clock menu more-arrow no yes

Filed under:

Hok Talk: How accurate were the 2018 ZiPS?

New, 17 comments

Many of you hate projection systems. Are you justified in this case?

Whit Merrifield prepares for an at-bat.
You probably guessed that Whit Merrifield outperformed his projections by a fair bit.
Photo by Kiyoshi Ota/Getty Images

Royals fans have a love-hate relationship with projection systems. They love to hate them. This is particularly true of the PECOTA projections which are, funnily enough, named after - but in no other way related to - former Royals infielder Bill Pecota. In 2014 and 2015 PECOTA didn’t particularly like the Royals and the Royals completely blew the projections out of the water both years.

For the record, PECOTA was quite generous to them last year with a 66 win projection. The Royals finished the year with a cool 58 Ws.

But there are a lot of other projection systems out there. Perhaps one of the most famous (or infamous) is Dan Szymborksi’s ZiPS projections. They attempt to estimate a player’s stats and fWAR rather than team record but you can project out a record using them if you like. But even leaving record calculations out of it we’ve had some people ask how the projections compared to the final result and it sounded kind of fun so let’s check this out.

If you add up all the zWAR of the projected position players, ignoring playing time, that played for the Royals last year you get a total of 12.3 zWAR. The final fWAR total of all Royals position players was 10.7 fWAR. That’s pretty close, all things considered. There were some pretty drastic differences, of course. Whit Merrifield, for example, was projected for 2.5 zWAR before collection 5.2 fWAR. Adalberto Mondesi was projected for 0.9 zWAR and collected 2.8 fWAR.

There were some over-estimations, too. Salvador Perez was projected for 2.6 zWAR but injury and ineffectiveness reduced him to 1.7 fWAR in reality. Paulo Orlando was projected for 0.1 zWAR in 364 plate appearances but collected an amazing -1.0 fWAR in only 93. ZiPS generously suggest Alcides Escobar was worth 0.3 zWAR when in reality he was worth -0.3 fWAR.

Others were pretty accurate. Alex Gordon was pegged for 1.4 zWAR and came up with 1.6 fWAR. Mike Moustakas was pegged for 2.5 zWAR in a full season and gave the Royals 1.7 fWAR in approximately two-thirds of a season. Drew Butera pulled in -0.1 zWAR and a very similar -0.3 fWAR.

What about the pitching side, though? The projection systems failed us a bit on that side. The Royals projected out for 14.0 zWAR but attained a mere 5.1 fWAR. So what the heck happened? If you guessed that it starts with Blaine Boyer, Justin Grimm, and Brandon Maurer you guessed well.

I couldn’t find projections for Boyer but Grimm nabbed a 0.4 zWAR while Maurer was thought to be able to achieve 0.6. Instead, those three went -0.9, -0.5, and -0.8 fWAR respectively. Enny Romero was another guy I could find no projections for and despite only throwing four innings for the Royals he accounted for another -0.4 fWAR.

It wasn’t all bad news for the Royals pitching staff, of course. Brad Keller pulled off a 2.5 fWAR campaign despite being projected for only 0.5 zWAR. He stands as the highest over-achiever. Jakob Junis came away with 1.3 fWAR compared to 0.8 fWAR. And Kelvin Herrera managed to hit his projected zWAR at 0.7 fWAR even before he was traded.

The most accurate projections were probably Kevin McCarthy (0 zWAR, 0.2 fWAR) and Glenn Sparkman (0.2 zWAR, 0.1 fWAR in a little more than half of the projected innings.) Ian Kennedy was projected for 1.0 zWAR and put together a 0.8 fWAR campaign. Which just goes to show that the more major league experience these players have the more likely the projections are to be accurate. Which is exactly what you should expect from them. More data should always mean better accuracy in projections.

So if you take the projected zWAR of this team and add it to the 0 WAR base (about 48 wins) you end up with ZiPS projecting the Royals would win approximately 74 games. Obviously, that’s even more inaccurate than PECOTA was this season so we should not use the system that way without putting in a bit more effort. I made no effort to re-calculate the numbers based on playing time or to fill in the gaps of the zWAR for players for whom I could not find projections.

If you’re wondering how accurate that 48-win number is for a 0 WAR team the Royals should have won 64 games based on their team fWAR. However, Bill James invented a concept known as the Pythagorean W-L record which attempts to normalize out some of the luck inherent in a team’s season by calculating how many wins they should have had based on runs scored and allowed. The Royals’ Pythagorean record was 62-100 which is pretty dang close the number of wins you would have expected the Royals to have with their team fWAR.