Here are the top picks my programs produced for use in Beat the Streak. This post mostly explains the ideas behind the calculations. In addition, this post shows tests on the Neural Network (NN). This post discusses an NN that includes the ballpark. I recently update the models, and the results of those tests are here.

First, the Log5 Method picks:

0.356 — Starling Marte batting against Luis Perdomo
0.335 — Adonis Garcia batting against Wily Peralta
0.332 — Ender Inciarte batting against Wily Peralta
0.331 — Josh Harrison batting against Luis Perdomo
0.326 — Matt Kemp batting against Wily Peralta
0.324 — Nick Markakis batting against Wily Peralta
0.321 — Freddie Freeman batting against Wily Peralta
0.314 — Adam Frazier batting against Luis Perdomo
0.310 — David Freese batting against Luis Perdomo
0.310 — Jeff Francoeur batting against Wily Peralta

Note the weakness of the Log5 method in this case. Luis Perdomo is so bad, someone like Adam Frazier, with a short put productive career, makes the list. Log5 is telling us to pick against the pitcher, with two very bad ones on the mound this evening.

The NN produces this list:

0.356, 0.755 — Starling Marte batting against Luis Perdomo.
0.275, 0.742 — Jose Altuve batting against Hector Santiago.
0.285, 0.737 — Daniel Murphy batting against Trevor Bauer.
0.304, 0.732 — Xander Bogaerts batting against Luis Severino.
0.331, 0.730 — Josh Harrison batting against Luis Perdomo.
0.277, 0.725 — Dee Gordon batting against Matt Moore.
0.301, 0.725 — Mookie Betts batting against Luis Severino.
0.332, 0.724 — Ender Inciarte batting against Wily Peralta.
0.267, 0.722 — Paulo Orlando batting against Chris Sale.
0.335, 0.721 — Adonis Garcia batting against Wily Peralta.

Note that the NN appears to be more concerned with batter ability to get hits than pitcher ability to allow hits. So Paulo Orlando makes the list against Chris Sale, and Jose Altuve makes the list against Hector Santiago.

The NN with park included modifies the list like this:

0.356, 0.761 — Starling Marte batting against Luis Perdomo.
0.275, 0.742 — Jose Altuve batting against Hector Santiago.
0.331, 0.738 — Josh Harrison batting against Luis Perdomo.
0.285, 0.737 — Daniel Murphy batting against Trevor Bauer.
0.304, 0.736 — Xander Bogaerts batting against Luis Severino.
0.332, 0.732 — Ender Inciarte batting against Wily Peralta.
0.301, 0.730 — Mookie Betts batting against Luis Severino.
0.335, 0.730 — Adonis Garcia batting against Wily Peralta.
0.277, 0.727 — Dee Gordon batting against Matt Moore.
0.267, 0.724 — Paulo Orlando batting against Chris Sale.
0.284, 0.724 — Martin Prado batting against Matt Moore.

Harrison moves ahead of Murphy, and the Braves hitters move up a bit. Starling Marte is still the unanimous first choice.

Let’s take a quick look at the two players who finish second. The Parameter lists look like this:

• Parameters: [Pit 2016, Pit 2014-on, Bat 2016, Bat 2014-on, ML 2016, Park 2014-on]
• Jose Altuve: [‘0.205’, ‘0.206’, ‘0.310’, ‘0.304’, ‘0.232’, ‘0.244’]
• Adonis Garcia: [‘0.329’, ‘0.289’, ‘0.253’, ‘0.255’, ‘0.232’, ‘0.234’]

Garcia is an interesting player. He has a decent batting average but seldom walks. That makes him a good choice for getting a hit streak going, since he gives himself plenty of opportunities to get hits, and has a decent chance of succeeding. Note that Altuve is so talented that even against a very good pitcher and defense, he still posts a high probability of success. The NN is telling us that the log5 method may not be that even Steven. Log5 rates the Garcia/Peralta matchup very high, but the NN regressions says no, the batter figures in more highly.

As always, your best picks will fail to get a hit about 25% of the time. Good luck!