, at least a little. Also, splits such as home-road and day-night are incredibly dangerous in small sample sizes. Severino has dominated day games this year, but in the past three seasons, his ERA was a full run higher than it was at night. Did he suddenly learn to love the sunshine, or is that statistic more noise than signal? I think that needs to be ignored.
