Is Fangraphs a reliable website? Project 2019 Brewers to finish last?

[JackNicholson1974]

I am wondering if the website Fangraphs.com is a reliable website for their advanced data projections?

 

I ask because I stumbled upon their 2019 projections and they project the Brewers to finish last in the NL Central in 2019 with a record of 77-85.

 

https://www.fangraphs.com/depthcharts.aspx?position=Standings

 

Anyone have any thoughts on that? Should we expect that much regression?

[Ron Robinsons Beard]

No ... they are not, in my opinion.

 

I mean, of course we are going to see some regression. But we also saw several players regress last year that should hopefully rebound. But I would peg that record projection at about a 0.5% chance of happening. The Brewers will/should be significantly better than that.

[homer]

They have a projection system that even the writers know comes out with some odd looking results from time to time. Here is from the season review article that came out today:

 

The larger point is that the Brewers don’t have much in the way of dead money and are awash in flexibility. I think they start off the season as underdogs to win the division, but they’re a team with realistic hopes to once again topple their wealthier rivals.

 

https://www.fangraphs.com/blogs/elegy-for-18-milwaukee-brewers/

[KeithStone53151]

Fangraphs projections are decent for guys that are under about 32 and have played at the mlb level for at least 2 years. Obviously any guy can have a bad year or career year. They seriously struggle to predict how very young players will perform at the mlb level and when players will decline. Looking at fangraphs projected standings is a complete waste of time. If you're curious how Yelich or Shaw or Perez might do this year, fangraphs projections are probably pretty reasonable. Then again, you could just as easily average their last 3 years and get roughly fangraphs projections. Last 3 years for Yelich average 302/382/507 vs fangraphs 296/381/514...close enough.

 

When I look at fangraphs, I consider it a starting point and add in my own assumptions. Arcia for instance showed enough promise at the end of 2018 and in the playoffs that I expect him to exceed his projections. Kratz on the other hand, they were probably a bit too nice to and he'll probably be worse than his projections. I'm frankly a bit blown away by how bad they anticipate Aguilar being. I expect him to regress as well but not to be a below average hitter at 1b. Something in the 800-825 OPS range for Aguilar would be a good projection, fangraphs dropped him all the way to 770 from his 890 in 2018.

[sveumrules]

FanGraphs does not create projections, they only host them.

 

Nineteen of their forty writers picked us to make the playoffs last year, with seven picking us to win the division.

 

https://www.fangraphs.com/blogs/fangraphs-2018-staff-predictions/

 

ETA: Former FG writer August Fagerstrom was hired away from the site by the Brewers & is currently listed as "Analyst - Baseball Operations" for the team, so clearly they think the information provided is of some value. I saw him throw out the first pitch at a game I was at (pretty sure they said he won the chance to do so at an employee party or something) & he did alright...for a NERD.

[homer]

FanGraphs does not create projections, they only host them.

 

Nineteen of their forty writers picked us to make the playoffs last year, with seven picking us to win the division.

 

.

 

yeah that's actually a better point. I believe their team projections are just totals rolled up from the player projections. Could be wrong though.

 

FanGraphs Depth Chart projections are a combination of ZiPS and Steamer projections with playing time allocated by our staff.

[sveumrules]

This article by Jeff Sullivan (published February of 2018) takes a look at historical projections from 2005-17 & does a really good job of explaining the limitations & most reasonable applications thereof...

 

https://www.fangraphs.com/blogs/lets-make-sure-were-honest-about-projections/

[Lathund]

Worth pointing out that "Fangraphs" aren't projecting anything. They host two different projection systems on their site; Steamer (Which comes out very early in the season) and ZiPS (Which is in the process of being rolled out). Close to the start of the season they will combine the two, and adjust depth charts, playing time and such. Only then will things like likely platoons, unconventional bullpen usage etc. be factored in. For a team that platoons very little and that has 5 traditional starters who all pitch 180+ innings the projections will be fairly reasonable even now. But for teams like the Brewers or Rays you'll have to wait a bit to get accurate projections.

 

Also projections are what they are; Up until something like 2/3 into a season they're still more predictive of rest of season performance than season to date stats are. We as fans will naturally tend to focus on the projections that don't agree with our views, usually the ones that underrate our own players. What a projection system can't do however is to take any kind of scouting data, or anything outside the game, into account. No projection system could say that since Wade Miley completely changed his pitch repertoire around midway through the 2017 season, he would suddenly have a breakout that went completely against his career numbers. Or that Tommy Pham got contact lenses. Or that someone revamped their swing. They also can't fully project breakouts, other than I guess with players who had a breakout year in terms of performance but not in results. And since using multiple seasons of data is usually more accurate, the projections for breakouts will lag behind some.

 

Projections will also tend to cluster teams more closely together than how things end up. Which is fairly obvious when you think about it; a projection that results in a statline of .250/.350/.450 is some sort of mean outcome, or most likely outcome, from 1000s of simulations. DRC+ shows the standard deviation next to it's values, which is something that should be done more often. Anyway, because it presents like that, naturally the extreme outcomes, whether it be 2018 Christian Yelich or 2018 Chris Davis will be way off any projection.

 

TL:DR: Projections are a lot better than most fans give them credit for. They also have some weak points, especially this early. ZiPS (Gradually being rolled out at FG) and PECOTA (At Baseball Prospectus) will be available soon and are usually more accurate. Some of the Brewers projection seems a bit off, but not egregiously so. A lot comes down to how playing time is distributed in the projections for instance, particularly with pitching. Another part of it is the defensive metrics; DRS, FRAA and OAA liked Brewers defense better than UZR did last year, and so that'll play into projections based off UZR too. It also seems to pretty much straight up translate FIP to ERA. Which is reasonable for players with little or no track record, but less so for those who have consistently outperformed, or those who play in front of a good defense. Aguilar is mainly down due to projecting a lower BABIP, which isn't unreasonable to do even if it may not end up that way. Some of the pitchers are a bit of a head-scratcher though I must admit.

 

Also consider that the Reds have made several additions, so have the Cardinals. Brewers are likely to be more active going forward than they are. That'll play into it, and the other (IMO better) projections will also be a bit higher on the Brewers. Who will still be somewhat underrated by the projections; which I remember the authors being very well aware of last offseason, but which is hard to do anything about; start meddling with opinion in these projection systems, and they lose their value.

[long ball]

I think Fangraphs is a good resource. I generally prefer Baseball Prospectus now that they've ditched TAv as their primary offensive stat in favor of newly developed DRC+, but I like Fangraphs because I think the site is more navigable.

 

As far as projections, they all have limitations. As others have mentioned, Steamer has some limitations especially this early in the off season. I was actually joking around with some of Brewers twitter a couple days ago about steamer projections for 2019 currently having the Brewers as a last place team.

 

Once zips and pecota roll out and rosters fill out we will have a better idea of what the projections are. Anecdotally, I think some of the strengths the Brewers have don't get measured effectively by projections. I don't think depth is accurately quantified by just innings pitched, plate appearances, and defensive innings. Example: the Brewers played their best baseball in September 2018 and in hindsight it was very predictable. Counsell was able to deploy his 35ish players in the best possible position to succeed and eke out extra situational value. Now the Brewers roster isn't as deep right now as it was in September 2018, but by continuing to improve at the margins (ie Gamel, Spangenberg, Claudio as complements to the returning roster and with minor league options to help over the course of April-Aug) the Brewers could reel off a similarly impressive September in 2019. I also don't think it accurately accounts for an elite (or awful) bullpen. The replacement level used for relievers is different than that used for starters which makes sense in a traditional model, but as the Brewers continue to "blur the lines" between relievers and starters that replacement level becomes less applicable.

 

In summary, I wouldn't get too worked up about what steamer shows as the standings right now and even if zips and Pecota peg the Brewers finishing somewhere other than 1st or 2nd in the NLC.

[CrewFanBrewMan69]

I think it definitely has a blind spot for us because how we are built and how we win is a little different. Last year required a crazy bullpen year and defense, those are volatile and hard to predict year to year and it’s probably hard to sustain the bullpen success we’ve had (in the projections eyes). Our rotation is also weak in their eyes, partly because the guys who have major league innings are league average or so and/or guys with only minor league track records. Also our defense/shifting was so good that it probably won’t project us to be as good there again because either it’s not likely to replicate or has not adapted correctly to account for the shifting. We also had like 14 wins out of Yelich and Cain, I don’t think it’s crazy that the projections don’t see that happening again and the rest of the offense is guys with not huge samples or impressive ones.

[Fear The Chorizo]

Fangraphs is pretty good at evaluating, comparing, and projecting individual players.

 

IMO, fangraphs is among the last places I look for team record projections, because their methodology has proven to be about as accurate as throwing darts blindfolded for team projections.

[Thurston Fluff]

When a team uses new, innovative approaches it is hard to get accurate projections from analytic based evaluations. Analytics takes time to compile stats to evaluate. The very definition of new and innovative means that isn't available. Using traditional metrics for a new approach fits like a square peg in a round hole.

FWIW I wouldn't be surprised if Aguilar does regress considerably but I also don't think this team's success last year was due to a bunch of guys having a career year. For every Aguilar there was a Santana.

[NYChez]

Vegas has the Brewers in the top third of MLB for 2019 World Series odds. That translates to around 88 wins. Note that the Brewers were 6th favorite before other teams made significant additions.

[CrewFanBrewMan69]

Vegas has the Brewers in the top third of MLB for 2019 World Series odds. That translates to around 88 wins. Note that the Brewers were 6th favorite before other teams made significant additions.

Vegas odds are less reliable/accurate than projection systems from Fangraphs, 538, BR, etc. Vegas prices things to get equal action on both sides and adjusts odds to limit exposure, not necessarily project what is most likely to happen.

[Ron Robinsons Beard]

Vegas has the Brewers in the top third of MLB for 2019 World Series odds. That translates to around 88 wins. Note that the Brewers were 6th favorite before other teams made significant additions.

Vegas odds are less reliable/accurate than projection systems from Fangraphs, 538, BR, etc. Vegas prices things to get equal action on both sides and adjusts odds to limit exposure, not necessarily project what is most likely to happen.

 

In this case, Vegas is WAY more right than Fangraphs. That Fangraphs projection is crap. I don't care how they justify it. It's still crap.

[NYChez]

Vegas has the Brewers in the top third of MLB for 2019 World Series odds. That translates to around 88 wins. Note that the Brewers were 6th favorite before other teams made significant additions.

Vegas odds are less reliable/accurate than projection systems from Fangraphs, 538, BR, etc. Vegas prices things to get equal action on both sides and adjusts odds to limit exposure, not necessarily project what is most likely to happen.

 

Looks like you have an easy path to becoming rich!

[homer]

Vegas has the Brewers in the top third of MLB for 2019 World Series odds. That translates to around 88 wins. Note that the Brewers were 6th favorite before other teams made significant additions.

Vegas odds are less reliable/accurate than projection systems from Fangraphs, 538, BR, etc. Vegas prices things to get equal action on both sides and adjusts odds to limit exposure, not necessarily project what is most likely to happen.

 

I could go into the predictive nature of futures markets but I will not.

[jerichoholicninja]

At this point, Ben Gamel and Claudio are the only additions the team has made that will likely impact the major league roster. We've lost our best SP and a starting infielder and replaced neither, and last year relied on Yelich and a bullpen full of guys playing out of their minds. If the season started today I could see this being a .500 team.

 

I'm not trying to be a Debbie Downer, I'm just pointing out Stearns hasn't gotten to work yet, so I can't blame anyone for looking at the current roster and not seeing many wins there.

[Fear The Chorizo]

At this point, Ben Gamel and Claudio are the only additions the team has made that will likely impact the major league roster. We've lost our best SP and a starting infielder and replaced neither, and last year relied on Yelich and a bullpen full of guys playing out of their minds. If the season started today I could see this being a .500 team.

 

I'm not trying to be a Debbie Downer, I'm just pointing out Stearns hasn't gotten to work yet, so I can't blame anyone for looking at the current roster and not seeing many wins there.

 

Fangraphs' 2018 brewers record projection, after cain and yelich were added - 79 wins. Actual result - 95 wins in 162 games

 

Fangraphs' 2017 brewers record projection - 73 wins. Actual result - 86 wins.

 

Based on their recent track record, a 77 win projection means the crew may threaten 100 w's without adding anyone else

[Eye Black]

The Fangraphs website is a great resource. The record projections seem inconsequential when compared to the overall content hosted by Fangraphs. Painting the entire Fangraphs platform as unreliable based on some team records spit out from the Steamer projection system input seems silly.

 

If you look a little closer the low projection is almost entirely due to the lack of certainty surrounding the Brewers pitching staff. The Steamer system gives the Brewers starting pitchers very little credit. Steamer projects all five Cubs starters to be above 2 WAR, with four of them being around 2.5 WAR each. The same system doesn’t project the Brewers to have a single pitcher near 2 WAR, with the top projection is for Jimmy Nelson at 1.6 WAR (next highest on the entire pitching staff is projected at 1.2 WAR). It also is projecting Zach Davies for 161 innings pitched next season, which isn’t something I would bet on either.

 

In my opinion it isn’t worth getting too worked over these projections, the Brewers are well equipped to outperform them by excelling in areas where the projection system has limitations.

[Ron Robinsons Beard]

At this point, Ben Gamel and Claudio are the only additions the team has made that will likely impact the major league roster. We've lost our best SP and a starting infielder and replaced neither, and last year relied on Yelich and a bullpen full of guys playing out of their minds. If the season started today I could see this being a .500 team.

 

I'm not trying to be a Debbie Downer, I'm just pointing out Stearns hasn't gotten to work yet, so I can't blame anyone for looking at the current roster and not seeing many wins there.

 

Fangraphs' 2018 brewers record projection, after cain and yelich were added - 79 wins. Actual result - 95 wins in 162 games

 

Fangraphs' 2017 brewers record projection - 73 wins. Actual result - 86 wins.

 

Based on their recent track record, a 77 win projection means the crew may threaten 100 w's without adding anyone else

 

Fangraphs may have its value, but this proves that it's record prediction algorythm is broken.

[3and2Fastball]

Fairly easy to project that Christian Yelich will not hit like steroid era Bonds in September of 2019. Difficult to project whether other players will pick up the slack, whether Jimmy Nelson will be productive, how valuable Burnes & Hiura & Dubon will be, and whether our relief aces will regress or not.

[Eye Black]

Good point, and if production wasn’t already tough enough to predict, correctly projecting player usage is a nearly impossible task. Right now these projections have Hernan Perez with 513 plate appearances, Cory Spangenberg with 302 PA, Eric Thames with 211 PA, and Tyler Saladino with 191 PA. I will take the under on those four hitters combining for 1,200 plate appearances.

[Boomer5]

Fangraphs takes into account player values but they seem to struggle with the massive impact of Counsel in the totality of our performance.

 

I’ve never seen a baseball manager get more out of a roster than he does.

[coolhandluke121]

I think it definitely has a blind spot for us because how we are built and how we win is a little different. Last year required a crazy bullpen year and defense, those are volatile and hard to predict year to year and it’s probably hard to sustain the bullpen success we’ve had (in the projections eyes). Our rotation is also weak in their eyes, partly because the guys who have major league innings are league average or so and/or guys with only minor league track records. Also our defense/shifting was so good that it probably won’t project us to be as good there again because either it’s not likely to replicate or has not adapted correctly to account for the shifting. We also had like 14 wins out of Yelich and Cain, I don’t think it’s crazy that the projections don’t see that happening again and the rest of the offense is guys with not huge samples or impressive ones.

 

I agree. I mean, what's the FG projection for manager/gm/analytics WAR? They're milking a lot of extra wins out of these guys. Not a lot of guys who contributed the last 2 years are coveted individual talents if you think about it, so I don't think FG's pessimism about the sum of individual talent on the team is actually that far off. It just misses the part of their value that goes with being able to fit into the larger strategy they're employing. That includes having options, being a good fit in a platoon, being able to handle long relief, and being a versatile defender.

 

As for next year, I think it's a combination of uncertainty about the young pitchers and some possible regression from guys like Aguilar, Yelich, Chacin, Cain, and Jeffress. I personally think Burnes, Peralta, and Woodruff will help them have a pretty strong staff, but I seriously doubt they'll have the most all-stars in the NL once again next year. I'm kind of bracing myself for a temporary step backwards, but I support them wanting to keep Hiura and their young pitchers instead of trying to keep pace with some of the other moves teams are making for the short term.