What are they and how do we use them?
There are so many stats for baseball and not all of them have value to predicting outcomes. The standard batting average, RBI, and ERA stat for example fit into that category. Stats are valuable when their context has meaning. For handicapping, we need to know who will win the game and how many runs will be scored. A batting average, RBI count, or ERA will not provide us context leading to the outcome of winning the game. This is where the sabermetric stats come into play. As a rule, sabermetric stats are designed to either provide value a player has to team wins, or measured quantity of value to teams wins when compared to other players. The key is they are associated to wins which is what we are trying to determine.
In writing about 3 True Outcomes, I used definitions for FIP and xFIP. Let’s revisit them now. FIP is Fielding Independent Pitching. Its definition: FIP takes a pitcher’s strikeouts, walks, and home runs allowed and translates them into a number scaled to ERA. Think of it as what the pitcher’s ERA should be if the defense behind him turned batted balls into outs at a major-league average rate. Ideally, FIP measures how dominant a pitcher is at limiting baserunners and scoring chances for the opponent. xFIP: is a predictive stat estimating the pitcher’s future FIP based on his current pitching data. It is a great tool for looking at upcoming games. It normalizes a pitcher’s home run rate based on his flyball percentage vs actual home runs hit. This predicts how a pitcher’s FIP will be over his next several starts. SIERA is Skill Interactive Earned Run Average. SIERA: quantifies a pitcher’s performance by trying to eliminate factors the pitcher cannot control by himself. But unlike a stat such as xFIP, SIERA considers balls in play and adjusts for the type of ball in play. Whereas FIP and xFIP are not utilizing balls in play, SIERA does, thus SIERA is the most accurate predictive model between the three. They all appear like the way an ERA looks. SIERA tells us a couple of things. Strikeouts are good, even better than FIP suggests. Walks are bad, but not that bad if you do not allow them to score.
What do they tell us? It is important to know the variation between FIP and xFIP numbers. You should always have them listed side by side or together. The reason is FIP tells the value of a pitcher independent of the defense and xFIP tells you about the ability level of the pitcher and what should be expected. We can see regression and progression here which is what we are looking for. The next charts show two pitchers’ game logs. CLE Shane Bieber (who was the #1 pitcher based on WAR) and SDP Zach Davies (who was the #25 pitcher). I arbitrarily chose #1 and #25 to see what if any the variance in FIP and xFIP was for the two. First is Shane Bieber.
Second, this is Zach Davies.
Using the scales, Bieber was better than excellent! His xFIP is also lower than his FIP. It means his talent and ability is greater than his performance. Zach Davies is interesting. His xFIP is higher than his FIP which means his talent and ability is lower than his performance. He is also showing up in the average area for FIP and below average for xFIP. Then compare his ERA to these results. No matter you slice it up, Davies, although a good season, was not an above average pitcher. This is where we need to investigate SIERA. Remember, FIP and xFIP are largely good for strikeout pitchers and Davies is not that. He has balls in play often so should look at these indicators. Using SIERA, Bieber had a 2.52 and Davies had a 4.32. Again, we see Davies struggle with a below average rating in SIERA.
The reason I provided a game log for both pitchers is for you see the variation form start to start. We have access to this information! We can see what the pitcher’s predictive stats are prior to the game. HINT: THIS IS EXTREMELY VALUABLE INFORMATION. Look at the games Davies pitched. He threw 12 games. Our predictive stats tell us he should allow 4 runs on average in those games but we know his performance was better than his ability so we should assume 3.5 – 4 runs he will allow on average. He allowed 3 or more runes 7 times out of his 12 starts. We can use this to our advantage! Bieber in the same sense should only allow 2 runs or less. He started 12 games. He allowed 2 or less runs 9 times!
To express how valuable this knowledge is means on a scale of 1-10, this is a 10! You cannot properly evaluate a baseball game without this knowledge. I use this on a rolling scale. I want to know how a pitcher is throwing today, and a recent history. I use 1 moth of starts which equates to 5-8 starts. I use their FIP, xFIP and SIERA just as described. This tells me an almost complete pitching story with accuracy! I go further into detail with more areas of study, but this is a meat and potatoes part of the equation.
Hope this was helpful. I am available for questions regarding any of my articles or need clarification of any of the data I use. Ask lots of questions!