2010 FIP Challenge Results Part II
Earlier today, in Part I of the series, I published a chart of 38 pitchers who had a difference of 0.50 or greater between their FIP and xFIP at the All-Star break and their 2nd half ERAs. Here I want to go into more detail rather than just giving a raw score for the two metrics.
In rating the two systems, I considered the metrics to recommend keeping a pitcher if at the All-Star break they were at 3.50 or lower, to listen to a trade if they were between 3.51 and 4.00, to actively look to sell the player if they were between 4.01 and 4.50 and to either sell or cut a pitcher if they were above 4.51.
Of course, we also have to consider what the pitcher’s actual ERA was at the break, too. A pitcher could still be a sell candidate if one of the metrics was significantly higher than his ERA. For these extreme cases, I considered a difference between 50-75 points to be a “listen” candidate, while above 75 to be a “sell high” guy.
Felipe Paulino – pitched in just 5.2 innings after the break. Officially a win for xFIP, but one we should probably dismiss due to a lack of playing time.
Francisco Liriano – His second half ERA was better than his first half mark, but both metrics thought he was outstanding before the All-Star break. This is a clear win for xFIP.
Anibal Sanchez – His 2nd half ERA (3.44) was a near-perfect match for his first half FIP (3.46).
Clay Buchholz – Both systems thought Buchholz was not nearly as good as he was in the first half. FIP had his as a keep while xFIP said he was an active sell. Since Buchholz did even better in the second half of the season than the first, this was a clear victory for FIP.
Josh Johnson – Both systems had Johnson as a keeper, but xFIP did a better job predicting his 3.50 post All-Star break mark.
Daisuke Matsuzaka – After allowing 4 HR in 71 IP in the first half, Matsuzaka served up 9 HR in 82.2 IP after the break. Big win for xFIP.
Johan Santana – His 3.00 ERA in the second half almost identical to his first half mark of 2.98. xFIP had Santana as a cut, so an easy win for FIP.
Jason Vargas – Eight of his 14 starts after the break came on the road and he allowed eight of his 10 second half HR away from Safeco. Big win for xFIP.
Justin Verlander – In the last three years, Verlander has posted an ERA over 5.50 in the month of April. He was terrific from May 1st through the end of the season again in 2010. Easy win for FIP.
Barry Zito – A lousy second half of the season made Zito a spectator for the Giants in the post-season. xFIP did an outstanding job predicting Zito’s collapse.
Ubaldo Jimenez – Just like with Zito, xFIP was just about perfect predicting Jimenez in the second half.
Tom Gorzelanny – Both systems saw Gorzelanny as a pretty good pitcher, but xFIP came closer to his second-half collapse.
John Danks – Again, the crystal ball for xFIP was right on target for Danks.
Tommy Hanson – In both seasons in the majors, Hanson has outperformed his xFIP. He turned it up a notch in the second half of 2010, thanks as much to his .233 BABIP as his 7 HR in 100.1 IP.
Matt Cain – Another pitcher with a history of outperforming his peripherals, Cain beat his FIP by nearly a run and his xFIP by nearly two runs in the second half of 2010.
Clayton Kershaw – Just as good in the second half of the season as he was in the first.
Cliff Lee – Many people wanted to eliminate Lee from this study last year, as he went from a pitcher’s park to a hitter’s park. But Lee outpitched his xFIP after the break in 2009. No such luck for Lee this year while following a similar story line of moving to a tougher park for pitchers.
C.J. Wilson – Like Kershaw, he was remarkable consistent between halves and ended up as a win for FIP.
Livan Hernandez – Those of us who kept predicting the bottom to fall out for Hernandez in 2010 are still waiting. Meanwhile, his second half ERA of 4.02 was a perfect match for his first half FIP.
Doug Fister – I imagine even the staunchest FIP supporters were shopping Fister every chance they could.
Fausto Carmona – After back-to-back seasons with a BB/9 over 5.00, Carmona allowed just 29 BB in 94.0 IP after the break last year. That had more to do with it than HR rate (10 HR in 94 IP) for why FIP was a clear winner.
Gavin Floyd – Fantasy owners did not know start from start what to expect from Floyd, but xFIP did a nice job of predicting his second half ERA.
Mark Buehrle – Like Wilson and Kershaw, Buehrle was a model of consistency with his ERA between halves this year. However, I worry about his K/9 rate and would be shocked if he was on any of my teams next year.
Brandon Morrow – With the Blue Jays out of the race, they decided to shut down a healthy Morrow after his first September start, which limited him to 46.1 IP after the break. A polar opposite to Buehrle, Morrow posted a 10.95 K/9, up from 8.14 a season ago.
Johnny Cueto – Cueto’s second half ERA was in the range predicted by his first half FIP and xFIP. But it was just 0.05 away from his FIP.
John Lackey – His second half ERA of 3.97 was a nice match for his 3.89 career mark but I doubt that makes too many Red Sox fans happy about his season and the team’s remaining obligation to him.
Kevin Correia – The first player on our list to have a big discrepancy between his FIP and xFIP due to a high HR rate, Correia did not show much regression in the second half of the year. Those who thought he would rebound, especially considering his home park, were disappointed. Correia allowed 13 HR in 82.1 IP in Petco this year.
Cole Hamels – Just like in 2009, one pitcher from the high HR rate side completely turned things around to become one of the best pitchers in the second half. Hopefully, Hamels has more luck in 2011 than 2009’s entry did. Rich Harden had a 5.58 ERA in 20 games with the Rangers this year after having a tremendous post-break performance (2.55 ERA) in 2009.
Nick Blackburn – Both systems would have advised cutting Blackburn, who responded with a 3.94 ERA in the second half, as he showed a big across-the-board improvement, including a microscopic 1.83 BB/9.
Kevin Millwood – Most of the players with a high HR/FB rate come back as not worth the risk by both systems. But xFIP said Millwood was significantly better than he showed in the first half and did an excellent job projecting his post-break ERA.
Jeff Karstens – An sore shoulder led to just one appearance in September. I regret the pain suffered by Mr. Karstens but it’s probably just as well that it played out that way.
Zach Duke – Hard to believe he made the All-Star team in 2009. Since then he is 11-23 with a 5.52 ERA.
James Shields – Both systems predicted a big bounce-back performance in the second half by Shields but that never materialized. He continued to give up HR by the basket, saw his K/BB ratio drop by over a full point and saw his BABIP increase to .362 after the break.
Brian Bannister – Limited to 25.2 IP in the second half due to rotator cuff tendinitis.
Randy Wolf – Both systems saw Wolf as waiver wire fodder but he had a 2.67 ERA in his final 13 starts, which was right after I placed him on waivers in a dynasty league.
Doug Davis – Elbow tendinitis kept Davis from pitching after the All-Star break.
Ricky Nolasco – FIP was nearly perfect with its Nolasco forecast, a marked departure in recent history, as he has generally underperformed his peripherals the past two seasons. Of course, Nolasco pitched just 47 innings after the break due to knee surgery.
Ian Kennedy – Neither system thought much of Kennedy going into the break but xFIP had a brighter outlook. Meanwhile, Kennedy put it altogether after the All-Star game, with Quality Starts in seven of his last nine outings. He finally started pitching well in his home park. In his last three games in Chase Field, Kennedy allowed 4 ER (0 HR) in 19 IP.
*****
When I started this comparison in 2009, my belief was that you would be just as well off using either system. After last year, there was a definite raw advantage for xFIP but now with two years worth of data, the two systems are basically even. Overall, there have been 72 pitchers who’ve had a 0.50 or greater difference between their FIP and xFIP at the All-Star break. Here’s how they did if you used their first half FIP or xFIP to project their second half ERA:
xFIP – 37
FIP – 34
Push – 1
Both systems have strengths and weaknesses. Generally speaking, xFIP does a better job with non-elite pitchers with low HR rates while FIP does a better job with elite hurlers. So, if a Tom Gorzelanny is cruising along with a sub-7.0 HR/FB rate, it appears you should look to sell high. But if it’s Justin Verlander, perhaps you should hold onto him.
We know that over the long haul that xFIP is the better metric to use for most pitchers. The issue here is that for one season (or one partial season) there may not be enough time for regression to fully kick in. Let’s look at Tim Lincecum. In the first half of 2009, he had a 3.9 HR/FB rate. In the second half of the season he had a 7.5 HR/9. This year he had a 9.9 HR/FB ratio. He has been regressing towards a normal HR/FB rate since the first half of 2009. But it did not all come in the same season.
Readers have suggested using first half ERA, or a mid-point between first half FIP and xFIP or an average of all three to see which one best predicted second half ERA. I think these are worthwhile suggestions and perhaps ones that we can use in the future (going back retroactively, too) as our sample size increases. We can also eliminate pitchers who did not pitch substantial innings and look for other trends and anomalies as our population gets bigger.
This started with a claim by my friend and colleague Derek Carty that FIP was basically useless for fantasy purposes with other metrics like xFIP available. Right now it appears FIP is making a case not to be tossed into the trash can by fantasy players.