Marcel for the rest of 2008

As expected, Jorge Velandia is set to improve. WOO HOO!!

4 months ago gte619n Comment 32 comments 0 recs |

Story-email Email Printer Print

Comments

Nice to see that Westbrook will somehow make 3 starts the rest of the season.

I did a Google image search for "Andy Marte." It turned up zero results.

by emd2k3 on Jul 21, 2008 10:09 AM EDT 0 recs

AWESOME.

Dear Mr. Sabean, I hear you have a reputation of being stupid. Want to deal Lincecum or Cain? You can pick THREE of these 4 players for either: Borowski, Dellucci, Blake, Byrd.

by westbrook on Jul 21, 2008 4:48 PM EDT up 0 recs

Summary: guys who haven’t played much will massively improve their stats. Barfield’s OPS will show infinite improvement. Guys who have played a little and benefit from SSS (Fasano) will regress to their statistical mean.

Does everyone buy ERA+WHIP as a valid stat?

Finally, I fearlessly project that Jake will have an ERA of 0.00 for the balance of 2008.

by FredOx on Jul 21, 2008 10:32 AM EDT 0 recs

I buy WHIP + ERA when I’m feeling lazy.

I buy K/9, K/BB and xFIP much more when I’ve got the time.

by gte619n on Jul 21, 2008 10:40 AM EDT up 0 recs

Hey, guys: Sal Fasano is an Indian!

by fleerdon on Jul 21, 2008 11:22 AM EDT up 0 recs

And our pal.

Steel Nick

by nickjs21 on Jul 21, 2008 12:31 PM EDT up 0 recs

Stop being needlessly mean to him, the both of ya!

by NickFantana on Jul 21, 2008 3:44 PM EDT up 0 recs

I don’t think I’m very impressed with these projections. I think this is what Chuck means when he charges curve fitting: just assume that a player’s rates will return to a weighted career average and add it all up. Sure, it’s likely they’ll get sort of close in general, but I don’t think this method isn’t producing anything that’s of particular interest, that it’s peering all that deep behind the numbers and coming up with something really noteworthy. The LGT group estimate for Marte in the predictions thread did just about the same thing, right, and we’re only guessing.

Where’s the real surprise in there? Maybe it’s Aubrey. If they hit the mark there, I’ll be impressed.

.882 OPS from Hafner the rest of the way? Improvements for Barfield and Velandia? Negative regressions for Peralta and Perez. Hmm. I dunno. I’m skeptical.

by jhon on Jul 21, 2008 12:54 PM EDT 0 recs

Not exactly going out on a limb predicting improvement for Barfield, since going 2-for-4 on the return would raise his average by 200 points. Aubrey has been trending in the right direction at Buffalo – over his last 10 games, he’s hitting .375 with two doubles, a homer and 9 RBI. For the month, he’s posted a nice .781 OPS, down a bit from May but a significant improvement over June. I also like that he’s not striking out much, with 28 in 260 minor league ABs.

by FredOx on Jul 21, 2008 1:36 PM EDT up 0 recs

a nice 781 ops? he’s 26. 781 ops in AAA are not that nice, especially as his “improvement” over June

by Gradyforpresident on Jul 21, 2008 2:37 PM EDT up 0 recs

agreed. Aubrey’s done a good job of maintaining a reasonably high batting average, but his injuries seem to have collectively drained his power. theoretically, his power should have been increasing over the past several years. Instead, his ISO numbers have continually gone down after posting a +.200 mark for his first 1.5 seasons. This season that number is down to about .120, making his worth almost nothing. I’m not opposed to swapping him with Garko for a month just to see what happens. But nobody should expect anything of great value from Aubrey.

by APV on Jul 21, 2008 2:44 PM EDT up 0 recs

Poor choice of words – all I meant was he hasn’t sucked as much lately, and the marginal improvement suggested above is therefore not unreasonable. I don’t like either Garko or Aubrey as a long term solution.

by FredOx on Jul 21, 2008 3:01 PM EDT up 0 recs

long-term solution at 1B = Beau Mills at the moment. Possibly Victor Martinez till that takes hold.

by APV on Jul 21, 2008 3:06 PM EDT up 0 recs

Yeah, that’s about right.

One of my prof’s once told me that, on average, you can’t drown within 50 feet of shore in Lake Erie, cuz, on average, it’s only four foot deep.

Resident LGT beer kinda sewer

by mauichuck on Jul 22, 2008 5:00 PM EDT up 0 recs

For reference, here’s a little introduction on what how the Marcels are calculated and what they do and do not represent.

by APV on Jul 21, 2008 1:23 PM EDT 0 recs

That clears things up. Thanks.

by jhon on Jul 21, 2008 1:43 PM EDT up 0 recs

Key point, direct from the caretaker of the monkey:

So, what to do? Trust the forecast for a group of players, but don’t go betting on any one forecast. There’s not a single person in the whole world who can help you there. There’s no book, there’s no program, there’s nothing to help you with any single forecast. That’s why we play the darn game, and that’s why we love its drama.

by FredOx on Jul 21, 2008 1:51 PM EDT up 0 recs

exactly. and as a “group of players” the Indians should be better in the 2nd half than they were in the first half because they were disproportionately worse than expected in the first half.

by APV on Jul 21, 2008 1:57 PM EDT up 0 recs

Speaking for myself, it’s a lot harder to buy into the “progression to the mean” stuff after we spent all offseason talking about how we were under-projected for 2008. We ended up being WORSE than our projections and now we think the team will perform up to the same projections we were dismissing as being overly conservative before the season.

Although it lacks rational thought, my current attitude towards the Tribe is to be underwhelmed and skeptical until I see significant improvement from every facet of the organization (including front office and training staff, which is actually most depressing.)

by NickFantana on Jul 21, 2008 3:49 PM EDT up 0 recs

I don’t entirely agree with the notion that we’ll now “perform up to the same projections”. The most likely outcome now is that we regress towards a new mean, which is probably lower than the pre-season projection because of this year’s performance. Take Garko as an example. I thought coming into this season he’d put up a big-year as he’s entering his “prime years”. He obviously has dramatically under-performed those expectations. This does mean that he’s still likely to show some improvement (he’s better than he has been), but maybe not as much as you might have expected early in the season. His poor performance might indicate his true talent level isn’t what it looked like coming into the season, bringing down expectations for him somewhat. So for Garko’s case I do think he’s better than he’s playing right now, but I don’t think he’s as good as he looked coming into the season (which wasn’t that bad, but actually not that great either).

by APV on Jul 21, 2008 3:55 PM EDT up 0 recs

As Ryan hinted in his latest Recap, Garko is fast approaching one full year of sub-700 OPS.

I agree with your basic concept here, though.

by Jay on Jul 21, 2008 7:33 PM EDT up 0 recs

Right. When you’re projections are off just re-calculate your target. This way you’re never wrong.

Lee’s good – he won 18 games – wait no he’s not, look at his peripherals – now he’s bad – told you so, his peripherals stink – now he’s good again – it won’t last, look at his peripherals – wait he’s is really good – that’s because he’s changed what kind of pitcher he is – etc. etc. If you project long enough – like until Cliff’s 45 years old – he’ll be bad again and finally, finally you’re right.

Resident LGT beer kinda sewer

by mauichuck on Jul 22, 2008 4:58 PM EDT up 0 recs

ok ok….before I get off your lawn I’ll put up one more response. And Chuck, I know you know all this, but…

A projection system like Marcel is intended to give you the most likely outcome (projection), given a certain model. As it turns out, the model Marcel is based on is a pretty simple one and includes some basic input data that has shown to have strong historical significance in the performance of baseball players. What the model obviously doesn’t include is personal observations of the sort, “wow, Cliff Lee really learned how to pitch”. It takes some consideration of the fact that Lee has shown evidence of real improvement in his control, but it does so without completely forgetting Lee’s control problems of the past.

My intention in posting the link to the Marcel’s reference page was to avoid one of these inane conversations about what a projection model like Marcel (or PECOTA, or CHONE, or ZIPS….) is and is not intended to say. Misinterpreting the intention of a model is not a problem with the model but a problem with the interpreter. Anyone who has watched this season has seen that Cliff Lee is actually a much better pitcher than his performances over any previous portion of his career would indicate. Given this, you would expect him to continue to outperform projections for him. The degree to which you would expect this will depend on how sustainability his improvements are, a subjective scouting criteria.

by APV on Jul 22, 2008 5:14 PM EDT up 0 recs

All I’m trying to say is this: just because it has the trappings of a science – you know numbers and math and stuff – doesn’t make it a science. It’s not. And just because you use something that looks like scientific method doesn’t make it science either.

Here’s a little pearl from one of my engineering profs. “Never apply logic to an illogical system”.

Resident LGT beer kinda sewer

by mauichuck on Jul 22, 2008 8:01 PM EDT up 0 recs

of course it’s not science. it’s attempting to outline possible scenarios for the future and their likelihood, but doing so in a systematic way so you can easily interpret them. Who said it was science? No hypotheses are being presented and rejected.

by APV on Jul 23, 2008 7:58 AM EDT up 0 recs

Well I fail to see where Mr. Marcel acknowledges the art in these “projections”.

Here’s his hypotheses: this is the likelihood of a few scenarios. A%, B%, C% Now does anyone go back, look at all of his projections and determine the accuracy of his method – you know like scientists would? If they have I haven’t seen it.

Much of this stuff looks a lot like elections projections to me. That’s a well honed system. The pollsters make projections based on a pre-determined sample and then re-calibrate their methodology depending on the outcome. I would imagine all of the “sabermaticians” doing the same. You weigh various factors, make a projection, determine how close you are and your “batting average” and then recalibrate. Again, almost like scientists. It’s really not that hard. But most of these guys, kinda like Miss Cleo, rarely acknowledge their misses.

I guess what sticks in my craw is the almost slavish devotion to the stats themselves. There’s a hell of a lot to this projection business other than past performance.

Resident LGT beer kinda sewer

by mauichuck on Jul 23, 2008 8:57 AM EDT up 0 recs

I lean towards the more strict Popperian view of science, so I’m not always a fan of comparing between likelihoods or probabilities. But as long as you can deconstruct how those probabilities or likelihoods were derived, it’s hard to say the model is wrong. It’s a model. You can assess the accuracy of that model against known results. Marcel is the simplest of these projection systems (hence it’s name), which makes it the easiest to deconstruct. And like it says, it’s not designed to give you a perfect picture of the future for every single player individually. But collectively, it does a pretty decent job when you look across a broad sample of players.

There is no perfect way to conduct futurology.

by APV on Jul 23, 2008 9:18 AM EDT up 0 recs

See, now if Darwin came out with his projections and stated that some sea-faring creature would in fact “break-out” and become amphibious based on its tendencies, chuck would have blown that theory “out of the water” because no “eveidence” existed.

by talonk on Jul 23, 2008 9:47 AM EDT up 0 recs

But most of these guys, kinda like Miss Cleo, rarely acknowledge their misses.

Right, but most of them are usually pretty darn close in their respective predictions. Neither system provides a be-all-end-all prediction, but both provide reasonable expectations.

Burn on, big river, burn on...

by Turkmenbashi on Jul 23, 2008 4:37 PM EDT up 0 recs

and I actually think there have been a whole bunch of articles which have gone back and compared the various projection systems at the end of the season. you can’t evaluate where a system like PECOTA messed up, because its details are proprietary, but system’s like Marcel (which is really just meant as a baseline) and CHONE are open.

by APV on Jul 23, 2008 4:58 PM EDT up 0 recs