The View From Wisconsin

When you are attempting to measure the performance of a pitcher, you generally look at three key areas – decisions, earned runs and baserunners. Usually, a pitcher who does well in each of these three areas will be considered a heavy favorite for the Cy Young Award at the end of the season.

Statheads like myself have always had a nagging feeling that these three areas could be combined in some manner to produce a formula that would give some sort of measure of a particular pitcher's performance over a given time frame – month, season, career, whatever. Instead of just dealing with ERA and Wins, we want something more, something that gives us the overall picture in one stat or set of stats.

The Rotisserieans of the 1980's, in creating the game that has now spanned the globe to cause the fantasy sports explosion, used a little-appreciated tool to measure pitching performance. They gave it the less than elegant name "WHIP Ratio", indicating what exactly it measured: the ratio of walks and hits to innings pitched. Others have tinkered and toyed with it (Baserunners per Nine innings, for example), but the basic formula is powerful enough.

WHIP Ratio: (H + BB + HB) / IP

In trying to develop that "holy grail" of a pitching statistic, you realize quickly that you could very easily use winning percentage, ERA and Ratio to determine a pitcher's Efficiency – that is, the level of pitching performance you would expect of him in each of the three areas. Unfortunately, Winning percentage is a "positive" stat (you want it to be larger, not smaller), so to make it where all three measures are the same, we have to subtract it from 1 – essentially creating a "losing percentage." Average all three together and you have what is called Pitching Efficiency Ratio.

Pitching Efficiency Ratio: (1 – Winning Percentage + ERA + Ratio) / 3

Now, you should realize that PER won't be enough to separate the wheat from the chaff. If a pitcher didn't allow any baserunners in his only appearance, he'd have a PER of .333, unless he got credit for the win – in which case, it'd be .000. For example, Kenny Greer pitched in one inning for the New York Mets in 1993, striking out two in one inning without allowing a baserunner. He got the win, which means his PER is 0. That doesn't tell you much about his performance – other than the fact that he had one good inning in one game.

So, to compensate, we need to pair PER with a stat that shows some form of longevity – a counting stat that indicates both performance and sheer numbers. The best such stat for a pitcher is, obviously, wins. So, we divide wins by PER to get what I initially called Pitching Performance Points. The theory, of course, is that the more wins a pitcher has, and the lower his PER, the higher his PPP score would be.

Only one problem, though – it doesn't work well for pitchers whose win total doesn't reflect their contribution to their team. Take the following example from 1990:

• Doug Drabek 22-6, 2.76, 33 GP, 231.1 IP, 190 H, 56 BB, 3 HB, 1.076 Ratio, 1.351 PER, 16.3 PPP
• Dave Stewart 22-11, 2.56, 36 GP, 267.0 IP, 226 H, 83 BB, 5 HB, 1.176 Ratio, 1.357 PER, 16.2 PPP

According to our simple formula, Drabek had only a slightly better total (16.3) than Stewart (16.2) – but Drabek was clearly better with fewer losses, a better ratio and a lower PER. The forumla also doesn't work in Greer's case – one win, divided by zero, means an infinite result. Just using wins isn't enough; you need to express a pitcher's winning percentage as part of the our "performance counting stat."

Fortunately, Bill James actually came up with a way to do exactly this in his book, Whatever Happened To The Hall Of Fame? He stumbled upon a statistical anomaly of sorts, which he called Fibonacci Wins. His theory is, pitchers whose winning percentage is greater than .618 (the Fibonacci number) have a better Fibonacci Win total than someone with the same number of wins and more losses. Or, a 22-6 record is definitely better than a 22-11 record. The formula is simple:

Fibonacci Wins: (W x W%) + (W – L)

In our example, Drabek had 33.3 FW in 1990; Stewart had 25.7. This translates to a 24.6 PPP for Drabek, and 18.9 for Stewart. Though both had phenomenal years in 1990 – probably two of the best seasons a pitcher has had in the last two decades – Drabek was a distinctly better pitcher than Stewart that season.

There is one more stumbling point in our PPP formula – relief pitchers. Even the best reliever in the game won't get more than a handful of wins and losses in a season – if that. For them, the use of the Relief Points formula (the one used to determine the Rolaids Reliever of the Year Award) is the most logical choice.

Relief Points: (3 x SV) + (2 x W) – (2 x L) [– (3 x BS)]

A short note: because Blown Saves are a recent statistic, using them is optional in this formula. As long as you are consistent in how you figure the formula, you should be fine.

To strengthen the numerator in our PPP formula, we add Relief Points to Fibonacci wins, divide by two to average them, and then divide the result by PER. For individuals like Mr. Greer, all we need do is just divide the sum of his Relief Points and Fibonacci Wins by two. This formula actually works quite well for pitchers from the 19th Century, as the totals do a good job of expressing pitching performance of both "pinch" pitchers and your starting workhorses.

There is a distinct problem, however, in the modern era. Since 1951, the five best single-season pitching performances, using this formula, are Dennis Eckersley, 1990 (146.7); Jose Mesa, 1995 (104.3), Eric Gagne, 2003 (96.5), Trevor Hoffman, 1998 (94.0), and Eckersley again, 1992 (89.7). Care to guess what happened when we added Relief Points? We "over-weighed" our PPP formula in the favor of Relief Pitchers.

So, we need to adjust the totals so that all pitchers are on the same footing, in relation to the contribution towards a team's succes. The easiest way to do this is to adjust Relief Points downwards, based on the number of innings in which a pitcher has appeared. Thus, we take Relief Points, multiply by innings per game, and divide by nine – essentially adjusting RP by the percentage of a typical game that the pitcher appears therein. This resulting formula is one we will call Cy Young Points, as it is a relatively accurate measure of whether or not a pitcher is a credible candidate for the award in a given season.

Cy Young Points: [(RP x IPG/9) + FWP] / PER

In analyzing the data from every season going back to 1871, our previous iteration of the formula actually works better for pitchers from the pre-"Modern" era of 1901 to present. When we analyze individual players from those years, the innings per game adjustment is unnecessary, since pitchers completed games on a regular basis. That began to change as we entered the 20th century.

The best CYP season in the last 50 years, not surprisingly, was Denny McLain's 1968 season where he went 31-6 for the Tigers. With a PER of 1.013, his CYP total was an astounding 95.2. The only pitchers to come close since have been Ron Guidry (1978, 88.4), Dwight Gooden (1985, 85.9) and Greg Maddux (1995, 73.3). Sandy Koufax's stellar 1963 season (25-5, 306 strikeouts, 77.1 CYP) is the fourth best total since 1955. Koufax is the only pitcher to have four 50+ CYP seasons in the last five decades; from 1963 to 1966, he had CYP totals of 77.1, 58.0, 70.9 and 70.7. Juan Marichal is the only one to have more than three.

The obvious question, of course, is what pitchers have the highest career CYP totals. It'd be a pretty good guess if you said Cy Young himself; he is one of only three pitchers to have more than 600 CYP in his career with 655.0. Unfortunately, he is not the leader in his own category. That honor goes to Christy "The Big Six" Mathewson, who translates a 373-188 record with 28 saves, a ratio of 1.071 and a PER of 1.180 into a CYP total of 688.9 – the best ever in major league history. Interestingly, Al Spalding's totals over seven seasons comes in just behind Mathewson and Young, with 653.4. Spalding's numbers are magnified by his small 65-loss total.

The rest of the top 10 list isn't much of a surprise: Walter Johnson is fourth with 571.3; Grover Cleveland "Pete" Alexander is next with 560.7. The rest: Eddie Plank, 465.8; Kid Nichols, 463.3; Mordecai "Three Finger" Brown, 456.1; Robert "Lefty" Grove, 454.1; and John Clarkson (the 1880's pitcher), with 453.2. The highest modern-day pitcher is right behind them at 405.2, one Roger Clemens. Greg Maddux isn't far behind, with Tim Keefe (394.8), Warren Spahn (374.5) and Charles "Old Hoss" Radbourn (374.3) between him (352.8) and Clemens. The best true reliever, Hall-of-Famer Dennis Eckersley, is down at 22nd (319.3), wedged between Jim Palmer (325.9) and Tom Seaver (319.1). Randy Johnson is the only other current player with 300 or more career CYP, sitting at 26th with 302.1.

The worst pitcher in baseball history, according to CYP, was John "Happy" Townsend. He posted a 9-6 record in his rookie season of 1901 for the Phillies, with a decent 7.8 CYP total in 19 games – and never again had a positive CYP total in his remaining five seasons in the majors. He went to Clark Griffith's Washington club in 1902, where he promptly lost 16 games, had a 4.45 ERA, and posted a –7.5 CYP total. He followed that up with a 2-11 record in '03, with a 4.76 ERA, a 1.595 Ratio, and a –8.9 CYP.

That wasn't the worst of it, though: in 1904, he went a ridiculous 5-26 for the Senators, though his ERA dropped by a run. His CYP for the seasons was –29.5, the second-worst single-season total in baseball history. The worst came one season later, when Fred Glade posted a 6-25 record with the Saint Louis Browns. Coupled with a 1.602 PER, his CYP total of –33.8 is still the single-season record.

After one more terrible year with Washington (though not as bad as his 1904 totals), Griffith traded him to Cleveland in 1906. The new surroundings of League Park didn't help, as Townsend lost 7 of his 17 apperances. He was out of the majors by the end of the year. He would live to the ripe old age of 72, dying in his home state of Delaware.