THT Mailbag: Offseason edition

In other news, teams win more when they play better

During Game 1 of the World Series, it was mentioned that Kazuo Matsui was batting significantly better when the rockies won then when they lost. This led the announcers to state that Matsui was a “key” to the rockies success. I was wondering how rare this is, and what the batting splits in wins/losses look like on average.

– Tim M., Atlanta, GA

John Walsh: You sound like you’re skeptical about Matsui’s split being noteworthy. Obviously, a team will tend to win the games when their offense hits well and lose those when they hit poorly. To see how big of an effect it really is, I looked at the average batter performance broken down by whether his team won or lost. The results for the year 2000:

| WinLoss | PA    | AVG   | OBP   | SLG   | OPS   |
| W       | 98101 | 0.306 | 0.387 | 0.509 | 0.896 |
| L       | 92160 | 0.234 | 0.306 | 0.364 | 0.670 |

As you can see, there’s a big difference, over 200 points of OPS. I don’t know what Matsui’s numbers were this year, but in any case, it’s safe to say that most batters hit significantly better in games their teams win. You were right to be skeptical.

David Gassko: Matsui had an .829 OPS in Rockies’ wins this season and a .616 OPS in Colorado’s losses, which is actually a slightly smaller split in OPS than what John found. In other words, the idea that he was the key to Colorado’s success is, not surprisingly, bunk.

Merkle’s boner

If a runner is on third, and there are two outs, if the batter grounds out, it’s the end of the inning. If the runner on third touches home plate before the batter is thrown out at first base, the run does not count because it was a force play at first and the inning ends. I know this to be true.

Given that, I have a hypothetical situation: Tie game, bottom of the 9th, bases loaded, two outs. The hitter hits a clean single to left field, the runner on third scampers home easily. The runner on second is the jubilant type and instead of running to third, he runs halfway and starts jumping up and down in celebration … maybe he even runs towards home plate to go hug the winning run.

Hypothetically, couldn’t the left fielder throw the ball to third base and force that runner out? I mean sure the run crossed the plate first, but it’s still a force out isn’t it? Wouldn’t the inning end and the “winning” run not count? I’ve sure this has happened in the history of walk-off baseball … someone on base didn’t run to the next base on a walk off hit. Why don’t the fielders force them out and wipe that run off the board?

This has been bugging me recently.

– Andrew H.

Sal Baxamusa: You’ve just described “Merkle’s boner,” one of the most famous plays in early baseball history. Late in September 1908, the Chicago Cubs and the New York Giants were locked in a first-place tie coming down the stretch. On Sept. 23, the teams played each other at the Polo Grounds. Giants ace Christy Mathewson held the Cubs to one run, but the opposing pitcher, Jack Pfiester, had matched him through eight innings. In the bottom of the ninth inning, Al Bridwell singled home what looked to be the winning run with two outs. But the runner on first base was the 19-year-old Fred Merkle, who, upon seeing the game-winning hit, turned around and headed to the clubhouse to celebrate … without touching second base.

An alert Johnny Evers, the Cubs second baseman, called for the ball—although it is still disputed whether the ball thrown to him was the actual game ball or not, as fans had already mobbed the field—touched second base, and Merkle was ruled out on appeal. The game was declared a tie.

The Cubs and Giants ended the season in a tie, and the Cubs won the one-game playoff to determine the pennant. Had Merkle bothered to touch second base, it’s very possible that the Cubs would not have even played in the 1908 World Series, which would have made 2007 the 100th anniversary of North Side futility.

Has it happened since? I’m not sure. But given Merkle’s story, no ballplayer should take the game-winning base for granted!

Carpe diem

There’s not much competitive pressure to get better in the National League. Will a powerhouse&like maybemdash;a healthy Phillies team—step up and force other teams to get better? All these 88 win teams don’t do much for me.

– Steve, Boston

Using Recurrent Neural Networks to Predict Player Performance
Technology is rapidly advancing possibilities in decision-making.

David Gassko: The problem is that most of the big-money teams live in the American League. Of the top-five teams in payroll this year, only one, the Mets, played in the NL. That’s a large part of the reason that the AL is the stronger league right now. Premier free agents tend to come from the National League to the American, and stay there.

With that said, there definitely are some National League teams playing in markets large enough that it would make sense for them to make a splash and rise above their competition. Payoffs for an improved Mets, Dodgers, or Cubs team would be huge. You can probably throw the Phillies into that category as well. The question is if any of them will do it.

The Cubs made their big splash last year and with their sale pending, they’re probably going to sit still this off-season (though I don’t think that’s a very good move for them). The Mets are unafraid to spend and with their own cable channel, they need to win year-in and year-out, but I’m not sure that the Mets can really seriously upgrade at any positions. Jorge Posada would have really helped but it looks like he’s staying with the Yankees. Johan Santana is one possibility, but he will cost a lot in terms of prospects, not to mention money.

That leaves us with the Dodgers and Phillies, both of whom could have made a huge splash and a huge upgrade by signing Alex Rodriguez. As good as A-Rod is, however, I’m not sure he would have made the Dodgers as good as the best AL teams. They would be one of the best teams in the National League, but that doesn’t really matter if you can’t compete with the American League clubs.

With A-Rod off the market, the Phillies could still establish themselves as the team to beat in the National League. This is a team full of players hitting their peak right now; no one would benefit more in the National League from adding one more impact player.


Let say a pitcher has strikes out x batters per batter faced and a hitter strikes out y times per plate appearance. I am wondering what is the best estimate of the fraction of the time this pitcher strikes out the hitter (ignoring handedness). My first guess was stupidly something like (x+y)/2. But then I thought maybe the league average would be important, because the pitcher strikes out x league average hitters per plate appearance, the hitter strikes out y times per plate appearance against a league average pitcher.

So let z be the league average strike out rate per batter faced or plate appearance (I assume they would be the same). Then maybe the batter strikes out y*(x/z) times per plate appearance, y being his baseline strikeout rate against league average pitching and how much better or worse the pitcher is compared to league average. (Of course you can think about it for the pitcher’s perspective and get the same number x*(y/z).)

Is this a good estimate? Have you guys worked this out already?

Dave Studeman: This has indeed been worked out—not by us, but initially by Bill James and subsequently by other baseball researchers such as Tangotiger. James called his version the “Log5 method” and Tango refers to his as the “Odds Ratio.” They’re just about the same thing.

Here’s how the Odds Ratio works. Let’s say you have a “true” .300 batter facing a “true” .200 pitcher in a league with an average of .250. (I’m talking about batting averages here, but I could be talking about any percentage). First, you convert each of these percentages into ratios of success to non-success. So the .300 batter’s odds are 3/7, or .429. The pitcher’s ratio is 1/4, or .25, and the league average is 1/3, or .333. You then multiply the two individual ratios, and divide by the league ratio, or:

.429 times .25, divided by .333

The result of that is .321. You then turn that odds ratio answer back into a percentage like this: .321/(1+.321) for an answer of .243. So, if a .300 batter faces a .200 pitcher in a .250 league, the expected outcome is .243.

The Odds Ratio (or, if you prefer, the Log5 Method) is one of the most powerful mathematical concepts in sabermetrics, and can (and should!) be used in a variety of situations that involve percentage matchups. Your example of strikeout rates is a great one. Another is one team’s winning percentage against another’s (though you have to vary the formula a bit in that case). You can read more about the Odds Ratio in this post.

Print This Post

Comments are closed.