Archive - projections RSS Feed

Cameron on college stats

March 4, 2009 in college baseball, links, projections, Sabermetrics with 3 Comments

by Myron Logan

Dave Cameron writes so much about baseball that if you’re not disagreeing with him once or twice a week, you probably aren’t paying attention. That is, of course, meant as a compliment to Dave, as he is right on the money most of the time.

But I don’t think I agree with this post on the utility of college baseball stats. First, he talks about the hazards:

“From the use of metal bats, the huge variances in quality of opponents, some parks that heavily impact run environments, and the smaller sample of games played, there are all kinds of adjustments that need to be made to try to translate NCAA statistics into something that resembles context-neutral. And, once you’ve done all that work, there is still limited value in the numbers.”

Those are all good points, I think, but I don’t see why they render the stats useless. The two biggest adjustments to make are probably the park and quality of opponents, and that can certainly be done. Those adjustments don’t necessarily make the stats predictive, but they are a step in the right direction.

Anyway, Cameron later goes on to say this:

“Good hitting prospects hit well in NCAA ball, but so do less good hitting prospects, and just using numbers, it’s basically impossible to tell them apart. We’re big fans of statistical analysis here, obviously, but we also need to know the limits of what numbers can tell us. When it comes to college performances, scouting reports are what you want – the guys hitting the fields everyday and looking at swings and athleticism do a better job of predicting which college players will hit in the majors and which ones won’t.”

I’m just not sure that is true. Brian Cartwright has some interesting stuff on the projection side. After the stats are translated, he finds that most players perform relatively similarly in college, the minors, and the majors.

If you read one of my Q&A’s with Chris Long, I think you’ll get the sense that he *certainly* does not ignore college numbers, or even put them on the back burner.

It’s also clearly apparent that scouting plays a big role. I think the best projection system for college players would involve combining both adjusted statistics and scouting reports, in some fashion. The only people really able to do that, at this point, are guys like Long, who have the access to tons of scouting reports that we really don’t. You can add things to the adjusted numbers like body type, swing type, bat speed, etc., sort of like PECOTA, combining numbers, physical traits, and actual scouting. I have no idea if this is actually being done, but I’d guess that someone is doing it.

Anyway, I don’t have a clear answer as to how to weight the stats and the scouting. I’m not sure anybody does. But I think it’d be just about as silly to ignore the numbers as it would be to ignore the scouting reports. And when you have both, like they definitely do in front offices, and like we sort of do with Baseball America-type sites, why ignore either?

CHONE on the NL West

February 21, 2009 in links, projections, San Diego Padres with 4 Comments

by Myron Logan

Rally has posted his CHONE team projections for the NL West, and it has the Padres at second in the division with an 80-82 record. Now, second place might not mean that much, as there’s a difference of three wins between second and last (and a difference of five between first and last in this remarkably close division) . But, still, 80 wins is better than I expected from any projection system, and CHONE is very well respected.

There’s some additional discussion at BTF, with a good portion of it surrounding the Pads’ surprising win total. I agree with the posters there that 80 seems a bit high, but hey, what do I know? One reason we do these projections is because our perceptions aren’t always on the money, especially  when we’re talking about 25+ players and how they are going to perform is the coming year. That said, we should probably weigh this appropriately with other systems, and add in our intuiton/additional information when we can. But, still, it’s good news, and I think we can use some of that.

UZR Updates

February 11, 2009 in contracts, fielding, Mike Rogers, player evaluation, projections, Sabermetrics with 1 Comment

by Mike Rogers

Fangraphs just keep getting better. They now have updated the UZR Defensive numbers to include outfield arms and double play runs. Back when The Hardball Times updated their 2008 outfield arms data, Myron looked at it and helped bolster his idea that Brian Giles should be moved off of the right field postion and switch to the oppostie corner. So, lets take a look at Brian Giles now with the UZR outfield arms update.

Brian Giles’ arm is bad. Like, on the extreme end of the worst bad. I’m talkin’ -19.5 runs bad over the last three years.  When the talk about outfield arms was being bandied about as being incorporated into UZR, it has been said that it only really effects the guys on the ends of the category — the very good (Jeff Francoeur) and the very bad (Brian Giles). Giles’ arm is averaging -6.5 runs off of his defensive value on average from 2006-08, and that’s not weighting it at all which would change that since he’s declined each of the last three years: -4.2 in 2006, -5.8 in 2007 and -9.4(!) last year. So, let’s just call it -6.5 runs, over his average of 140 games played in those three years. That would then become about -6.9 runs or we’ll just call it -7.

Defensively, as I noted in the comments of Myron’s post I linked to earlier, the arms ratings really puts a dent in Giles overall value. My comment noted that without arms ratings his defense is +4.42 over the last 4 years. Run that to a Wins Above Replacement conversion using CHONE’s projected .346 wOBA (and a league average of .332), and I get +8.52 offense, +4.4 defensively, +20 for replacement level and -7.5 for positional adjustment, converted to wins above replacement I get 2.4 WAR. Multiply by 0.85 to account for playing time and that’s 2.06 WAR — a bit above-average.

However, if you account for his arms ratings, and to keep it on the 4 year average like I used in my comment, his 4 year arms ratings comes out to -19 (2005 was +0.5 for him in RF). Averaged out, that’s -4.75 runs per year with his arm. Run this into a WAR conversion and his WAR drops to +1.97 WAR. A one-year deal on the open market for a 1.97 WAR player is $9.07 million. Value for a 2.4 WAR player for a one-year deal on the open market? $10.96. Basically, his bad arm is worth about $1.89 million to the bad in terms of his value.

PECOTA’S standings

February 10, 2009 in links, projections, San Diego Padres with 5 Comments

by Myron Logan

Here they are. Also, everything on that page is free, so we can mess around with it a little without worrying about pissing anyone off. Anyway, let’s park adjust some runs scored, runs allowed numbers. In the table below, you’ve got for the NL West, predicted runs scored/allowed, and park adjusted predicted RS/RA:

Team Record RS RA paRS paRA
Dbacks 91-71 818 731 779 696
Dodgers 83-79 761 746 777 761
Giants 79-83 702 716 695 709
Rockies 76-86 829 891 761 817
Padres 74-88 708 770 770 837

Average runs scored/allowed in the NL is 785. So, how about that? Once again, the Padres offense is its strong point, nearly league average, when park adjusted, and right with the top dogs in the division. The pitching, on the other hand, is projected to be much worse, 52 runs below average, and near the bottom of the league with the Pirates, Astros, Marlins, and Rockies.

In terms of the division, the Dbacks are pretty clear favorites, by PECOTA. They are led by a rotation that is anchored by two of the best starters in the league, Brandon Webb and Dan Haren. PECOTA also projects a 3.77 ERA (unadjusted) and 155 innings out of Max Scherzer, the Dbacks young right hander.

The Dodgers could really use Manny, as PECOTA/Clay Davenport project major playing time out of Juan Pierre in left field. While Pierre’s fielding and base running certainly cuts the gap between his and Manny’s value, it’d still be a big improvement to add Manny’s bat. Further, Pierre could be better utilized as a pinch hitter, pinch base runner, and replacement fielder.

As bad as things seems for the Pads, well, they really aren’t *that* bad. The offense is pretty decent, and with a couple of holes plugged up, could really be excellent. The pitching, outside of a few guys, obviously needs to be rebuilt. A bullpen, I think, can be fixed on the fly, in one offseason. The rotation, though, will be tougher. The Pads need, at least, two or three guys who are legitimate starters, depending on whether or not Peavy stays with the club or not. Whether those guys emerge from the organization is still an unanswered question, but I have a feeling there will need to be some shopping in free agency or the trade market, at some point (most likely, next offseason).

With quite a few interesting story lines, like the ownership transition, the high draft pick, and – oh, yeah – the games, this season should be a fun one to follow. And who knows, if some things break right, it’s not out of the question that this ‘09 team makes things interesting.

Hit Tracker projection system

February 6, 2009 in projections, Sabermetrics with 2 Comments

by Myron Logan

I think this projection system, created by Greg Rybarczyk, has a chance to be the best one out there. Why? Because it’s working with a different, more accurate dataset than the other systems, which are all of course developed by very smart people. There is, however, only so much you can get out of the traditional data.

If you’ve been around here for a while, you may remember me talking about what Greg’s model is trying to take into account: the idea that a double isn’t necessarily a double, a homer isn’t a homer. All hits aren’t created equally, especially when you’re trying to predict the future. Why count a bloop double the same as a rope off the wall, when you want to know how a player is going to perform in the future?  We’d all take the rope, right?

For the most part, these things are supposed to even out over time, and that’s what the other projection systems assume. They also regress data back toward some mean, to try to account for the noise. Well, the Hit Tracker model takes it one step further, and actually attempts to correct for this problem using weather information, batted ball speed off the bat, spin off bat, etc. The article I linked to, if you’re into projections, is a must-read.

By the way, my thoughts regarding this were certainly not original. I remember reading an interview of Mark Shapiro where he stated that he thought there was a lot of unexplored work to be done with offensive stats. Surely, they were talking about it in the Spalding Baseball Guides in, like, 1912* : ) The thinking isn’t original, but as far as I know, nobody has actually gathered the data or put the model together, like Greg has. I’m also looking forward to HITf/x, which will apparently debut this year, and could also advance hitting projections (among other things, like fielding analysis). Good time to be a baseball geek!

*And, no, I’m really not kidding. There were some super-advanced articles in those things.

Page 5 of 10« First...«34567»...Last »