28 October 2015

Looking At The Orioles 2016 Steamer Predictions

Click here for Matt Perez's archive.
Fangraphs has recently uploaded the first batch of projections via the Steamer projection system for the 2016 season. This is one of the first projection system to release a batch of projections and therefore can be used to get a basic understanding of how the 2016 Orioles look before the start of free agency.

Before starting on this task, it’s worth noting that Steamer has already stated on twitter that their playing time projections aren’t particularly sensible at the moment. Indeed, this system is even giving guys like Travis Snider, Delmon Young and David Lough playing time for the 2016 Orioles’. I believe that they have recently removed playing time from their projections.

After removing soon-to-be free agents, Steamer projects the 2016 Orioles to earn 10.8 position player and 9.5 pitching fWAR, implying that the Orioles should be expected to win 68 games. The Orioles’ free agents are worth another 10.6 fWAR total, and thus the Orioles would be a 79 win team if they didn’t lose a single free agent. All in all, this is a glum projection for the Orioles’ and sits in stark contrast to Jon’s prediction made earlier in the month. So, it’s interesting that a closer look suggests that these WAR totals don’t necessarily predict the same thing that other Steamer statistics show.

On the offensive side, there are relatively few surprises. Steamer projects Schoop will regress and end up with a .249/.288/.422 line while projecting that JJ Hardy will have a bounce back year with a .239/.280/.363 line. It isn't clear whether Steamer takes into account his labrum injury and is a potential shortcoming of the model. 

The first step to projecting offensive production is to take the plate appearances projected for each player (no longer available) and the players positions. The remaining plate appearances at each position would be taken by replacement level players who performs 10% worse than the average player at his position. For the sake of simplicity, I will presume that each position (including pitcher) has the same of plate appearances in 2016 as they did in 2015. Once that’s complete, the depth chart looks like this and the results per position are shown below.

The results show how the Orioles’ lineup is weak. Steamer projects the Orioles to be top-ten at only third base (Manny Machado) and center field (Adam Jones). They expect the Orioles to be slightly below average at catcher and second base while receiving terrible offensive performances at first, shortstop, left field, right field and DH which come as a surprise to no one. When inputting these results into the lineup simulator used by Jon in his post, the Orioles should be expected to score 661 runs. In 2015, this would have been the 20th most runs scored in MLB and would tie for 12th highest in the AL with the Angels. All in all, it seems plausible that Machado and Jones wouldn’t be enough to save the offense but would be enough to keep it from being the worst in the AL.

This is also 14 fewer runs than Jon’s method projected for the Orioles. Given that Steamer thinks that players like Reimold and Urrutia are better offensively than a replacement level corner outfielder, Walker is better than a replacement level first baseman while Joseph and Clevenger are better than replacement level catchers, it is likely that a depth chart projecting those players to play significant innings would further narrow the distance between the two projections.

The next task is to do something similar for the pitching side. Steamer released projected innings pitched and earned runs allowed by each pitcher. I presume that each pitcher pitches the amount of innings that Steamer predicts and that replacement pitchers throw the remaining innings. A replacement level starter should have an ERA of 5.76 while a replacement level reliever should have an ERA of 4.63. On average, the Orioles pitching staff pitches roughly 1450 innings per year, of which the bullpen throws 540 and the rotation throws 910 while the Orioles’ fielding allows 35 unearned runs. The chart shows the results.

The Orioles may lose O’Day and Chen but will still have a strong bullpen and good fielding. The Steamer projections suggest that the bullpen will be deep and therefore be able to throw quality innings. Even without O’Day, and with Steamer projecting significant regression, the Orioles’ bullpen is still tied for #17th in the majors. All ten of the relievers on the Orioles' depth chart are better than replacement level and therefore the Orioles' bullpen should improve as the depth chart is updated.

The rotation is still projected to be poor with the loss of Chen and ranked only 25th in the majors as they were in 2015. Steamer believes that Orioles’ starting pitching will regress to the mean and therefore make up for the loss of Chen. Overall, this method projects that the Orioles will allow 721 runs. This is good enough to for 19th in MLB and 10th in the AL.

If the Orioles were to allow 721 runs and score 661 runs then they should win 75 games. This is three fewer wins than the 78 that Jon projected but is considerably closer to his projection than the 68 wins that Steamer projects via WAR. Furthermore, improvements in the depth charts will help the bridge the gap.

Steamers’ WAR projections suggest that the 2016 Orioles should consider rebuilding while their other statistics suggest that a few more pieces can turn the tide. It is interesting that they each tell different stories and probably means that the Orioles are in a better situation than a first look at Steamer might indicate.


Brooklyn Guy said...

Can we get some perspective on these predictions? What did the predictions show vs. actual performance over the past few years?

Jon Shepherd said...

One standard deviation is about 8 games, so, yes, they mean something, but you are looking at a broad bell curve. For the most part, projections for last year ranged from around 78 to 83 last year. I had them at 85 with my system.

Matt Perez said...


That link argues that Steamer is one of the more accurate projection systems out there. Then again, this is just a first draft and probably won't be as polished as the final product.

My understanding is, and could be wrong, is that studies have shown that these predictions have some validity. I'd consider them unbiased opinions of each clubs talent. This is important because fans of a club can have biases that make them overestimate the talent on their club. The fact that this judges clubs impartially and with some accuracy means it deserves a seat at the table.

I also wouldn't argue that these should be taken as gospel. Projections aren't going to be 100% accurate.

Jon Shepherd said...


We have done some others.

Chicago Curmudgeon said...

The real answer comes if you look another year out. The fact is that help is not coming from the minors, that the only players on the roster that rate to get significantly better are Machado and Gausman, and Jones is already on a decline. The Orioles have taken their shot at a title before Machado turns into a free agent unless there are radical changes in team economics.

If the team wants to take its shot while it has Jones and Machado, it probably has to spend $75MM more a year. Maybe $100MM more in its most expensive phase, then collapse back to a rebuild having cemented Machado to a long-term deal in his first run at free agency.

I hope the need for the spending spree would convince ownership that it has to make larger, smarter steadier investments in minor league and international talent acquisition to avoid the situation it is in now.