This is interesting. HUGE disclaimer that I'm not handicapping baseball at all this year...real life getting in the way. I hope to be back by the playoffs. That said, this sorta thing is how I think about baseball betting, so I'll quibble with it for the sake of discussion and without putting enough work in to know if I'm right or wrong.
I like that he's using xFIP to evaluate pitchers, it's the best opponent/field neutral metric we have. Where I'd quibble is why he needs full game run lines and full game pitching stats at all. You can presumably build a 1st inning probability model as he has here and keep it agnostic of the full game numbers, it's simply "is the probability different enough from the betting odds to make a bet?". I don't want the noise of two pitchers that are theoretically exactly the same in the first inning but different in the 6th b/c one is a veteran and gets more rope and one is young and gets pulled at the first sign of trouble. The full game market efficiently accounts for this, I just don't know why I'd care to include it when a simpler calculation would suffice. I'd also wonder about first time through the order vs literal first inning stats, but if he put this much work in what he landed on has to have performed better on a back fit or maybe they're the same and larger sample size wins. The hitting side is a little more obvious, wRC+ is a fine choice and first 4 or 5 hitters are fine with me depending on what back fits better.
I'd also say that if he's seeing inefficiency early and that motivates making bets before he knows lineups then something about the market has changed/adjusted. I was never a big first inning guy, but the guys I know that are used to wait until nearly first pitch b/c they were always betting yes and betting no was a bit of a square side. You'd pretty literally never lay favorite juice on a yes and also pretty literally never bet a no, so this is probably a reflection of the exotic markets getting more efficient as bettors get smarter.