[Business Week] Extreme Moneyball: The Houston Astros Go All In on Data Analysis

Yeah, it's not like OK pitchers ever have great stretches or anything. Bartolo Colon last year with his 2.65 ERA proved he is a star. RA Dickey, Cy-Young winner, was clearly a player the Mets should have held on to. It's not like the peripherals pointed to these guys regressing and that they regressed or anything... Let's just look at a month stretch and evaluate out based on that. The perfect "buy high" strategy.

 

Cosart also did implode - just a month ago to a tune of an ERA over 8 in July...

 

Young pitchers do that sometimes. Even good ones. Why is that sample size over one month more instructive than his last 4 starts where he's given up a total of 2 runs? Or the sample size from last season where he finished with a terrific ERA?

 

I'm sorry, are you really comparing a 24 year old pitcher who is in his first full season (after having a very promising debut, with room for improvement) to a former phenom who turned into a roider, and a journeyman who developed a knuckle ball because he had no UCL (and happened to win a cy young)?

And you're back with your analysis that this is the best Cosart will ever be... At 24... When in fact he's actually been even better as a Marlin (so maybe the Astros didn't sell him at his "highest").

You're also saying "one good month", when Cosart put together a run of consecutive quality pitching from Late April till July (12 starts) earlier this year, along with his great debut last year.

Wait, I know... His "peripherals" sucked. Well, now they don't and he's dominating. Young pitchers in their first full seasons tend to make adjustments like that.

 

Absolutely - it's not. His season ERA this year is around 4. People keep pointing to one month periods where he's good and saying things like "wait for him to implode" - so I'm pointing out there are months where he's imploded.

Again, he's had a good ERA for his career - but ERA is a terrible predictor of ERA going forward. His other stats have been dismal even when his ERA was good - outside of a decent June and now a great August, which coincided perfectly with a new pitching coach.

No - I'm pointing out what happens when you look at ERA's over small sample sizes as your predictor of future performance. As I've said numerous times, stats show that ERA is a poor predictor of future performance. So when people keep using ERA exclusively to predict Cosart's future performance, I point examples of the problem there.

And again, you rely on the stat that I keep telling you is a terrible predictor of future performance to get your great stretch - and one that I guarantee you the Astros don't use much. You can keep talking about it, I'll keep ignoring it, and we'll just keep talking past each other. I'm explaining what the Astros are doing and why. You can accept it or not.

I also said maybe the Astros will be wrong. It's not an "adjustment" to the majors - he's now doing things he's never done in his MLB or MiLB career. I suspect the change to Miami was helpful with a new pitching coach and new look at his approach - similar to what the Astros saw in "fixing" McHugh. If he can sustain it, he has a chance to be an outstanding pitcher. The Astros use the data they have to make the best decision at the time. They see failing pitchers they think they can fix like Colin McHugh and buy them low. They see succeeding pitchers that they think won't sustain it and sell them high. Get used to it - that's the entire concept of everything they do and it's going to happen a lot more. This is the "fully commit" part of their analytical strategy. It's often going to involve trading away players that are having good years; it's often going to involve picking up players that seem crappy. They'll be right sometimes. They'll be wrong sometimes. If their system works, they'll be right more often than they are wrong.

 

He's "never" done? Again... this is his first full season. Being close minded to the point that players in their first full seasons can NEVER improve will not be a successful long-term strategy for any player/position.

The Astros themselves have enough examples of pitchers who take a little longer to "get it" and eventually outperform what they did in the minors or in the majors thus far (Keuchel, Oberholtzer, and your recent example of McHugh).

I have no problem with buying low and selling high... provided that it clearly makes the MLB team better (for which it is still a huge unknown whether or not this trade did that). I'll also question any system that says they have "too much" young quality starting pitching... and I doubt even Luhnow would agree with that sentiment, as he just came off a draft where he was trying to acquire 3 HS starting pitchers for his system.

My biggest fear is that this "system" could work exactly as it should (one which seems to overvalue prospects/unknowns vs. established MLB players), and the big league team never escapes mediocrity or reaches actual MLB success before the next wave of big league players are traded for prospects.

 

When you walk people your whole minor league career and continue to do it in your major league career, correcting that is not an example of "adjusting" to the major leagues. It is an example of doing something you have not done before. If he didn't walk people in the minors and started doing it when he came up because the hitters he was facing were better, then fixing that is an adjustment. If he's able to drop his walks going forward, that's more of an example of a change of in approach or a step forward in his development as a whole. And yes, it's something he's never done in his entire professional career; no idea if he walked lots of people in high school or not.

Absolutely. They are looking beyond the current results and determining whether they think a player can be good or not - that's the whole point of the statistical system.

That's never going to happen, though. If you're buying low and selling high, you're by definition trading someone who is performing well for someone who is not or is filled with potential. Buying low and selling high almost always will result in an immediate worsening of your current team with the intent to improve it in the long run.

It doesn't overvalue prospects or undervalue established MLB players. Take Carter and Keuchel and McHugh as an example of MLB players it valued and held onto - in Carter's case, for much longer than fans were willing to wait. Based on that, you can see assume they see something different in Cosart and are not just trading all their MLB-level talent all the time.

Cosart is basically the definition of what the Astros' methodology is about - if you're opposed to the trade because he's doing well, that's fine - but you're really essentially saying you're opposed to the advanced-analytics system they are using (which is also fine). The point of what the Astros do is to find value where other people don't see it, and find where other people see value where the Astros don't see it. The pro-Cosart arguments boil down to 3 things: he's young and thus can improve; he has good "stuff"; he's had success thus far in terms of ERA. These are all great and all true - but they are exactly the things that the "non-analytics" people have looked at for decades. Those are the people that the Astros target as trade partners - the people that will value Cosart highly whereas their metrics value him less. The Astros' goal is to find the inefficiencies in what other people value - to find where they can project less success and thus ID trade-high candidates. Keeping a guy like Cosart when the underlying stats don't project as much success is the definition of the "old school philosophy" of focusing on what your eyes see rather than what the data tells you. And that may be the right way to go - but the Astros are fully committed to the data-driven methods, so this is a part of going "all in" on that strategy.

 

I feel you've gone off the deep end and clearly over-valued Cosart's "mediocrity" as being the sole reason for this trade. I stand by the fact that getting Moran was the driving force. And each additional quality start that Cosart churns out makes Moran's "expectations" go up as well.

My "fear" of the system is that if there's another highly drafted prospect available next year, they will likely continue to part with the "we know who they are now" major leaguers like Kuechel, Carter, Folty, or McHugh... after all, chances are this is as good as Kuechel and McHugh are going to be. You could make a Cosart-like trade every single year if you wanted. And while analytically, the trade could be the "right" one... I'm just going to have a hard time rooting for a team that can't pass up on replenishing the farm over and over again.

Also, I was opposed to the trade because I thought Cosart could improve... now that he's improved, I'm supposedly "against everything the Astros stand for" which is really not the case.... just that I'm seeing the Astros "system" to have flaws when they can't foresee improvement from 24 year old pitchers in their first big league season.

 

The guy has pitched 209 innings in the Major Leagues now. He has a 3.27 ERA over that span. At some point that has to mean something. He's 24 years old and was under club control for the next 4 seasons.

I get that maybe ERA now doesn't extrapolate to ERA later...but his ERA over his time in the minors was good too. And of course none of that assumes any sense of improvement from experience or from coaching. Personally, that's not the kind of guy I'd be trading away...from my view, you can never have enough pitching, particularly when it's good, young and cheap. I don't assume for one second that all the pitching talent we hopefully have at the minor league level will extrapolate to being able to come up to the major league level and maintain a 3.27 ERA over 209 innings.

 

And I'm saying your fear is unfounded and comes from not understanding what the Astros do and how they make decisions.

Then you're going to struggle with the Astros for Lunhow's tenure, unfortunately.

Well then you should stop pointing to his ERA as evidence of how good he is. If you're opposed them actually making decisions based on the advanced analytics they are seeing in Cosart, yeah, you're essentially opposed to the system. You're asking them to use surface analysis instead of their stats methodology.

Except that's total nonsense. They foresaw improvement from other 24 yr old pitchers and held onto them even when they sucked - this is the part you seem incapable of seeing. You are under the mistaken impression that because they don't see something you do in one specific player, it must apply to ALL players. The whole point of the Astros system is to see what you're not seeing - to see the difference between a bunch of 24 yr old guys that might look similar on the surface.

 

I agree with you - but the Astros likely don't expect Cosart's talent to extrapoloate to a 3.27 ERA over the next 209 innings either. This is where you have to decide whether to trust the analytics or not. It's easy when the results match what the stats predict. The hard decisions are when the results are better or worse than the analytics suggest. That's when you decide to stick with a Keuchel and his 5+ ERA or a Carter and his sub-0.200 BA or take a chance on a McHugh. Or trade a way a high-performing player like Cosart. If you're not willing to make those decisions, then you're choosing your eyes over the the analytics - which is fine, but that's not the Astros' front office approach.

 

I appreciate the way you've defended the front office... and I understand your admiration as it definitely gives hope to internet nerds everywhere that they too could run a baseball team someday... however, I'm not inclined to think they're "invincible" in every single move/decision they make and this very well could be an example of the Astros giving up early on a player who will go on to have a great career (better than the unknown careers of the prospects they got back).

All teams have trades or make moves where they come out on the wrong end... some happen immediately, some take time. Even the Astros ultra intricate "system" that you so admirably defended (and attribute every "hit" to) is capable of these mistakes.

But, as you've already admitted that they "could be wrong"... we're just going to have to sit back and watch. We'll have to see if with each promising start Cosart has, and each article that gets written about why he's better now, and why the Astros were foolish to let him go (and the expectations for Moran grows... already exceeding what it was when he was acquired), will the Astros practice pattern change as a result of "mistakes"?

I'd say there's already been a few mistakes this year attributable to the "system" that they not only corrected, but likely will impact how they do things in the future:
1. They essentially admitted sending Appel to A ball was a mistake (as most of us here suggested)... if they had to do it over again, he likely gets the full amount of time to train (that he would have had if not for the appendectomy), and starts off at AA. I know they don't like demotions or skipping levels, but that's essentially what ends up happening when they got him out of A ball for a change of scene (and not performance).

2. They admitted that they were burned by the "unknowns" when it comes to drafting high school pitchers in the draft (in this case, an "injury" concern that if really something of substance, would have likely impacted him at some point in college).... and they plan on correcting it with their supplemental pick next year ("We expect to get a player that will likely be on the big league team prior to when Aiken was going to...")

Finally, i'd appreciate if you back-off the elitist viewpoint as in pertains to this front office...mainly because you don't work for the Astros, you don't know why they really make the decisions they do... and when you say things like "You just don't understand why they do things,"... or "you're going to struggle with being a fan of this team..." it comes off as not only arrogant but simply delusional from the standpoint that you may actually believe that you work for the Astros or really do have a full insight on to why they do things (when you don't).

Fans of this team (whatever fans they have left) can, will, and should continue to question moves they don't agree with... it doesn't mean they don't "understand" the front office, or simply don't "trust" the analytic approach. I fully understand why Luhnow made the trade, I support the full analytical approach... and I still can have an issue with the move.

I'm also not going to be naive enough to believe that every single "right" decision they make was based on endless amounts of data/research, and every single "wrong" decision they make wasn't really "wrong". From a sell high perspective, I though they should have gone all out to trade Castro last year (and the "leaks" proved they were trying to do that), but they held on to him because the right trade wasn't there (just like the right trade may not have been there for Kuechel at this year's deadline). It doesn't mean they hang on to these guys because the deep analytical theorems says there's something there that makes them untradeable.

 

Here's the problem - you say that, but I don't believe you (and frankly, most others) truly do when push comes to shove. Let's step back for a second. We can safely assume the following:

1. The Astros system is designed to predict, as best as it can, the future performance of players through the use of advanced analytics. Whether it does a good job of that, we have no idea, but that's the purpose of it.

2. The Astros' goal is to identify and keep/develop the best players they can find. Cost is also relevant in a general sense, but not here since everyone involved is under club control and not getting paid for a long time.

3. A 24-yr old pitcher projected with a 3ish ERA is about as good as you can reasonably expect to find.

From there, we can safely deduce that the Astros' black box believes that Cosart is likely to regress. If not, there would be no reason to make the trade because any prospect you get - unless you're getting a phenom like Harper or Strasburg - is not going to project to better than the above stats and come with more risk due to being a prospect.

So here's a simple question:

If you have a pitcher that you see is young and doing well, but the data says he shouldn't be and isn't likely to sustain it, which do you believe?

If you believe the data, then it means you believe he's overperforming and the right organizational move is to trade high. If you believe the actual performance, then you invest in the guy and hold onto him. But, by definition, the latter choice is saying you don't actually trust the system. You trust it when your eyes and the data matchup, but when it comes down to having to trust the data against what your eyes see or your own analysis tells you, you're picking against the data.

Again, there is absolutely nothing wrong with that, and I would suggest at least 25 teams in MLB would go that route. Maybe even 29, because the Astros' front office is at the far extreme in terms of their focus on analytics. I'm not even arguing that the front office is right - this is a big grand experiment that has never been done before and we have no idea what the results will be. What I am saying, though, is that the Astros are committed to this experiment. And my opinion is that they have to see it through. You can't say you're going to use this analytic system but then go soft on it when a tough decision has to be made because then you never really are fully implementing the system. You're really just using it to confirm your own biases instead of actually trusting it.

FWIW, the same human psychology applies in the stock market and in all sorts of other real world things - everyone wants to buy low and sell high, but when their stock that they bought low goes upa bunch, it's much harder to sell - because there's always a belief that you picked right and it can get better. So an emotional stock trader often gets the "sell high" part wrong. Gamblers have the same issue when deciding when to walk away from a craps table or whatnot. There is a natural human tendency to see something at its peak and overvalue it and see something at its trough and undervalue it. That's one reason people always say you have to take the emotion out of the stock market and try to stops and limits and the like to make your decision making more "objective". Or for a gambler to set absolute standards for when they walk away (if you double my money, you leave the table and take a break no matter what, etc). The Astros are attempting the sports version of the same phenomenon.

 

Does everybody remember J.A. Happ?

Over his first 29 starts, 40 games overall, he had an ERA of 3.27. All of his peripherals were bad, but his ERA was good. Lo and behold, his ERA and his peripherals eventually met.

Could Cosart improve his peripherals, absolutely. But it's the Astros job to project how good players will be long term, and the Cosart they traded didn't project extremely well.

 

The problem with that question is that you make the assumption that the young pitcher will be fixed in his ability to improve/change over time (i.e. - "He is who he is... at 24"), along with you cherry picking the stats/data to say he shouldn't be good.

Did I believe Cosart (as a rookie) wasn't as good as his ERA said last season, yes. Do I believe he needed to improve his control and work in his curveball more to continue to reach his potential, yes. Do I believe he "could" improve/make adjustments to get to that point, yes. The last point seems to be where we're at an impasse... you believe he wasn't going to change based on his minor league stats and major league stats, I believe that all 24 year old's (position players as well as pitchers) can eventually "get it". Based on how he's looked over the last month, there's a good chance Cosart has made that adjustment that you thought he was incapable of doing (and don't worry... the peripheral stats now support him doing what he's supposed to be doing).

Additionally, look at other factors... is his arm sound? (by all accounts, no serious arm trouble, and he's been pitching pro ball since 2008) Do his mechanics predict possible future injury? Does he possess a solid work ethic? What's his attitude like? (immature from what some said, then again he's also got a pretty big competitive streak). Again... all things that aren't going to be spit out by a computer that I'm fully confident this team still takes into account.

Lastly, stop speaking for the Astros... you're not a member of their front office. This is basically your opinion vs. mine. You're not privy to even a fraction of the decision making that went into this deal, or what this front office does on a daily basis... despite one or two articles/year that discuss it. Also, despite what you "believe", I do fully support the front office and their approach... but I won't do it blindly (or be a sap) and just go along or agree with every move they make.

I understand its an experiment... just like its an experiment to see how much losing fans will continue to endure at the big league level to the point that even just simply looking respectable may not bring them all back (and again, more than 2/3'ds of this city couldn't care less about the Astros right now).

 

Quit looking at simply peripherals and ERA... what about his age? What about his stuff? What does he need to improve? Does he have the work ethic and mental makeup to do that?

Because right now, peripherals and a low ERA at debut are the only things that Happ and Cosart have in common.... a better comparison between soft tossing lefties with good numbers would probably be Happ and Keuchel.

And even then, a guy like Happ at 27 gets you Roy Oswalt for a pennant run? Much rather be in a position to make those sorts of trades in a year or two.

 

Yeah, come on guys. Only post things Nick will agree with.

 

And to add to my previous post, another "mistake" the front office admits to (and rectifies) as they fire Bo Porter.

Again, they're going to make mistakes during this experiment... and likely change the way they do things multiple times, throughout the process.

 

Or just agree with everything the team does at all times... they're always right, even when they're wrong.

 

Some of us would disagree on how good Cosart's stuff is. Merely throwing 96 MPH does not equate great stuff. What i see is a pitcher that lacks an out pitch, right now anyway.