Defending the Stolen Base Pod Projections

Last week and earlier this week, I highlighted a group of hitters who my Pod Projections projected stolen base upside and stolen base downside compared to Steamer projections. Until I performed the comparison, I had no idea I was so bearish on steals, relative to both Steamer and to 2017.

In my downside post, I compared PA/SB marks. However, those numbers were not weighted by plate appearances, so we’re going to try this again (I had included Steamer in the downside post, but since their forecast average was close to the 2017 actual mark, I kept this comparison to just 2017 to simplify things). There are 325 hitters in my projection set who recorded at least 10 plate appearances last year.

PA/SB Comparison
2018 Pod Projected PA/SB 2017 PA/SB
78.2 69.9

That appears to be a relatively large decline in stolen bases. But is it really? A 600 plate appearance season would yield 7.7 steals using my forecast, and 8.6 steals using 2017. That’s about a 10% decline. I guess that’s rather significant. And a significant difference caught the eye of frequent commenter TheTinDoor, who immediately challenged me:

Thanks. I get that you’re not projecting league-wide stats, you’re projecting individuals, so ultimately it may not matter. But being that far off 2017 would scare me if I was in the projections business…

Challenge accepted. Let’s dive in to determine if I’m off my rocker or there’s something to this expected decline in steals.

Like I did in identifying the upside and downside guys, I divided the Pod Projected PA/SB by the 2017 actual mark. The lower the percentage, the more bullish I am, and the higher percentage, the larger the decline I’m projecting.

Let’s compare some metrics between the Bearish and Bullish group. I have excluded any player that either stole 0 bases in 2017 or I’m projecting them for 0 steals. I also only included hitters who stole at least eight bases in 2017 as they weigh more heavily in the group average than players who stole just a base or two.

That left with me just 16 hitters with a 2018 Pod Projected PA/SB % of 2017 PA/SB below 100%. So my two groups are composed of 16. You might cry small sample size, but it was clearly big enough to perfectly tell the story.

PA/SB Comparison
Group Age 2018 Pod Projected PA/SB 2017 PA/SB 2018 Pod Projected OBP 2017 OBP 2018 Pod Projected ISO
Bullish 25.6 32.0 35.2 0.326 0.327 0.149
Bearish 28.1 63.9 37.9 0.336 0.351 0.170

These averages are almost too good to be true. Let’s start on the left. Speed is a skill of the young. You know this. I know this. Everyone knows this. Players lose speed as they age. This is a fact. I had totally forgotten that at the end of 2013, I asked Jeff Zimmerman to create a stolen base aging curve for me. He did just that. Rather than demand you click the link to find the curve, I have reproduced it:

Stolen Base Aging Curve

Would you look at that — stolen bases remain relatively stable through about age 27, then they decline, and keep declining. My Bullish group is still safe, averaging just 25.6 years on Earth. On the other hand, the Bearish group is a whopping 2.5 years older, averaging 28.1 years. That’s precisely the age in which stolen base totals begin their rapid descent. Of course, it’s even worse for the big steals guys who swipe at least 20 bags.

Moving to the right is now a comparison of PA/SB for each group, first my projection, and then 2017 actual. You can see that I’m far more bearish on the Bearish group than I am bullish on the Bullish group.

Next up is OBP, a huge driver of stolen base attempts, for obvious reasons. You’ll notice that I’m projecting the Bullish group for a virtually identical OBP as in 2017. For the Bearish group, I’m forecasting a 15 point decline in OBP. That drop in OBP clearly provides part of the explanation for the drop in PA/SB versus 2017. You can argue that my projections are wrong/bad for the Bearish group, of course, but that’s a topic for a different day.

Last, a comparison of projected ISO (isolated slugging) ends the table. In my long-time projecting experience, I have learned that power guys who steal bases often decide that attempting a steal is no longer worth the risk. A lot of Billy Hamilton’s value comes from his speed and stolen base prowess. He couldn’t afford to stop displaying his best skill or dramatically reduce it.

Power hitters are different. Their offensive contributions are driven by their propensity for extra-base hits, namely home runs. Is it really worth the injury risk to also steal 20 bases? No, probably not. You can see the huge variance in these types of players using two cherry-picked examples of two of baseball’s best power bats, Manny Machado and Bryce Harper. In recent seasons, each of them suddenly stole 20+ bases, but then followed that up with 0 and just 4 steals, respectively, the next year. So I will always regress a power hitter’s PA/SB more heavily than a guy who relies on his speed to provide value.





Mike Podhorzer is the 2015 Fantasy Sports Writers Association Baseball Writer of the Year. He produces player projections using his own forecasting system and is the author of the eBook Projecting X 2.0: How to Forecast Baseball Player Performance, which teaches you how to project players yourself. His projections helped him win the inaugural 2013 Tout Wars mixed draft league. Follow Mike on Twitter @MikePodhorzer and contact him via email.

10 Comments
Oldest
Newest Most Voted
Inline Feedbacks
View all comments
Alan
6 years ago

Good article, and I would respectfully disagree with the comment to the previous article suggesting there is a natural league-wide value that projections must relate to. If you were projecting wins or innings pitched, there would be, but the game could exist with zero stolen bases or with a steal attempt every inning, and sadly we are trending towards the former. Steals are strategic choices, just like sacrifice bunts, and there is no reason for them to “regress to a mean” like a HR/FB rate or similar, other than there may be modest “natural variation” in mix of game situations, opponent skill at countering steals, or a player’s OBP that could make a small difference in season-long steal totals.

TheTinDoormember
6 years ago
Reply to  Alan

I don’t have access to the specific Pod Projections player pool (specific 325/342 players), so here are the league-wide numbers PA/SB% numbers for the last 5 years:

2013: 68.6 PA/SB
2014: 66.5
2015: 73.3
2016: 72.8
2017: 73.3

Clear movement away from SBs between 2014 & 2015 season; no real change the last 3 years. These numbers ARE closer to Pods than the specific player pool (69.6 in 2017). You’re right of course, there’s no reason that these NEED to align the way Wins do.

My broad point: I think it’s reasonable to expect league-wide SB trends to be similar to the last 3 years. Pod’s arguments are perfectly reasonable for any individual player who’s aging. I wouldn’t argue with any specific player’s projection.

Taken as a whole, though, projecting the player pool to drop by 12% (Pod’s projected numbers) hasn’t really been explained. Yes, as players age, they’ll steal less. As they hit for more power, they’ll steal less. But this was true in 2015, and 2016, and 2017, and yet league-wide rates haven’t changed. Some players are aging down the curve, but there must be another offsetting group of players who are stealing more (or are coming into the league).

Mike, do you believe league-wide trends will change dramatically this year? You haven’t really been arguing that point (your posts are more narrowly focused on individual players). It’s also possible that league-wide rates stay steady while a sub-group declines, but the size of your cohort would lend me to expect roughly similar…

TheTinDoormember
6 years ago
Reply to  Mike Podhorzer

Well to be fair, the league numbers actually help your argument since the league PA/SB (73.3) was closer to your projections (77.8) than the pool (69.6) in 2017. Your non-projected players must have stolen at a considerably slower pace in 2017 to bring that average up nearly 4 PA/SB?

It would certainly be interesting if your projections got “individuals right more often”, but on the misses, were severely skewed to under-projecting SB. A few big outliers?

This may come across as pedantic, as I don’t have a better way to do it and I’m hung up on the 10-12% overall decline as opposed to individual players. Maybe put a reminder to look at 2018 results in October?

Mattabattacolamember
6 years ago
Reply to  TheTinDoor

I think it just speaks to the overall unpredictability of individual stolen base totals. Mike identifies this and stays conservative. If you build a projection system that has the goal of valuing players for fantasy baseball, you would not want to rely heavily on an unstable part of many players games.

He also brings up a good point that there is not necessarily a way to regress to the mean. Because of this, being conservative on stolen bases means you universally lower everybody’s totals. Jarrod Dyson from 30 last year to projection maybe 25 this year. That is not offset by saying Jose Abreu will trend towards the mean and steal 5 instead of 1, rather you won’t count on him stealing at all. If you imagine this throughout each individuals projection, suddenly you see how a large expected drop is possible without something being fundamentally wrong with the overall system.

The outliers who outperform the projections will happen, but there simply is no reliable way to predict those breakouts.

The Strangermember
6 years ago
Reply to  Alan

This is true, but in general you probably wouldn’t want to forecast a change in the league-wide environment without a good reason. However, the comparison also only looks at the players who had at least 10 PA last year. I wonder how many SB could reasonably be expected from players who didn’t have 10 PA last year, and whether that might move the projections more in line with last year’s numbers.

I also suspect there’s an effect where some number of utility infielders and 4th/5th outfielders end up playing regularly because of injuries or vastly outperforming their projections and get the chance to rack up a bunch of steals. But that’s not something you can easily predict at the start of the season – the sensible projection for any one of those guys individually is that they might steal bases at a decent rate but they won’t play enough to put up big totals.