30-game winner Paul Skenes?! A new formula to bring pitcher wins back to life

There have been 2,664 pitchers who have made at least 30 career starts since 1901.

Three of those pitchers -- or one out of every 888 -- own a career ERA below 2.00. Two of them are Hall of Fame deadball era greats: Ed Walsh (1.82) and Addie Joss (1.89). The third is Pittsburgh Pirates superstar Paul Skenes.

The chances of Skenes, who has made just 39 career starts, remaining in that class are slim. That's nothing against him. It's the reality of math and the era in which he plays. The careers of Joss and Walsh overlapped in the American League from 1904 to 1910, when the aggregate ERA was 2.61. The collective ERA in the majors since Skenes debuted is 4.04.

This season, Skenes' 1.85 ERA leads the majors, and he's first among all pitchers in bWAR (4.4). The latter figure is actually tops among all National League players, period. The current numbers generated by my AXE system and the futures at ESPN BET both mark Skenes as a solid favorite to win his first NL Cy Young Award.

Incidentally, Skenes' won-loss record for the woeful Pirates is a meager 4-6. Should we care?

Won-loss records for pitchers are no longer part of the evaluative conversation, so if your response to the previous question was "no" then congratulations for paying attention. If your response was anything else, then it's almost certainly because you're in a fantasy league that still uses pitcher wins, not because you think Skenes' record actually tells us anything about his true value.

But what if I could tell you this and prove it: Skenes' real won-loss record is 11-5, the win total tied for the third-most in the majors. I'm going to explain how I got there, but first, let me explain why I think it matters.

Just to illustrate how starting pitchers were written about for most of baseball history, I pulled up the 1980 MLB preview from the Sporting News and went to the page where the Pirates (defending champs at the time) were analyzed. Here's a bit on their pitching:

"The Pirates last year won without a 15-game winner. The staff won in bunches. Five pitchers won 10 or more games."

There were no other pitching statistics in the staff outlook. No ERAs, no strikeout rates, nothing about walks. This was it. This is just how pitchers were discussed back then.

It's good that we understand how to assess pitchers now at a deeper level and, even back in 1980, people like Bill James were already doing it. But pitching wins still meant something as one of the baseball statistics James might allude to as having achieved "the power of language."

That is: To describe a pitcher as a 20-game winner had real meaning. It was an avatar for quality, and if someone was a five-time 20-game winner, that was an avatar for greatness.

Pitcher wins have always been an imperfect measure, but its flaws have ballooned over time as the game and the responsibilities of the starting pitcher have evolved. Last season, 41.3% of decisions went to relievers. One hundred years ago, that number was 18%.

A good win statistic clears away a lot of contextual noise. In every game, you have two starting pitchers, on opposing teams, pitching on the same day, at the same ballpark and in the same weather conditions. While starters will never admit they are competing against each other ("my job is to get the opposing lineup out" is the standard refrain), they actually are. Their job is to pitch better than the other pitcher, because doing so means giving up fewer runs than him and, if you do that, you win. Well, at least before the bullpens get involved, but a good win stat would filter out that factor, too.

Take anyone who has ever pitched for the Colorado Rockies. The Rockies have been around for more than 30 years and it's still exceedingly difficult to make heads or tails of their pitchers because so much of their data has to be greatly adjusted for ballpark context. And, while park effects are necessary and sophisticated, they are also estimates.

The Rockies have never had a 20-game winner. The closest was Ubaldo Jimenez, who won 19 in 2010, when he also became one of two Rockies starters to top 7 bWAR. (The other was Kyle Freeland in 2018.) Jimenez is Colorado's career ERA leader as well, with a mark of 3.66. Every other qualifying Colorado starter in franchise history is at 4.05 or above.

Thus, when we talk about the best pitchers of the current era, Rockies pitchers are almost always going to be left out of the conversation. Their numbers just don't seem telling or comparable.

This is where a better win statistic would be so useful. Because whatever the precise effects Coors Field might have on a game's statistics on any given day, a good win stat would be comparing two starters on that field in almost exactly equal conditions. If we do it that way, maybe the Rockies do get some 20-game winners on their ledger.

Is such a win stat possible?

For me, the pitcher win should strictly be the domain of a starting pitcher. This dictum is clouded by the use of openers to start games and bulk pitchers who are used like starters but just not at the outset of games. For now, let's try not to think about that.

The question about each game I want to answer is this: Which starting pitcher was better in that game? The starter who becomes the answer to that question gets the win; the other gets the loss. And that's all. It's as simple as that. Every starter in every game gets a win or a loss and no-decisions don't exist.

Well, the no-decisions would still exist, because I'm not proposing that we erase traditional won-loss records from the books. There's too much history attached. Early Winn is remembered in part for clinging to his career in pursuit of 300 wins, and he finished with that number exactly. Cy Young is remembered for his unbreakable career record of 511 wins. Likewise, Jack Chesbro's claim to immortality is that he owns the modern single-season record of 41 wins. We don't want to erase those things -- we want to add to our understanding of starting pitchers.

Something I've proposed on a number of occasions is to use James' game score method to assign wins and losses. In fact, I've tracked game score records for several years and for this piece, I expanded my database back to 1901 to see how the historical record might look.

There are other game score methods, but I like James' version for its simplicity, though the modified version created by Tom Tango for MLB.com has the same virtue. With either, you can look at a pitching line and easily calculate the game score in your head, once you've got the formula down. (If you can't do that calculation, study more math.)

I also would try to account for short, opener-style outings. I use James' version but dole out a heavy penalty for going fewer than four innings. To avoid ties -- when the starters end up with the same game score -- you can give the W to the starter on the winning team.

Awarding pitcher wins like this isn't perfect. The conditions for the starters aren't truly equal because the quality of the lineups they face won't be the same. When Skenes beat Yoshinobu Yamamoto earlier this season, for example, his task against the Los Angeles Dodgers' lineup was a bit more difficult than Yamamoto's figured to be against Skenes' teammates. Likewise, the quality of the defenses behind opposing starters won't be the same in any given contest.

Despite those disparities, the mandate for both starters is identical: Out-pitch the other guy. And you know what? The game score method of assigning wins and losses to assess the success of that assignment works pretty well.

Let's call a game score win a GSW and a game score loss a GSL. Do you know who owns the single-season record in GSW?

It's Chesbro, still. In fact, his 1904 feat looks just as impressive by this method. Here are the top five seasons by GSW:

Jack Chesbro, 40-11 (1904) Christy Mathewson, 35-9 (1908) Iron Joe McGinnity, 34-10 (1904) Mathewson, 34-12 (1904)

Ed Walsh, 34-15 (1908)