Although I am only about a third of the way through my hitter projections, I am already bored and have been itching to start up on pitchers. So after getting a taste of the methodology and process I employ in coming up with projections on the offensive side of the ledger, it’s time to look under the hood of my pitcher projections. The exciting part for you is that these projections will differ from the forecasting systems much more so than the hitter projections. So there is much more room for debate and of course having the fun of being right. Without further ado, it’s time to dive into how I project pitchers.
Like the hitter projection introduction, I will go fantasy category by category and describe what metrics I project that eventually lead to the official projection.
I first project a pitcher’s batted ball breakdown, as in, his ground ball, line drive and fly ball rates. These are typically pretty stable from year to year so it usually ends up being similar to the previous season. If a big jump or decline in ground ball rate was experienced, I will look deeper at any pitch selection changes and decide how much of that, if any, is real. Usually I will regress back to career average at least in some part. For rookies, I use Statcorner.com for minor league rates, while being aware that minor league ground ball rates are higher than in the majors.
I then project HR/FB ratio. For a veteran starting pitcher, I rely heavily on the career average, completely ignoring the prior year’s mark. I always stay within the 8%-12% range for these pitchers. Even though this is a so-called luck metric that most stat-heads would argue should regress toward around 10% (or whatever league average is), it is clear that some pitchers either do have slightly better (worse) skills than others and/or get helped (hurt) by their home park. Matt Cain is the most obvious example, so I would not project him for a league average rate.
Relievers are a different animal as they seem to be able to post much lower HR/FB ratios on average, for whatever reason. Since they rarely have the sample size necessary to be confident in their true skill level, I do regress them heavily toward league average, while still acknowledging that they do likely have better HR/FB suppression skills than starters. For example, I projected Mariano Rivera for a 7% mark last year, which is the lowest rate I will project for any pitcher.
Next, I project BABIP. Like the HR/FB ratio, I do regress toward the league average (around .300), but for starters with a long track record of apparent BABIP suppression ability, I will go as low as .270 (last year’s Chris Young projection). Among relievers, Rafael Soriano received the lowest BABIP projection overall at .260. In dealing with rookies, I almost always assume about a league average BABIP and HR/FB ratio.
Moving on is BB/9. From a graph that will first be published in my Second Opinion article, we now know that young pitchers improve their control into their late 20’s, and then see steady decline. This was intuitive anyway and something I always assumed, so I have typically incorporated this logic into my projection. I also look at F-Strike% which has a strong correlation to walk rate to see if there is a disconnect.
Second to last is K/9. Pitchers on the whole peak early in strikeout rate so it is never a good idea to assume a young pitcher will keep improving his K/9. I primarily use SwStk% to look for disconnects there to see if there is some hidden upside or downside. I will also research strikeout rate spikes or drops from the previous year to see what may have been the cause, and how likely it becomes the pitcher’s new level.
Many of those projected metrics get thrown into an expected LOB% (or strand rate as Baseball HQ terms it) formula. Like the two expected BABIP formulas I use for my hitter projections, this expected LOB% formula guides me to project the pitcher’s LOB%. For young pitchers with a limited track record, I will go with whatever number this formula spits out. For veterans, I frequently decide to adjust it, taking into account the pitcher’s career average. Relievers are treated a bit differently as they actually strand a higher percentage of runners than starters. Since I also project games and innings pitched, the formula is able to differentiate whether I am projecting a starter or reliever and it adds a couple of bonus points to a reliever’s expected strand rate.
Now I could breathe again, as all the projections required to spit out my ERA projection are done. Since I’m crazy, I actually have 8 different ERA formulas in my spreadsheet. They are xERA (Baseball HQ), SIERA, DIPS, QuikERA, ERC, FIP, FIP Base Runs, and Base Runs. Some of these formulas use my BABIP and HR/FB ratio projections, others ignore them and assume league average rates. Since I do not want my official ERA projection to ignore my BABIP and HR/FB ratio projections, I have modified the xERA formula to take them into account. This now becomes my official ERA projection.
I assume that the other forecasting systems simply take some sort of weighted average of previous season’s ERAs, throw in some other factors, and then spit out a projection. On the other hand, I actually calculate an ERA based on my projected peripherals. I think this is a huge advantage and I don’t think any system actually does this. Please let me know in the comments if I am wrong!
Once I have all the metrics projected to calculate an ERA forecast, it’s easy to then project WHIP. Based on my projected BABIP and walk rate, a hits allowed projection, as well as a walks total, is spit out to be used in the WHIP formula.
Not much needs to be said here, as I project innings pitched and K/9, and then voila, a strikeouts projection pops out.
I hate wins as a fantasy category and every time I threaten to switch it the following year, but of course that never happens. My wins projection is based on the same Pythagorean formula used to estimate team wins and losses. I first look at the latest projected standings at various sites that also project runs scored by each team and average them together. This number serves as the projected run support for the pitcher. The pitcher’s ERA is then multiplied by 1.1 to estimate total runs allowed, not just earned runs. Last, a constant is added to the formula based on the average number of innings pitched it takes to record a decision (a win or a loss). This constant is a little higher for relievers.
All these numbers are then thrown into this modified Pythagorean formula which spits out my win projection. This is fantastic for me, because trying to project wins manually is not an enjoyable task.
This is the most manual category, as unfortunately I have yet to find a formula that I was confident enough in to use instead. Projecting saves is basically a crap shoot, so I look at the numbers everyone else typically will. First off, I completely disregard the whole “he can’t close” BS. Ryan Madson had that tag and he was excellent in the role given a real opportunity. Middle relievers can’t save games, only blow saves, and many times when these pitchers get labeled as not being up to the task of closing games, the sample size is ridiculously small. So I will generally look at my projected ERA and then how safe the pitcher’s job is. This is the most art (as opposed to science) required in doing my pitcher projections, which is why I typically like to draft the more unproven closers with strong skills on the cheap.
Print This Post