An Early Look at the Projected Standings
We’ve had the 2016 Steamer Projections up on the site for a while now, but until this morning, the only way to look at the aggregate team projections was to look at a team’s total projected WAR and eyeball how that might translate to wins and losses. WAR is a good enough proxy to get you in the right neighborhood, but because of differences between the leagues and the fact that wins aren’t perfectly linear, ideally, you want to run the raw numbers through a run estimator and then use BaseRuns to convert those runs scored and allowed numbers into an expected win total. Well, as of today, we’ve updated our Projected Standings page to do exactly that, taking the individual Steamer projections and the playing time projections from our depth charts to produce estimated win-loss records for every team in baseball.
Probably to no one’s surprise, the Cubs currently stand atop the projections with a 95-67 forecasted record. The Cubs were excellent a year ago, built around a core of exciting young players, and then added Jason Heyward, Ben Zobrist, and John Lackey in free agency. Yes, they’re projected to win two fewer games than a year ago, but that’s simply a function of the fact that projections are attempting to project context-neutral performance, not accounting for wins that can be added (or lost) due to clutch performance; taking away the effects of sequencing naturally results in a smaller spread from top to bottom.
So, instead of looking at the projections relative to a team’s 2015 win-loss record, here are the current Steamer projections — these can and will change as more free agents sign, trades are made, and the depth charts become more clear as we get closer to the season — compared to each team’s own BaseRuns expected record from a year ago.
The difference that immediately stands out the most is the Red Sox, who played like a mediocre team last year but are again projected as the AL’s best team on paper, according to these forecasts. Of course, our projections liked the Red Sox a lot last year, and that didn’t work out so well, so I’d imagine there will be a decent amount of skepticism related to that forecast, given how the team has actually performed the last two years. Adding David Price and Craig Kimbrel definitely helps, and the system expects Hanley Ramirez and Pablo Sandoval to go from a combined -4 WAR to to +4 WAR — an upgrade roughly equivalent to replacing a AAA scrub with Mike Trout at one position — so it’s hard not to see this team as likely to be a bit better, though I’d take the under on the idea that they’re the second best team in baseball at the moment. It’s clear, though, that once again, Steamer really likes the Red Sox roster.
The other big jump also belongs to the team named after footwear, but this one is probably a little more likely to be accepted; the White Sox make a jump into the middle-tier of teams thanks to replacing scrubs with Todd Frazier and Brett Lawrie, plus adding a competent catcher in Alex Avila, along with adding in some positive regression from Melky Cabrera. The team is still lacking a big league shortstop and a couple of back-end starting pitchers, but the Frazier move in particular was a huge improvement, and the White Sox hopes of contending last year might come to fruition this year if they catch some breaks.
On the other end of the scale, the Blue Jays and Astros are expected to take the biggest steps backward, though in both cases it’s more about inevitable regression to the mean than anything about the team’s off-season decisions. The Blue Jays just got so many great performances from guys who can’t be reasonably expected to have the same season again — Kevin Pillar and his +4.3 WAR, for example — that there was basically nowhere to go but down. The Astros fall a bit less and are still projected as the best team in the AL West, but Steamer doesn’t expect the team to get the same kind of production from part-timers like Jake Marisnick or Marwin Gonzalez again, and it will be difficult for Dallas Keuchel to repeat his Cy Young performance from 2015.
By and large, though, we don’t see a ton of huge swings. Here’s a plot of every team’s expected record (according to BaseRuns) from last year versus their projected record (also according to BaseRuns) for 2016.
The Red Sox stand out on that graph as well, but most teams are in the same general vicinity of where their context-neutral performances from a year ago pegged them. And that big giant clump in the middle is why every team in the American League believes they can contend this year; the AL is full of okay-but-not-great rosters, and no one is ready to write-off 2016 when everyone is starting from mostly the same place.
Of course, these projections aren’t gospel. Once all the ZIPS forecasts are done, we’ll fold those numbers into our depth chart projections, which will change some things, and there are plenty of other good projection systems out there as well, many coming to different conclusions. So don’t go put the mortgage on these numbers, especially given how big a roll context plays in determining a team’s final record. But in terms of setting up a decent baseline for the upcoming season, it’s nice to have win-loss numbers to give us an idea of the various gaps between teams.
Dave is the Managing Editor of FanGraphs.
I’ll bet the Pirates play better than ~.525 ball.
Projections are conservative and are different than predictions.
But done right, they essentially should be about 50/50 on any one team.
what does 50/50 mean in this context?
50% change they are better, 50% change they are worse.
change change to chance
How did i make the same typo twice in a row?
auto correct made the same correction twice in a row.
There are lots of way of projecting. One would use the median guess (as you say) others would try to make an unbiased prediction. There are other methods too (such as the MLE) that make the type of wild guesses that fans like to make. Surprising, really, that sites like FG don’t use the MLE to pander a bit more.
So what’s the point?
If you were to forget the actual wins and losses, but just think of it as:
Cubs – Favorite
Cards/Pirates – Fighting for 2nd
Reds/Brewers – Fighting not to be 5th
Would that be better?
I believe the Reds/Brewers will actually be fighting TO be 5th.
Once the Cards trade for Blackmon, and then sign Davis they will project to have the best record in Baseball.
It’s a good baseline projection for the Buccos though and given how these systems can’t always figure in things like pitch framing and shifts I would agree with you.
Dave, can you give some more detail how FG’s comes up with the runs-against estimates? Does Steamer generate the components needed to make a BaseRuns-against estimate? Or do you instead adjust the Steamer ERA projections using the run environment you estimate from the offensive components? It’s clear you’re not simply taking Steamer ERA and adding unearned runs.
It’s kind of amazing to consider that the Giants added Cueto, Shark & Span and it didn’t even move the needle.
Shark and Span aren’t all that good. I mean, good enough to sign and give $$$ to. But the Giants weren’t particularly bad at those positions last year.
Really? CF defense and SP depth were the Giants biggest weaknesses last year IMO…
The problem is steamer doesn’t really like what turned out to be their strength – the infield. All of the infielders are projected to regress quite a bit.
All of their infielders exceeded expectations and I think in most cases exceeded anything they’d done before. It’s not impossible that the improvements are all repeated (or improved upon), but that is not the best guess.
Either that or we are to believe that the Giants basically have four (4) 5-WAR true talent infielders at the moment.
I get the feeling you didn’t watch the Giants much last year. Beyond Bumgarner, the rotation was awful based on just about any metric.
I know it doesn’t fit the narrative, but the offense was good. SP was terrible. If Span plays at all, he will be better than Pagan. If Cueto and Shark are 2-3 WAR each, that would be an upgrade over what Lincecum, Petit, Heston and Leake gave them last year as starters.
I’m guessing Steamer doesn’t think Panik and Duffy will combine for 9 WAR again, that’s more than fair. I can’t remember if it was ZIPS or Steamer, but one pegged Posey as playing around 120 games. I think the needle not moving is more about the bats regressing (Crawford, Panik, Duffy, Posey, average year for Pence) , rather than offseason moves.
Shit, if Shark hits his projection (I think he will do better in that park behind that defense), he will deliver twice the value as the four-headed monster of Petit, Lincecum, Heston and Leake.
Same with Span. I think he can do better than -0.5 WAR if he gets on the field.
Very possible, but that’s more of a prediction… a projection is the most likely outcome… which tends to be conservative.
I’m a Cubs fan, I watched my team win 97 games and then add Heyward, Zobrist, Warren, and Lackey, along with improvements from the young players… and they’re projected to win 95… I think they’re going to win 100+, but it’s very possible they win less than 90… hence the 95…
Seriously, I would think any Cub fan would just temper things a bit and repeat to themselves “Washington Nationals” over and over.
This Washington Nationals = Chicago Cubs meme is so intellectually lazy that its tiresome. For starters, they are independent probabilities. Second, many teams exceed or fail to meet their projections for any variety of reasons. The Nationals were felled by injuries and a terrible manager who lost control over his team. Any team is at risk of injury. The Cubs clubhouse is a little more stable than Washington’s.
“The Cubs clubhouse is a little more stable than Washington’s.” Not sure you’re on the right site.
“This Washington Nationals = Chicago Cubs meme”
That’s not a thing.
(Someone pointing out, once, that the Nationals failed to meet expectations last year doesn’t make it a thing.)
It’s not one time. It’s every Cubs thread this offseason (of which there are plenty).
So…there are some people saying “They should be really good!” and some saying “whoa, not so fast! Nothing’s certain!”
In other news today:
Sun rises, sun sets
Earth spins on its axis
A Kardashian does something cringe-worthy
@ bly
In Washington, the team’s successful closer was displaced by a trade acquisition. The displaced closer pouted and sucked and then broke his hand punching his locker. In Chicago, the team’s successful shortstop was displaced by a rookie. The displaced shortstop fixed his mechanics and won the starting 2B job, performing much better after being benched.
To be sure, the respective situations of Drew Storen and Starlin Castro are more complicated than what I’ve described above. And to be sure, there is much more to judge the stability of a clubhouse than that. But the team managed by Joe Maddon had quite a bit more cohesiveness than the one managed by Matt Williams. They also didn’t have any choking incidents.
It is a mistake to use last years results as any kind of baseline for this years projections. You said the signings did not move the needle. Can’t look at it this way.
vr, Xei
Yet, that’s exactly what we are doing, because we are humans. I agree with you, not the best means of comparison. But the response was more toward “the Giants weren’t that bad at CF or SP last year.” Which is true. They were awful.
I’m more shocked at the Dodgers projections. There are tons of question marks, but the projection systems seem to favor the absolute best-case scenario for them all.
no, the projection systems never favor best case scenarios for any team.
The team is very talented and has plenty of depth in the pitching staff. The rotation that opens the season in LA will not be the same as the rotation that pitches in September and into October.
Because you can’t expect Belt, Panik, Crawford, and Duffy to all repeat their excellent 2015 seasons.
Why not?
They might repeat those seasons. He’s not saying they won’t. But you should never expect four guys who all just had career years to do it again. Even if you’re super optimistic and project them all to play at that level, you should expect at least one to underperform, get hurt, etc.
Really looking forward to the Red Sox continuing to prove how junk the aptly-named Steamer is.
Steamer can only crunch the numbers that used to show Ramirez and Sandoval as guys worth 2 WAR when they stunk and 5+ WAR at their best.
Last year, Steamer couldn’t smell the upcoming stink of these guys becoming the worst two everyday players in baseball. If you put them both up to +2 WAR, that’s most of the bullishness on the Red Sox.
steamer is projecting about what? 8 more war out of those two than they put up last year? seems generous.
i agree. i can see Hanley being worth 2 WAR as a DH if healthy (iffy in itself), but Sandoval just seems broken. last year was in large part just a continuation of trends that have been present for the last several years. his ISO continued to decline and he displayed ongoing ineptitude from the right side of the plate. Steamer has him putting up his best ISO since 2012, which is extremely unlikely given the giant LF at fenway and his uselessness from the right side. his defense cratered more than expected granted, but with that terrible body it’s not too surprising. in his previous 3 years, UZR only figured he was average and DRS thought he was below average anyways, so it’s not like he was good to begin with. i can see a bit more luck on balls in play and a little more defensive value just from sheer random regression, but what’s an empty .260 BA with awful baserunning and poor defense worth?
ack i meant giant RF at fenway. his uselessness as a righty doesn’t let him take advantage of the monster
@icehawk10
Could you expand on this?
“Steamer has him putting up his best ISO since 2012, which is extremely unlikely given the giant LF at fenway and his uselessness from the right side.”
A quick look at Sandoval’s career spray charts shows a hitter with a pull-heavy distribution from the right-side and an oppo-swing from the left side, this would appear to be the perfect fit for Fenway’s LF. Tony Blengino went very in-depth on this last year (http://www.fangraphs.com/blogs/hanley-panda-and-the-monster/):
While the switch-hitter obviously logs more at-bats from the left side, he hits more fly balls to LF/LCF by a fairly sizeable margin — in 2014, he hit 75 fly balls to LF/LCF and only 57 to RCF/RF.
“but what’s an empty .260 BA with awful baserunning and poor defense worth?”
If you kept his Steamer hitting projection (104 wRC+) but shaved a win off for defensive/baserunning incompetence, that’s a 1 WAR player over a full season. So still a 3 WAR upgrade over 2015, but 1 win less than Steamer thinks. However, the Sox are projected to be 5 wins ahead of the 2nd best team in the division, so this would hardly impact the complexion of the division
you’re right about the pull/push tendancies. thing is, he still has to hit the ball pretty hard to the opposite field to make use of the monster. his 10 doubles to LF last year were not meaningfully different than what he did in SF, despite the uber friendly conditions at fenway. from the right side he’s an extreme GB hitter and terrible in virtually every other way, so doesn’t take advantage of the wall at all.
the skills have been degrading for a while now. i just dont see an average offensive performer anymore, which makes him not good at anything really. like i said, some modest regression in some areas is in order, but it is a long way from where he was last year to approximating an average MLB 3B
I think anyone who either watched Sandoval play last year or simply glanced at his Fangraphs player page can agree that his 2015 stats look more like an outlier than indicative of a new true talent level, so his amount of doubles to LF in 2015 aren’t meaningful when projecting what he will do in 2016. The question we want an answer to is how many balls will he hit to LF when healthy and performing to his true talent leveL? This is what Blengino looked at when he analysed Sandoval’s batted ball tendencies last year:
“On all BIP, Sandoval’s REL PRD is a slightly below average 98. General contextual adjustments boost that figure to 108, but the Fenway effect ramps it up substantially higher to 117. Add back the K and BB data, and Sandoval’s 2014 REL PRD is 109, his true talent level is 120, and his Fenway-adjusted mark is 128.”
So let’s say his true talent levelOn all BIP, Sandoval’s REL PRD is a slightly below average 98. General contextual adjustments boost that figure to 108, but the Fenway effect ramps it up substantially higher to 117. Add back the K and BB data, and Sandoval’s 2014 REL PRD is 109, his true talent level is 120, and his Fenway-adjusted mark is 128.”
So even if you discount his true talent level heavily from what Blengino estimated, he should still be an above average hitter.
Any color commentary around the big drop in expected runs for the Kansas City Royals? What’s the narrative for the Royals if they outperform projections again this year? Luck or are they doing something that the models can’t pick up on?
Cluster Luck had them as the #1 benefactor in 2015.
https://thepowerrank.com/cluster-luck/
Take cluster luck, plus the notion that we know less about defensive WAR than we do offensive WAR, and perhaps what, if any degree reliever WAR is off a bit, and that helps me explain with some rational thought why the Royals have beat projections for a couple years.
Was the 2014 KC squad generally considered cluster lucky as well? I don’t remember.
Uhh I don’t know if Cluster Luck existed for MLB before 2015. Anybody else? I cannot find 2014 on his site.
Not calculated the same as that clusterluck site, but the “clutch” on FG’s sites suggests that the Royas did indeed benefit greatly from sequencing in 2014:
http://www.fangraphs.com/leaders.aspx?pos=all&stats=bat&lg=all&qual=0&type=3&season=2014&month=0&season1=2014&ind=0&team=0,ts&rost=0&age=0&filter=&players=0&sort=11,d
http://www.fangraphs.com/leaders.aspx?pos=all&stats=pit&lg=all&qual=0&type=3&season=2014&month=0&season1=2014&ind=0&team=0,ts&rost=0&age=0&filter=&players=0&sort=13,d
Well, ‘cluster luck’ basically wins best name for a sabermetric at least.
On a related note, has there been any work done to see if the Royals are the kind of team that can somewhat control positive sequencing? Maybe better said, have there been any connections between certain baseball skills/strengths and positive sequencing?
I’m definitely starting to get the feeling that the Royals skills as a whole (bullpen, defense, base-running prowess) twist sequencing probabilities into their favor.
Like anyone on this site would admit there’s a problem with the system…
Admit the system doesn’t predict with 100% accuracy? Sure, constantly
Acknowledge there are things the system doesn’t consider? Absolutely, all the time.
Will fans of the latest team to have merely two seasons of outperforming projections come down from their tower of smugness to offer some possible explanation as to what the projection system might be missing, rather than simply taking shots at the system that dared insult their team? No. Never.
Which system exactly? Baseruns projections, or how baseruns is used to predict # of wins?
Remember that year the Diamondbacks completely wrecked their Pythagorean estimate and Bob Melvin won NL Manager of the Year? Or those years when the Orioles outperformed their projections, and we were supposed to believe Showalter was eking hidden value from his roster the numbers didn’t capture? Or those other years when the Angels repeatedly beat the projections, and it was Scioscia who was the genius who knew something the computers didn’t?
These are very rough, very early projections. But we know there are aspects the projections don’t fully capture, and as dtpollitt says, it’s possible the Royals are disproportionately strong in those areas. However, we also know there’s always an element of luck involved, and gambler’s fallacy aside, it doesn’t always “even out” even over 162 games — which, as Cameron notes, is why the spread of results in real life for any given year is always greater than the spread in the projections. In fact, it may not even out over many more games than that. When a team outperforms for a couple of years in a row, it is entirely possible it is more than luck — but, as hard as it may be to believe, it’s also entirely possible it is nothing more than luck. As much as we might like the narrative to be otherwise.
We have reason to be somewhat suspicious of the projections, simply because we understand them and know their limitations. We should be at least as suspicious of “magical” teams even though, or especially because, we don’t understand them.
If you ask 30 people to flip a coin 162 times, you will both get a wide spread in those 162 outcomes among those 30 people, and you also observe some long streaks of one person getting one outcome in a few instances. It’s just how probability works.
It would actually be more unlikely to never observe any team beat their projections a few years in a row. That would actually suggest that there is some weird bias in the system.
Quick and dirty simulation using Excel and RAND()
Thirty people flipping coins got heads in a range of 62 to 92 times with three runs of 10+ heads or tails in a row
Do it again; 73-91 with five streaks
Again, 66-97 with five streaks
Again, 69-94 with seven streaks
Again, 68-96 with seven streaks
So, that is the sort of randomness involved
I’m definitely NOT a fan of the “God of Gaps” explanation of things.
If the Royals significantly outperform their projections 2 years in a row, rather than just chalk it up to luck … which seems to define the idea of “we don’t know” rather than just its implied “randomness” (unintentionally or otherwise), then I would first assume that there is something within or absent of the projections that accounts for the difference.
The whole “luck” narrative doesn’t jive with me very well. It’s a throw up of the shoulders, a “who knows?” … rather than a “we need to find out” or “there must be some valid explanation that doesn’t rely on the nondescript luck”.
Weren’t there articles on this very site explaining that basically one season the Royals never got hurt? Don’t they have the best bullpen in baseball? And aren’t we aware that that we, as a community, may undervalue relievers? Maybe we don’t value their defense enough? I would like to think that we would explore every option, test every possibility, before we just say “aw it’s luck” … because luck as we are using it in this sense doesn’t seem to be significantly congruent with the luck that we use to describe statistical randomness.
But, I could be wrong.
I’m obviously biased as a Blue Jays fan, but other than Pillar and Donaldson should we expect significant downwards regression? I can understand dinging Bautista and Encarnacion half a win just for getting older, but that will easily be offset by the extra 200-300 PA from both Tulowitzki and Travis.
check the zips projections. they are much more sensible projections for the jays. as they were last year.
Shouldn’t any projection system worth its salt expect Tulo to get hurt again?
Last year, the AL playoff projections for 2015 were:
CLE: 35, BOS: 34, LA: 29, SEA: 26, DET: 23, TOR: 17, BALT: 7, CHI: 6
OAK: 6, N: 3, TB: 3, HOU: 0, KC: 0, MN: 0, TX: 0
Well, you can tell how those went. And not one nationally recognized expert or baseball statistician – literally, not one – predicted KC to win 90 games, much less the playoffs. (I read on some sports statistics site that even in game 5 of the 2015 World Series, KC should still have a losing record based on their predictive model of offense and pitching). It could be that the game is entering a new era, shifting away from home runs (which won’t help you win) and power pitching (it’s overrated; contact hitting beats it most of the time) to contact, speed, defense, and good relievers (akin to Whiteyball). Of course, you can’t quantify team culture – if a guy swings for the fences every time, as opposed to buying into the team system and leadership and putting effort into contact and moving runners “down the line,” you’re more likely to lose. These projected and predicted models may have to be adjusted in the next few years.
what?
This study was done. Power pitchers are better than finesse pitchers, against all batters. By proportion, contact hitters have a smaller performance gap when comparing their performance against power pitchers and finesse pitchers, but they too are better against finesse pitchers.
Want to guess how the 2nd best team at not striking out performed against their baseruns?
It was the A’s, who underperformed their baseruns expected win total at a near historic rate.
RL–
Make sure you post your predictions for 2016! So we can see how you do beforehand, and not just after-the-fact sniping that is so, so easy (and tired).
Probably topical to bump…
http://www.fangraphs.com/blogs/the-difference-between-predictions-and-projections/
Interesting article. I don’t agree with the definitions specifically for prediction. Prediction is a forecast of the future. Projection is a forecast of the future based on past/present data and the accuracy of the projection is dependent on the input of the data. Ms. Cleo has made many predictions over the ages that do not fit the definition of prediction or projection in the referenced article.
Fair enough.
A more pointy-headed guy’s definition:
“A prediction is a probabilistic statement that something will happen in the future based on what is known today. A prediction generally assumes that future changes in related conditions will not have a significant influence”
“In contrast to a prediction, a projection specifically allows for significant changes in the set of “boundary conditions” that might influence the prediction, creating “if this, then that” types of statements. Thus, a projection is a probabilistic statement that it is possible that something will happen in the future if certain conditions develop.”
http://sciencepolicy.colorado.edu/zine/archives/1-29/26/guest.html
I’m kind of shocked that the Yankees are projected to be one of the better teams in the AL. I guess the projections see strength in being just OK at all positions except RF, and semi-competent depth in SP and RP. It could also be the fact that the projections don’t think anyone else in the AL is particularly good.
3 of the top 10 projected teams are in the AL east so it is hardly due to subpar competition.
why? they basically have 3war at every starting position and rotation slot, plus the best bullpen in baseball.
where does that follow? you can see 3 WAR from C pretty easily, and maybe if you squint hard at 1B and 3B, but where else? seems pretty unlikely from Ellsbury or Gardner or Gregorius or Castro. Tanaka and Pineda assuming relatively good health, sure, but that’s about it. in general it’s not a team that has a whole lot of high level talent that doesn’t come with massive health concerns, and features mostly middling decline phase performers after that.
conveniently I already nerded up the answer for you. here is their average bwar/fwar over the last 2 seasons, paced to 650pa or 32 starts:
Gardner 3.4
Rodriguez 3.0
Teixeira 2.9
Ellsbury 2.9
Headley 2.9
McCann 2.8
Castro 2.0
Hicks 1.6
Tanaka 4.3
Pineda 4.2
Severino 3.6
Eovaldi 2.3
Sabathia 1.3
you can hope for a sudden massive decline I guess but I wouldn’t count on it. They have a couple legit old guys but really aren’t that old overall….and suddenly they have a nice crop of kids trying to push their way into the lineup too.
This was a solid effort, but the first 6 names on this list are at the age where the typical -0.5 WAR/year decline is applicable.
Also not sure why we wouldn’t just look to the projections since ZiPS and Steamer are both available for the Yankees (and after all we are commenting on an article about projections). Both ZiPS and Steamer have projected McCann and Headley as close to 3 WAR players, but no other hitters. On the pitching side, an average of ZiPS/Steamer would have Tanaka and Pineda as the two pitchers that could threaten 3 WAR.
So I would agree with ice_hawk here, saying they have 3 WAR at every starting position and rotation slot is incorrect
this mini discussion started with questioning how the yanks’ projections could be so high….not sure using projections to support the projections really answers that question.
No I don’t actually have an issue with the Yankees team projections. I just thought that ice_hawk is right to question the suggestion that they have 3 WAR players everywhere, and that using last 2 years WAR wasn’t a particularly convincing defense of that suggestion because 1) their best hitters are post-prime and thus should be expected to worsen slightly and 2) a number of those players should not be projected for 650 PA/32 starts due to injury risk
Sure anyone looks good if you pace them up to 650pa or 32 starts, but that’s ridiculous. Only one guy even touched either of those marks on the team.
Pretty ridiculous to artificially increase each guys WAR just for your argument’s sake.
It’s not my fault that the Yanks are projected as a good team.
I’m just trying to help the people questioning it understand exactly how good the talent on that team actually is, and why they are projected to be a good team, yet again.
Yes, they have age and injury issues, but they are extremely exaggerated, and every team has those issues.
You can choose to underrate the “Old and Broken” Yanks yet again, but I’d advise against it, especially since they have a nice wave of young talent starting to break through.
Funny thing on the Astros, they significantly under-performed their BaseRuns last year and only won 86 games. So while BaseRuns projection shows that they were incredibly exceeded their likely true talent in scoring/preventing runs last year, they are still projected to win a game more than last season due to their randomness in sequencing hurting them in the standings last season.
Astros bullpen, while being 2nd in WAR, was 19th in WPA. I was not the biggest fan of the Giles trade, but it is easy to see why they wanted him.
I think it would behoove fangraphs to analyze the team projection success rate of Steamer vs. Zips.
iirc, Steamer was a much bigger reason why the much loathed Red Sox were overrated last year than Zips was, and there’s a similar difference between the two projections this year.
incidentally, Steamer was a much bigger reason why my beloved Blue Jays were underrated last year than Zips was, and there’s a similar difference between the two projections this year.
Steamer has some ‘splainin’ to do, imo.
it always seems to nail the indians, though, for some reason
You don’t recall correctly, ZiPS was very high on the Sox in 2015.
When Fangraphs released their 2015 projected standings with both ZiPS and Steamer, the Red Sox were projected to win the division by 5 games, as they are now with just Steamer projections.
Last year they also released first with only Steamer, which gave them a huge lead. That lead was cut significantly once zips was added, and there would have been no lead at all if we had glimpsed zips by itself first.
The difference is even bigger this year.
If we use zWAR projected based on Fangraphs’ current Depth Charts playing time:
Donaldson 6.8 —– Betts 4.9
Bautista 4.7 —— Bogaerts 3.0
Tulowitzki 4.2 —- Pedroia 2.8
Martin 3.7 ——– Ortiz 2.6
En’cion 3.6 ——- Bradley 2.2
Pillar 3.1 ——– Hanley 1.8
Saunders 1.8 —— Pablo 0.9
Travis 1.2 ——– Swihart 0.8
Smoak 0.6 ——— Castillo 0.7
Pompey 0.5 ——– Young 1.1
Cola 0.3 ———- Hanigan 0.7
Goins 0.3 ——— Holt 0.6
Thole 0.0 ——— Shaw 0.0
Barney 0.3 ——– Vazquez 0.3
Jays 31.2 ——— Sox 22.4
Stroman 3.5 ——- Price 5.4
Estrada 2.2 ——- Buchholz 2.6
Dickey 1.8 ——– Porcello 2.3
Happ 1.5 ———- Rodriguez 2.0
Chavez 1.4 ——– Elias 1.1
Jays 10.4 ——– Sox 13.4
Storen ??? ——- Kimbrel 1.2
Cecil 0.9 ——– Uehara 1.1
Osuna 0.9 ——– Kelly 0.9
Sanchez 0.6 —— Smith 0.8
Hutchison 0.4 —- Tazawa 0.5
Loup 0.3 ——— Ross 0.2
Tepera 0.1 ——– Layne 0.2
Jays ~4.0 ——— Sox 4.9
Total: Jays 45.6 — Sox 40.7
What we have here is a massive difference in the two projection systems – Steamer thinks the Sox are a good 5 wins better than the Jays, Zips thinks the Jays are a good 5 wins better than the Sox.
Based on last year, I’d lean towards Zips having a better read on the players on these rosters, but we’ll have to wait and see.
They had a pretty huge lead even after ZiPS was added. Could you please post what the projected standings were when it was 100% Steamer?
I also find it hard to believe that there would have been no gap had the projections been 100% ZiPS based. The graphical depth chart for the 2015 Jays added up to 36 WAR, the 2015 Red Sox added to 41.
if you give me the playing time allocations from the 2015 depth charts (I can’t find them) I’ll calculate the team zwars.
I don’t have them either, however you can simply pull up the 2015 ZiPS release for each team. The difference is large enough that I think even if you got creative with the playing times you wouldn’t be able to move the needle that much. I’ve found a tweet from Dan on 5/4/2015 that projected the Red Sox to finish with 85 wins and the Jays to finish with 81. At the time, the Sox were 13-14 and the Jays were 13-15. So I’m still confident that you are wrong about this. Every projection system thought the Sox had a superior roster, including ZiPS
Playing time projections are crucial in this case, though, given how much of Boston’s projected value came from their supposed great depth beyond the projected starters.
No, as I already mentioned the Sox were 5 WAR ahead of the Jays just on the graphical depth chart which is made up of starting players only. And as I also already mentioned, ZiPS was still projecting the Sox to finish with a better record a month into the season.
Look I see what you’re going for here. You want people to think that ZiPS projected the 2015 Jays to be better than the 2015 Sox, so you can use that to convince people that ZiPS is more reliable since they are projecting the 2016 Jays to be better than the 2016 Sox. But you are wrong, and I think you know it. Everyone thought the 2015 Sox would be better than the 2015 Jays, and ZiPS was no exception.
I thought the results showed Zips was much better at projecting veterans, while Steamer was better at younger players. I thought I read that on BtB or some place like that.
Actually I was being quite honest about the importance of playing time projections to last year’s zips.
for example, this is how I think the opening day lineups compared (including injured players and not their replacements):
Porcello 3.5 —— Stroman 3.1
Miley 3.1 ——— Dickey 2.7
Buchholz 2.0 — Buehrle 2.6
Masterson 1.8 – Hutch 1.8
Kelly 1.5 —— estrada 1.2
Uehara 1.3 —- cecil 0.7
Tazawa 0.9 —– loup 0.6
Layne 0.4 ——- sanchez 0.5*
Varvaro 0.4 —– redmond 0.3
Mujica 0.2 —— hendriks 0.3*
Wright 0.0 —– hynes 0.2
Breslow -0.1 —- Delabar 0.0
Pedroia 3.8 —- Donaldson 5.1
Betts 3.1 ——- Bautita 4.2
Ortiz 2.9 ——- martin 3.6
Pablo 2.8 —— Encarnacion 2.9
Hanley 2.7 —- reyes 2.7
Bogaerts 2.3 – Saunders 2.1
Napoli 2.0 —- pompey 1.2
Vazquez 1.6 – travis 0.9
Victorino 1.2 – smoak 0.2
Craig 0.3* —- valencia 0.6
Nava 0.6* —- carrera 0.6*
Holt 0.6* —– Izturis 0.1
Hanigan 1.4 – navarro 1.6
40.3 —- 39.8
* = cut their war in half for bench/bullpen roles as they were projected for starting playing time
I’d rather do a more precise projection, though.
Somebody did the math in October. In 2015, Steamer and Zips performed about equally having an RMSE difference of .01. Over 2014 and 15, Steamer outperformed Zips by a good margin.
Source: http://tangotiger.com/index.php/boards/viewthread/442/
That actually shows us that Zips alone beat Zips/Steamer combined, which means zips beat steamer by a healthy margin, as I expected.
Half true. You’re right I misread it but you ignored that zips/steamer combined beat zips alone over a two year period by a much healthier margin so steamer wins there.
yeah possible but then I’d want previous years too.
I was pretty skeptical about the Steamer projections last year so i’m at least glad to see some confirmation of that.
But it’s an interesting thing to look at this year – the BOS v. TOR comp in particular looks dramatically different depending on which projection you use.
I like OLIVER
He ruined the Brady Bunch though.
(In the same way you can “ruin” a mayonnaise sandwich by adding paprika – i.e., it’s probably not that good to begin with.)
race for the number 1 pick between the Reds, Brewers, Phils and Braves is going to be awesome…
I think the Crew have the inside track.
Correct me if Im wrong, but these projections dont actually model the season in any meaningful way, right? Theyre just the projected baseruns of each team, assuming that the players on each team perform at their mean projections and receive the playing time projected by Steamer. Right?
If thats the case, Im skeptical about their usefulness, even relative to the fundamental uselessness of any projection. If theres no modeling for the probabilities of the wide range of different outcomes for each player, the projection is missing a TON of the information that actually drives seasonal win totals. So, for instance, the projection will be flat-out useless about a team like the 2015 Cubs, which had a ton of elite prospects in the upper levels. Such players will provide tremendous value if they prove good enough, but wont be given the playing time to hurt much if they arent good enough, so a model that uses flat performance/pt projection wont accurately capture the upside of such a team and will generally underrate them. Similarly, the projections will be useless for teams that have extraordinarily good/bad depth to cover in the event of injury or unexpectedly poor play.
Now, I say Im skeptical, but Im not dismissive. I dont actually know how much that missing information will affect a projection. Maybe it doesnt make a huge difference. But I do lean toward not taking a projection seriously if its willfully ignoring a significant amount of information.
I am an econometrician and typically models like this arnt used in business or industry the way fangraphs and other sites are attempting to use them. There are macro models and there are microeconomic models. Micro models are hyper specific, usually only applicable to a lone company or business and are only accurate to for that business (or occasionally a totally homogeneous entity like a franchise). Macro models apply more to a segment, industry, or economy as a whole.
It is generally not advised to make macro generalizationa from micro models and vice versa. Using WAR to predict team success is using a micro model (the player) to model a macro concept (the team). So you have a model projecting the projections of another model.
Bryce Harpers World Series champion ring is ready,started working on it after the M.Scherzer trade.Should I begin on Joe Maddens.
Once you do a refresher on how to spell Maddon.
These look good to me.
I wish I could see the algorithm specifically used in WAR war projections, aging or improvment curvs, regular season records, and the weights given to each. I am an econometrician with PhD’s in stats and econ. I’m also a Royals fan and am curious to see what stat(s) is under/overwighted in the case of KC. 2013, 2014, and 2015 have seen KC significantly overachieve while others have underachieved without injury or significant under performance from individuals.
One of the big areas is that 2015 saw several key Royals play as key players. Steamer projects that Hosmer, Moustakas, Morales, and especially Cain put up career years in 2015 rather than putting up sustainable performance that will carry over into 2016. Even role players like Dyson are projected to produce less WAR in 2016 than in 2015 (even with Steamer assigning Dyson more plate appearances). The system also has Volquez worth a little less and the bullpen worth less with no Holland or Madson and Duffy moving to starting.
So, that’s the big question as I see it. Was 2015 a wonderful combination of several guys having career years, or was 2015 the year Everyone Improved and that higher level of production is now the baseline moving forward?
Team projections have nothing whatsoever to do with prior years’ results other than at the player level and only with respect to the contribution a player’s prior raw context-neutralized stats makes to his projection. Nor should they.
Now, you could make a (small) argument that some aspect of prior team stats could be incorporated into a team projection but only to the extent that there is a plausible reason (i.e., something that is not being captured by the individual projection). For example, one source of error for projections is personnel and playing time. It is plausible that some teams are better or worse than others at keeping players injury free, optimizing roster decisions, making late pre-season and in-season trades, etc.
Even then, the addition into the model of “team stats” not captured in the player projections would necessarily have to be small and would be very difficult and speculative to incorporate.
I have to imagine this type of adjustment is the second of three steps that Vegas must use in setting an initial over/under type line:
1. Best estimate from a forward-looking (projection) model
2. “Manual” adjustments to the model results based upon material known factors that are also known not to be captured in the model.
3. Any adjustment needed to cover known biases among bettors (ex. Cowboys spread shifting toward their opponent because of all the Cowboys fans betting).
Dave could you elaborate on the process in the first paragraph, specifically what causes the win projections to vary from WAR projections?
For example the Cubs are 52.4 team WAR which would be 100 wins but project to only 95 (now 96) wins. The Nationals project for 2 team WAR more than the Yankees but the win projections are NYY 87 and Wash 84. What is it about these teams that causes a 5-game swing? Seems like the NL teams move down which is strange since I assume the AL is the better league (not having 5 bad teams).
Was expecting more outrage from Mets fans.
We wasted it all on Jeff’s compiling of the NL projections off-season.
Using steamer etc, Mets came in at 79-83. Thought the pitching projection was low based on noahmets community research and said I thought they were an 84-85 win team and could push 90, but acknowledged I was total homer. Hurray for me and lucky guesses:
http://www.fangraphs.com/blogs/the-team-projections-and-you-national-league/
This year I’d say about the same, although I’ll go with an 87-88 win team but could push 92-93 or so. A lot depends on mid-year deals, etc.
But there isn’t much point in arguing with projections.
I’d like to see something like confidence intervals; which teams and players are we least certain about? The way Steamer is displayed to us is as a single outcome, when surely the simulations produce many possibilities and choose an average figure.
How the heck are the Diamondbacks still projected to have a losing record?! Right now, they’re my pick to win the NL West.
Maybe your pick is wrong.
It’s pretty pessimistic to think the Jays run production will fall by 3/4 of a run per game, with the same offence as last year (but without the half year of Revere, and with a full year of Tulo)