No Such Thing as Too Much Pitching, Right?

I  also feel the Twins need to sign 1 top flight starter and a flyer or two.  Things may work out, but squadering the chance is not good.

I still would try to do it on the cheaper side - sign Cobb, and give Tillman a 1-year prove it contract - like 1-6 or 1-7, with perhaps a club option for like 10 the following year. Then I would look to trade for Archer or as has been suggested on here Teheran.

This would give the team a rotation as follows:

Archer/Teheran

Santana

Berrios

Cobb

Gibson

with proven backup options ready to fill gaps/injuries:

​Tillman

Mejia

along with injury recovery options:

Pineda

May

Hughes

Duffey

and finally the rookie crop:

Slagers

Enns

Gonsalves

Romero

Jorge 

et al

I'm assuming a trade for Teheran or Archer would cost at least Gordon and perhaps one of those rookie SPs. Considering how our rotation would shake out for the next few years (with contracts expiring) - I'd do it in a heartbeat. 

I'd be willing to deal Kepler or Kiriloff along with Gordon and a MiLB SP prospect in a deal for Archer - but if we can get Teheran for something a little less I'd be fine with that too. I wouldn't deal Lewis in a deal for either of them.

Is there no longer any thought being given to trading for Chris Archer? Seems the ideal time to plug him in for Santana and then have an even stronger rota when Ervin comes back. The cost will be high, but do any of the Twins pitching prospects project to be better than Archer when they eventually get to the majors? 

Well, the only way to figure out why something doesn't make sense is to do a little digging.The only defense I've ever seen you use is to say something doesn't apply (without much support, and in a flippant way) or just to say that X stat is better than Y stat because that's what Statcast/Steamer or Teams use to determine value. I'm pretty sure that MLB teams are much more sophisticated than Steamer is, but I realize there are probably some really dumb GMs out there.

 

I think that the Steamer gang would probably tell you that they aren't very good, relatively speaking, at accurately predicting an individual player's numbers. I've seen no comparisons of Steamer to the 2016 or 2017 projections that they made other than they were generally better than the other 7 or 8 prognosticators, however they don't give the method they used to rank them. For instance, what percentage of their WAR projections are within 25% of the actual WAR? 

 

A somewhat sideways BAPIP IS considered in FIP. FIP is an attempt to eliminate defensive players from the picture, but it does it by weighing things that may or may not do what it tries to do. To assume that every batted ball is hit in the same way is not really a great way to do it.I freely admit that it would be a very hard thing to do, but because it's hard to do, and SOs are very heavily weighted, it doesn't really reflect what makes a pitcher effective. It would be a better stat if it made an allowance for the deviation from the player's BAPIP to the league BAPIP, because that WOULD give pitchers credit for being more fieldable.

 

Your example of an average from career BAPIP doesn't work because it doesn't look like FIP considers more than the single past year. So, apples to apples, only the single year BAPIP should be used to compare Darvish and Lynn to the projected FIP that was being discussed My points were to show why FIP wasn't a good measure for these two pitchers in this current year. Once again I'd bet Steamer would agree, because so many of Darvish's singles turn into doubles, and doubles to triples because Darvish can't hold people on base and he throws a ton of wild pitches.  

 

One other thing, I think Santana did have a BAPIP that was pretty high as a rookie. How do you suppose his WAR from that year looked compared to the following year's projection?

 

Maybe you could take some time to give us your take on why Darvish's ERA was 15% higher than Lynn's. Dig in to the stats a bit. It might help us understand why Darvish is clearly more productive than Lynn. 

If you're looking for a better analytic than FIP look at SIERA. It takes batted ball profiles into account (i.e. BABIP). It is an ERA emulator but it functions as an ERA predictor even better than FIP does. It too thinks Darvish is a significantly better pitcher than Lynn.

As for wild pitches, Darvish had 12 last season out of a total of 3054 pitches.  Another way to state that once every 3 games he has a wild pitch. Seems likely that has a statistically insignificant impact on his performance.

On an unrelated note I'm excited for the new statcast data to be incorporated into the analytics. I would think knowing average exit velocities and launch angles could greatly enhance an ERA estimator/emulator.

I am always a fan of building from within, but the Twins are truly well into the onset of their window to compete. With the rebuilt bullpen, the two remaining biggest holes for this team is a RH corner OF/1B bat and a true #1 SP. Although we obviously also need at least one more veteran SP as a contingency with question marks surrounding the position. We have the trade capital to land that #1 if an opposing team is in fact shopping one. We need to go on this, and then use our FA dollars to land that veteran lower-rotation veteran arm and that RH bat. We've already missed out on Cole, I hope we make every effort to land Archer because I doubt there are any other teams considering shopping a young controllable top-of-the-rotation type pitcher. 

FIP and BABIP are just stats which can be calculated on a career, season, or even a per game basis. They get more useful/predictive with a larger sample. They are not projections themselves. Projections like Steamer incorporate multiple years of BABIP and FIP (and/or its components).

The prior year's FIP is a projection, and the formula bears that out. I don't know how Steamer uses FIP, but his projection is pretty much Lynn's 2017 FIP with little or no correlation to previous years. I was pointing out that it's not a valid comparison between Darvish and Lynn.  It's not, because it doesn't account for Darvish's SBs and WPs, and their affect on his earned runs is projectable. They will likely cost 7+ runs every year unless he improves dramatically.  The only things that would make Darvish more valuable last year are more subjective than objective.

If you're looking for a better analytic than FIP look at SIERA. It takes batted ball profiles into account (i.e. BABIP). It is an ERA emulator but it functions as an ERA predictor even better than FIP does. It too thinks Darvish is a significantly better pitcher than Lynn.

 

As for wild pitches, Darvish had 12 last season out of a total of 3054 pitches.  Another way to state that once every 3 games he has a wild pitch. Seems likely that has a statistically insignificant impact on his performance.

 

On an unrelated note I'm excited for the new statcast data to be incorporated into the analytics. I would think knowing average exit velocities and launch angles could greatly enhance an ERA estimator/emulator.

Thanks for the info on SIERA, I'll take a look.

Each pitch doesn't tie to a runner advancing.  WPs and SBs do.  In addition to his WPs, he gave up 30% of the Rangers' stolen bases.  Together, they make a significant impact on his ERA.  It could easily be 15%.

I don't know how Steamer uses FIP,

 

[...]

 

It's not, because it doesn't account for Darvish's SBs and WPs, 

These would seem to be mutually exclusive statements. You can't claim to not know the math behind Steamer and simultaneously claim it doesn't take into account SB's and WP's.

Thanks for the info on SIERA, I'll take a look.

 

Each pitch doesn't tie to a runner advancing.  WPs and SBs do.  In addition to his WPs, he gave up 30% of the Rangers' stolen bases.  Together, they make a significant impact on his ERA.  It could easily be 15%.

According to Tom Tango's run expectancy charts Darvish's SB - CS added ~3 runs to his season total last year. Finishing the math that means if Darvish didn't allow any SB's last season (nor any CS) he would have had an ERA of 3.71. When you do the math on WP's (which was ~3 runs as well) it works out to an ERA of 3.69.  Combined they lower his ERA to 3.55.  Or a difference of 8%.

I'm not sure what this has to do with the larger point though.  Darvish > Lynn.

Edit:  I hate not being able to put in 2 spaces after periods.

Just sign Andrew Cashner and Wade Miley, both have decent track records and are former 1st rounders (so they have the ability to be a bargain).

Also, I could see both turning into effective relievers as well if someone (Gonsalves, Romero, Little, May, Duffey) steps up and performs. I especially see Miley (smaller frame) (if that matters?) having the abilty to be that max effort guy in the pen ready to be unleashed.

These would seem to be mutually exclusive statements. You can't claim to not know the math behind Steamer and simultaneously claim it doesn't take into account SB's and WP's.

 

ME:  They're referring to different things.  I DON'T know how steamer uses the FIP calculation in their projections, but I DO know that FIP does not allow for SB or WP.  I hope this helps.

 

 

According to Tom Tango's run expectancy charts Darvish's SB - CS added ~3 runs to his season total last year. Finishing the math that means if Darvish didn't allow any SB's last season (nor any CS) he would have had an ERA of 3.71. When you do the math on WP's (which was ~3 runs as well) it works out to an ERA of 3.69.  Combined they lower his ERA to 3.55.  Or a difference of 8%.

 

ME:  He is determining this value based on the runs scored in a game. The methodology isn't explained, and I wasn't able to find it, but it looks like he's saying that a SB is worth a larger percent of a run in a higher run scoring environment because there would be more bases advanced during the game.  For instance, if you were to steal in a 20 run game, you would be pretty likely to score, thus a SB in a 20 run game might be in the 4s or 5s (but as I think about it, I think it would reach an absolute max value).  His valuation might be applicable here, but he's got a lot of variables to work with.  By their nature this type of valuation generally uses plug numbers to allow iterations.  I don't know where he might have used them.

 

I've never seen a specific average value for SB or WP, but it wouldn't be as complex as his calculations were. I would say that a net 18 SB and 12 WP together would make more runs than that because they'd likely be to 2B, 3B or Home,  But, like I said, I don't know their values,, you could be correct.  

 

 

I'm not sure what this has to do with the larger point though.  Darvish > Lynn.

 

Edit:  I hate not being able to put in 2 spaces after periods.

I hate the single space after too.  I'm also not great at the quotes. sorry

Keep in mind, at its core FIP is just a shortcut to describe a pitcher's K rate, BB rate, and HR rate on an ERA scale. Projecting those rates is what creates the projection. (Although FIP can be derived from the projection too.)

You can see Steamer doesn't like Lynn's K rate or HR rate to bounce back. With a return to normal BABIP, that makes for another high FIP, and Steamer basically projects guys ERA right around their FIP. I think Steamer is maybe overly pessimistic on Lynn, probably because it is agnostic about the TJ surgery/recovery context, but a non-elite K rate guy coming off a career low K rate is a concern. He hasn't yet bounced back fully from TJ in terms of K rate (and velocity?), and there remains the chance that he never will, particularly past the age of 30.

Another projection is ZIPS which looks better for him -- gives his 2017 rates and thus FIP a little boost, and gives him a better ERA than FIP too. Still has him at .289 BABIP, a tick lower than Steamer but much closer to average than his 2017 mark.

https://www.fangraphs.com/blogs/2018-zips-projections-st-louis-cardinals/

In my opinion, Lynn is no slouch, but I still prefer Darvish, who is not without risk of course, but has already come back better/farther from TJ in terms of K rate and velocity, while a contract for Lynn would be a bet on that full recovery which hasn't yet happened. And Darvish has a higher ceiling (peaked at 5.9 bWAR, compared to 3.7 bWAR for Lynn) and probably a higher floor (he's basically been a 4+ bWAR performer, prorated to 31 GS, in each of his 5 MLB seasons so far).

I am sure the Twins have their own criteria that goes beyond any data we have (or could comprehend!) but from my vantage point, Darvish is the higher aim, and I would like them to aim high, given the circumstances.

Phew, I thought this thing was headed south fast.  Thanks for your takes, they're very insightful.  Bill

Yes, statisticians and actuaries have their hands full, how can we understand and apply? The more complex a stat, generally the less useful if becomes. People use them incorrectly, and the weighting of the  calculations are subjective.

I liked Sabermetrics when it discovered that bunting, stealing bases, switch hitting, and having enough LHP  and hitting in the rotation generally hurt more than they helped. Most Managers and GMs still overvalue that stuff.

ME:They're referring to different things.I DON'T know how steamer uses the FIP calculation in their projections, but I DO know that FIP does not allow for SB or WP.I hope this helps.

Thanks for clarifying. It seems you're a bit confused by some of these analytics. I will try and simplify the last 15 years of SABR thought into a few sentences. If you already know this I apologize and you can ignore me. 

As you said FIP only uses a few variables; K's, BB's and HR's primarily, which makes it an easy analytic to use. As such it can be used to estimate how a pitcher performed without taking into account things like the defense behind him, the ballpark he plays in, etc.. that the pitcher has no control over. So for 2017 Darvish struckout 209 batters and walked 58 (which FIP LOVES) but also gave up 27 HR's (Which FIP hates). That makes for an above average FIP and this is how it was originally meant to be used.

However, FIP can also be used as a projection system in two separate ways. First, the 2016 FIP correlates more highly to a players 2017 ERA than a players 2016 ERA to their 2017 ERA. So often a players 2017 FIP is used as a short hand for how a pitcher's ERA might perform the next season. Second, given a projected number of K's, BB's and HR's it can predict a 2018 FIP for players. This is where Steamer comes in. Steamer (and others like ZIPS) uses past trends in year to year fluctuations in a stat, say strikeouts, to predict what a players future number of strikeouts will be.

He is determining this value based on the runs scored in a game. The methodology isn't explained, and I wasn't able to find it, but it looks like he's saying that a SB is worth a larger percent of a run in a higher run scoring environment because there would be more bases advanced during the game.For instance, if you were to steal in a 20 run game, you would be pretty likely to score, thus a SB in a 20 run game might be in the 4s or 5s (but as I think about it, I think it would reach an absolute max value).His valuation might be applicable here, but he's got a lot of variables to work with.By their nature this type of valuation generally uses plug numbers to allow iterations.I don't know where he might have used them.

 

I've never seen a specific average value for SB or WP, but it wouldn't be as complex as his calculations were. I would say that a net 18 SB and 12 WP together would make more runs than that because they'd likely be to 2B, 3B or Home,But, like I said, I don't know their values,, you could be correct.

I didn't look closely enough at his methodology and that was more complex than the one I was looking for, sorry! Disregard my previous mathematical analysis. Though I still think you're overvaluing SB's and WP's. I'll look more closely and repost.

I still would try to do it on the cheaper side - sign Cobb, and give Tillman a 1-year prove it contract - like 1-6 or 1-7, with perhaps a club option for like 10 the following year. Then I would look to trade for Archer or as has been suggested on here Teheran.

 

This would give the team a rotation as follows:

 

Archer/Teheran

Santana

Berrios

Cobb

Gibson

 

with proven backup options ready to fill gaps/injuries:

​Tillman

Mejia

 

along with injury recovery options:

 

Pineda

May

Hughes

Duffey

 

and finally the rookie crop:

 

Slagers

Enns

Gonsalves

Romero

Jorge 

et al

 

I'm assuming a trade for Teheran or Archer would cost at least Gordon and perhaps one of those rookie SPs. Considering how our rotation would shake out for the next few years (with contracts expiring) - I'd do it in a heartbeat. 

NO!!! Please aim high.

Thanks for clarifying. It seems you're a bit confused by some of these analytics. I will try and simplify the last 15 years of SABR thought into a few sentences. If you already know this I apologize and you can ignore me. 

 

As you said FIP only uses a few variables; K's, BB's and HR's primarily, which makes it an easy analytic to use. As such it can be used to estimate how a pitcher performed without taking into account things like the defense behind him, the ballpark he plays in, etc.. that the pitcher has no control over. So for 2017 Darvish struckout 209 batters and walked 58 (which FIP LOVES) but also gave up 27 HR's (Which FIP hates). That makes for an above average FIP and this is how it was originally meant to be used.

 

However, FIP can also be used as a projection system in two separate ways. First, the 2016 FIP correlates more highly to a players 2017 ERA than a players 2016 ERA to their 2017 ERA. So often a players 2017 FIP is used as a short hand for how a pitcher's ERA might perform the next season. Second, given a projected number of K's, BB's and HR's it can predict a 2018 FIP for players. This is where Steamer comes in. Steamer (and others like ZIPS) uses past trends in year to year fluctuations in a stat, say strikeouts, to predict what a players future number of strikeouts will be.

 

I understand what you're saying, but I don't see numbers that support it yet.  Averaged over the only years that I was able to use them, both ERA and FIP averaged being off by roughly .5. I realize that's not a sample, but I'd love to see some actual comparisons overall. How close is projecting ERA and how close is FIP?  The people that use FIP say that it's better, but there should be a basis somewhere to establish it.  It wouldn't be hard to do if someone merged a couple of databases, and if it averages .5 off, it's not that much of a help, in fact it's kind of a hindrance because people don't bother to look at individual stats to see what's actually going on.

 

Please don't take this in any way to demean your valuations of players. It's that I see a lot of people that say one pitcher is crappy, and another is great based solely on WAR, when they really don't know if it's valid for an individual player. WAR definitions generally even use a disclaimer that says players within 1 (or something around there) of each other are about the same.

 

Thanks for the assistance.  

I understand what you're saying, but I don't see numbers that support it yet.Averaged over the only years that I was able to use them, both ERA and FIP averaged being off by roughly .5. I realize that's not a sample, but I'd love to see some actual comparisons overall. How close is projecting ERA and how close is FIP?The people that use FIP say that it's better, but there should be a basis somewhere to establish it.It wouldn't be hard to do if someone merged a couple of databases, and if it averages .5 off, it's not that much of a help, in fact it's kind of a hindrance because people don't bother to look at individual stats to see what's actually going on.

 

Please don't take this in any way to demean your valuations of players. It's that I see a lot of people that say one pitcher is crappy, and another is great based solely on WAR, when they really don't know if it's valid for an individual player. WAR definitions generally even use a disclaimer that says players within 1 (or something around there) of each other are about the same.

 

Thanks for the assistance.

There have been many year to year correlation studies done. Let me be upfront, none of our current analytics do a very good job of predicting future performance. That said they are still the best we currently have. Here is a matrix showing correlations from a few years ago. The best predictors of future ERA are FIP, xFIP, SIERA, tERA. They are more accurate predictors of 2018 ERA's than a players 2017 ERA. 

On a related note, has anyone seen the Steamer/ZIP correlation numbers?

To whom? If you mean us arm-chair GMs trying to pretend to keep up with the pros, perhaps. Not to the ones who designed the stat, though.

Ha ha.Yeah, mainly us. I'm sure the designers like them. If I was them I'd like them too.

There have been many year to year correlation studies done. Let me be upfront, none of our current analytics do a very good job of predicting future performance. That said they are still the best we currently have. Here is a matrix showing correlations from a few years ago. The best predictors of future ERA are FIP, xFIP, SIERA, tERA. They are more accurate predictors of 2018 ERA's than a players 2017 ERA. 

 

On a related note, has anyone seen the Steamer/ZIP correlation numbers?

Thanks for the link!  

Is there a glossary for this chart? I can't remember all of the abbreviations.

Here is a link to Fangraphs' pitching glossary page. You should be able to find most of them there including links for more indepth explanations if needed.