As so often happens, the rollout didn’t go quite as expected.
Spring data didn’t actually start loading until Monday, but should be on all player cards now. (Just the 2013 spring data).
The pitchers forecast pages were supposed to become analogous to the hitter pages – primarily through the addition of their top 20 comps. But it looks like that change didn’t make it all the way through the list – it seems to cut off somewhere between CC Sabathia and Jeff Samardzija. And there seem to be some odd problems – Trevor Bauer’s comp list has reasonable personnel but strange numbers, which I think might relate to having less than three years of work; I think innings are getting divided by 3, but runs are only getting divided by 2.
But in Miguel Gonzalez’s case, I find it very hard to believe that comp list belongs with that starting line – as though two jobs were running at once, and the other guys comp list made it in. Let me dig into those issues.
Update – Bauer problem solved, and yes, the less than three years bit had a lot to do with it.
Just as I’m getting ready to throw in the towel, I finally track down the bugs in the pitching projections that have been driving me up a wall for more than a month. And damned if it isn’t always the stupidest, piddlingest crap that is the hardest to find – I’d managed to switch the indices between the input league and the output league.
The way the program is supposed to work, the real stats are compared to the real league, and then adjusted to the new league. Somehow, don’t know when or how, the part that was supposed to move the stats from translation space back to the expected 2013 reality was going off on a weird angle.
And yes, that was holding me up from writing. I couldn’t face up to an audience without having a proper product (even if the product, with all its faults, was going out already.)
So start by checking out the Projected Standings tab. That’s the culmination of this winter’s work, as it stands now…early in the spring, with a lot of decisions yet to be settled. I would have liked to have had this done two months (and two of my own drafts) ago, but life doesn’t always move at our preferred pace.
|AL East||Won||Lost||Runs||Runs A|
Just to be clear about what goes in to the projections: all of the performance rates are based on the computer output. No manual intervention allowed. The computer also sets the upper limit for playing time. For a clear starter, I’ll stay within that limit, with a few exceptions (Stephen Strasburg, for instance, I bounced a good 30 IP above the computer, which is still a conservative 180 innings). Deciding on who’s the starter – and how long they’ll hold it, and who the likely replacements are – that’s all me.
And when I put all of those things together, I get an absolute free for all in the AL East, with four out of five
dentiststeams falling within an 84-87 win range. Looking at the Playoff Chances tab, which runs with these projections a million times, they are all in the 37-50% range for making the playoffs. The Orioles are well back, around 10% – as much as I don’t want it to be so, their performance last year was built on so much good luck that a repeat seems impossible.
Going back to the ‘Projected Standings’, if you click on the TOR tab you’ll get my full projection for the team. The team stats are sortable; just click the header line. Each player should be a link; click there and you should get a full page for the player, with stats for his full career, right through whatever he’s done in spring training so far this year. (I think – I’ve been able to run the ST stats manually, but they haven’t gone through an automated computer run yet). Everything is still subject to change, and I welcome your suggestions on what to add. And write on – topic selection has always been a weakness for me.
And so, to bed.
No sooner do I say I am back, then I come down with an inflammation of my stomach that’s wiped me out for a week.
I estimate that my gas generation this past week has vaulted me into a top 10 source for climate change.
Not that I ever really left. Kept updating everything behind the scenes, just didn’t say anything about doing it. I guess I hit a point where I felt like I was saying the same things over and over again, and wasn’t coming up with anything new. So I walked away for awhile, maintaining things as much as a reference for myself and inviting anyone who wants to to use it as well.
My personal baseball season gets under way tonight, with my first draft of the years…which I’ll follow with an auction for another league this coming weekend.
I’m still not sure I have anything new to say, but maybe I’m over letting that bother me. At least temporarily.
I’m in Phoenix this weekend, doing a interview with NHK television, primarily talking about the prospects for Japanese players in the US…which this year means lots of Yu Darvish. If you’ve seen the projections I’ve made, then you already know I’m very bullish on him. Simply put, he has the best statistics of any pitcher we’ve seen leave Japan – better than Matsuzaka’s – and, yes, I know about and have accounted for the extreme drop in Japanese offense last year. The only negative thing I have to say concerns his workload, where he exceeded 31 batters per start last year. That would normally be a very large concern, but let me make two points. One, that extreme offensive decline meant that a higher than usual number of these batters should have come with the bases empty – so pitching from the windup, and less likely arm strain. Two, every scouting report glows over his smooth, repeatable mechanics, which we also think leads to a higher pitch capacity.
I’ve been doing these interviews with NHK for about eight years now, we think since Hideki Matsui first signed with the Yankees. They are always a lot of fun to do; they feed me really open-ended questions (So, Norichika Aoki?) and just let me go, and I’ll rattle off whatever I can think of, with digressions into whatever sabermetric points I think have relevance to the case or explanations of my process or…really whatever. We did the interview outside, at the Dodgers’ training camp offices, with Clayton Kershaw throwing warmup pitches on a mound right behind me, and various players staring out windows right at us while running on the treadmill inside. Pleasant as I make it sound, the conditions were actually kind of nasty – there was a 25-30 mph breeze blowing in with the occasional grain of sand or dust, and temps only in the low 50s. We had to seek a more sheltered location because the sound man couldn’t hear anything but wind through our mics.
And in the middle of the interview, I made a point about the ongoing discussion about expanding to a second wild card team this year – a decision that was officially announced while I was being interviewed. It took a little while to sort through the logic changes that come with having two wild cards, but the post-season odds calculator is now running that way. The biggest thing to like about it, for me, is re-introducing a real race for something like the expected Yankee-Red Sox collision – it really matters who wins the division and who gets the wild card, beyond just a one-game home advantage.
I sent the latest update in this afternoon. The most significant change from the programming side is a bugfix that was giving pitchers too many innings, by about 3%. So pitchers who did have 200 innings got moved down to about 194, without any of their hit/run/walk/strikeout numbers changing (those numbers had already been set by the program; literally the last step is take all the batters faced that are left over, and convert those to outs, and that’s where the bug was.) That had the side-effect of requiring me to pump about 30 innings back from the starters to the bullpens, with some slight rearrangement for teams who were stronger in one than the other.
There were also quite a few changes to the hitters, but not from the program. As I was preparing for my first draft (tomorrow!) I came to the realization that I was just being too conservative with the PA I was giving to the top, no doubt they’re starting as long as they’re healthy, types of players. Those players got their PA set by my eyeball rather than the computer-derived number, with a strong favoritism towards the median value for the past five years.
I also blew the dust off the playoff odds routine and ran the current projections though it, producing the results you’ll find here. That chart uses the basic projections to set the performance for each team, and then plays the season a million different times using the actual schedule. A key feature of this model is that it does not treat the Red Sox, for example, as a definite .577 team, even though that’s what their latest projection says. That .577 rating is an estimate of the Red Sox value; it may be higher (key players may do better than expected), although there’s a lot more ways it could end up worse (injuries, in addition to just randomly worse performances). This model creates a spread of values used for each “year” run in the model – perhaps .578 in year 1, then .534 in year 2, .601 in years 3, etc etc up to a million. The spread is a curve that has a median value of .577 (meaning that there are just as many scores below .577 as above it) - but it has a longer tail to the low side, so the mean value will be a little below .577.
Unfortunately, the American League looks rather dull – my models clearly and unambiguously favor the Red Sox, Yankees, Tigers, and Rangers to make the playoffs, with only weak challenges led by the Angels and Rays. Fortunately, the National League looks to be almost totally up for grabs. The Phillies and Marlins provide a strong East race (the Nationals break .500, but there’s too much power in the division for them to contend). The Cardinals, Red, and Brewers have been repeatedly swapping positions through my various updates, so it is no surprise that they all have solid chances to win. And the West is even tighter, between the Giants, Diamondbacks, and Rockies – I would say that is officially too close to call.
And I have to note, with some sadness, the passing of Gary Carter. For ten consecutive years, from 1977-86, I have Carter rated (by WARP) as the #1 catcher in the National League; seven of those he was the #1 catcher in both leagues, and in 1982 he was the #1 player in the NL (losing the major league title to Robin Yount’s spectacular season). That creates a strongly affirmative answer the question, “Was he the best player at his position?”, the number 3 item on Bill Jmaes’ Keltner List. By one way I ran for comparing players all-time, I rated Carter as the #2 catcher – only behind Bench. It is ridiculous, beyond ridiculous, that he was not elected on the first ballot.
But Carter’s death holds a special poignancy for me, because, just a few months prior to his initial diagnosis, I had a similar run-in with the medical establishment – a shadow on a CT and overhearing a long and largely unintelligible word that ended with the suffix “-oma”. Mine, fortunately for me, turned out to be of a type that is as passive and benign as Carter’s was aggressive and malignant – but it took a couple of weeks to get the tests done to verify that, and I certainly have not forgotten the fear that came in the meantime. There, but for the grace of God, go I.
I know I haven’t written anything for a week, but I’ve been hard at work. For the last week I’ve been working on making improvements to the forecast algorithm, particularly the pitching side. Through December and January, I was able to incorporate the component scores into the hitter forecasts, and produce an improvement over the whole stat-line approach I had been using. I’ve been trying to do the same thing for pitchers, and just this morning cracked the ‘prior performance’ barrier. While I’m still working on the improvements, I felt good enough about them to incorporate them into new model run. While the changes were dramatic for some pitchers, the effects on teams wasn’t so large – the new method does not shake up the standings. But new standings, new depth charts, and new projections are on-line.
Like the old system, the projection is based on a Marcel-like baseline. Where it differs from Marcel is that the different statistics have different weighted averages, and I use the translated data throughout the process. Strikeouts are very heavily weighted towards the most recent season – roughly a 5-2-1, rounding off, with a small (~15%) regression to mean component. Walks and groundball rates are also highly slanted, though not as much as Ks. At the other extreme, hit rates have essentially no weight for seasons (1.2 – 1.1 – 1), and an 85% regression to mean, which is why stats like FIP work. Once the baselines are calculated for everybody, I go through a similar-player search, and then see how those similar players deviated from their baselines in the following year(s), and apply those deviations to the players. Once all this is done, I run the player through the translation routine backwards to get his stats back into an expected-2012 performance baseline.
I’m testing the new projection system against the set of all pitchers, who had 50 major league innings in 2011, who pitched for only one team in 2011, and who had a major league appearance in 2010. My note says that is 437 pitchers. I’m only looking at five top-level stats for judgment – hits, walks, strikeouts, homeruns, and runs allowed. The projection is normalized to the actual innings pitched in 2011, and I just look resulting errors tabulated.
Here’s the root-mean-square error you get from just using the player’s 2008-10 (major league) stats as your 2011 projection:
Hits 14.03 HR 4.09 BB 8.74 SO 12.72 R 11.48 Sum= 51.06
Same thing, but using his translated stats for 2008-10 as the projection:
Hits 13.76 HR 3.57 BB 8.01 SO 14.26 R 10.66 Sum= 50.26
Lower is better, so this gives us the not terribly surprising result that using reasonably adjusted minor league data in addition to major league data is better than major league data alone. Incidentally, if I use the luck-free runs allowed instead of actual runs – that would be calculated runs, using a normal number of H/BIP and HR/FB – the run error would drop almost a run, to 9.85.
Here’s the results of the program I’d been using to use projections for the past two months:
Hits 11.66 HR 3.27 BB 7.62 SO 13.03 R 9.88 Sum= 45.46
And here’s the results I’m getting from the new version, as of 11:00 PM Sunday night:
Hits 11.48 HR 3.33 BB 7.59 SO 13.16 R 9.47 Sum= 45.03
I’m more than a little annoyed at seeing the strikeout numbers trend backwards; on the other hand, the improvements everywhere else suggest that I’ve got a blind spot – a hole in my swing, as it were – probably a calculation error that should lead to a nifty improvement once I track it down.
In case you were wondering about over-fitting, I am also checking the routines against 2009 and 2010 pitchers, who are not part of the test set. The improvements there are about 3/4 size of the 2011, which suggests some mild overfitting, but not enough for me to be worked up over. At least not yet.
The Nationals added Edwin Jackson to their rotation, which gave me a good reason to run another update.
He’s a no-doubt rotation member, and actually goes straight to #1 on my rankings for them if you rank by innings. He, Gio Gonzalez, Stephen Strasburg, and Jordan Zimmermann make a pretty strong front four – perhaps not quite a match for the Phillies, but very strong. It also puts them in a position where at least one of Ross Detwiler or John Lannan goes underutilized (I have no idea where Chien-Ming Wang fits in, but I haven’t been able to run any projections that like him for 2012).
He pushes their projected record up a notch to 80-82. Their projected run totals – 4th best in the majors in the pitching, 6th-worst in batting – point to a seriously unbalanced team. Their offense has fallen into the “OK trap” – with the exception of Ryan Zimmerman at third, everybody on the roster looks to be an OK player for the position, neither good enough to help the team towards a title nor bad enough that an easy upgrade is available. Even the deservedly-maligned Roger Bernadina projects as an average major league hitter. You’re not going to do any better with a Marlon Byrd or Vernon Wells, who strike me as the kind of guys you might be able to get for a John Lannan. And no, I don’t think Bryce Harper is a savior for this season – the K rates he’s shown make him a bad risk for jumping straight to the majors.
Clay Davenport’s Projections for 2012
generated on 2- 2-2012
Three days later, and there’s a host of changes already made, which you can see here.
I spent six hours yesterday going through the pitcher projection codes, and did track down a couple of bugs, mostly in the way that the BABIP numbers (and thus hits) would track for each pitcher. This dramatically changes some pitchers, while others are barely changed; the main effect on the standings is that everything got narrower. The extreme teams moved 4-5 games back towards the middle, but I feel better about the way they look.
One flaw I noted – which I am guilty of every year – was that I was too conservative on the playing time for known, no doubt about it starters, listing too many guys at 80 or 85% instead of 85-90%. 90 is as high as I go. Cal Ripken, during the Streak, might have gotten a 95. Maybe. I went through and bounced a lot of hitters up a notch.
Surprised by the Tigers’ making the deal with Prince Fielder. I rushed him in to the projection in the simplest way – he gets the 1B slot, Cabrera gets the DH. I’m intrigued by the idea of Cabrera getting some time at third – its a bad-looking set of guys over there, currently. That will deserve a closer look. But it really solidifies the Tigers as Central favorites. Prior to the signing, I had them at 82 wins, with the White Sox at 80 and Indians at 78. Afterwards, they’re up to 88, while every other team in the Central is down 1, so they go from +2 to +9 over their nearest rival.
From the comments, SG says:
Clay, are the relief->starter projections adjusted for role? It seems like Chris Sale, Neftali Feliz, Aceves and Bard are projected to perform at the same rates they’d projected to have as relievers, but with more innings as starters. Relievers moved to the rotation should see a degradation in rate of performance, something like 15-20% higher in ERA. If that’s already in there, nevermind.
The master spreadsheet was making those role adjustments, but it turns out that the data being sent to the output was reading from the unadjusted area, not the adjusted area. That’s been fixed.
Slugging does look strangely low across the board for TB. Almost everyone is about 30-60 points lower than their zips projection. You may want to check those numbers.
There are a couple of ways to go with projections. The first time around, I ran them with adjustments built in so that the sum total of the AL projections, for instance, would equal the AL numbers from 2011. The slugging average of the AL last year was .408; the projections came out to .428. The corrections then went and knocked 20-30 points off everybody across the board.
That is a common problem with projections; they tend to be optimistic. The way the optimism fails, though, is not with an across the board cut, but because specific individuals fall dramatically short – think Ryan Zimmerman, Hanley Ramirez, or Stephen Drew last year. This time I ran the numbers without those adjustments, so the figures for individuals come up. On the downside, though, the numbers are now unbalanced – the sum totals for the hitters will not match the sum totals of the pitchers.
Any chance to get all the player projections linked in one excel sheet?
They’re now on the projections home page.
Anna says: Hi Clay, for the Giants, a couple of things stood out:
- Brandon Crawford is the projected starting SS, not Mike Fontenot
- Bochy-endorsed RF Nate Schierholtz needs a projection!
The Schierholtz thing was another bug – he was in there, but the NL players on each team with the most PA without leading in any position did not get printed out – a kind of DH mixup. And I did see Bochy’s comments on Crawford, I just missed adding them in from an even earlier version (I first set the master spreadsheet up around Thanksgiving). I still have my doubts about him lasting for the whole season, but I did go ahead and reverse the PT between those two.
A few comments on the Mets:
* I think the 72 win outcome is reasonable, and as a Mets fans, I’ll sadly take it.
* Barring injury, I’d be shocked if Ronny Cedeno gets 2x the starts at SS as Ruben Tejada.
* One of the Mets front office staff — I think DePodesta? — has said that the way the team intends to treat SP prospects is by giving them one-way tickets to NYC. That is, when a SP comes up, it is with the expectation that he not go back to AAA. I think you’ll see more of Hefner and Schwindin and less of Familia and perhaps Mejia. You may also see some Matt Harvey, if the team thinks he can go from September call-up to Opening Day 2013 rotation. But that’s a minor nit.
* Another minor nit: Beato almost certainly will not make the team out of camp.
My take on Cedeno is that he has “established starter” attached to his name, even though he sucks. Second glance does show that I probably was hasty in setting all the time to Ronny, so I will be revisiting that. And if the Mets are indeed tracking towards a low-70s (or, revised, mid-70s) figure, they’ll be a little more willing to let a guy who perhaps should be in AAA to stick around.
This will be regularly updated in the “Projected Standings” tab on the header menu. You can access all teams and players from the links.
Clay Davenport’s Projections for 2012
generated on 1-21-2012