Article: BREAKING: Twins to Sign Versatile Marwin Gonzalez

[Otto von Ballpark]

 

I have to wonder if this is a place where prediction models fail. They account for playing time but I question how accurately they predict subpar replacement players. 

 

For example, Gonzalez is likely a step down from Polanco. But if Polanco gets injured, how much of a step up is Gonzalez from Replacement Player X?

I was just thinking, it might be instructive to look at Brock's question in terms of luck. (Pardon the long post here, and don't take it as some long diatribe against Gonzalez - it's really just exploring the nature of projections and things!)
 

The projection I shared from Fangraphs is a median projection, assuming no luck or neutral luck. (An aside: the projection is not *predicting* that there will be no luck or neutral luck -- it's simply attempting to quantify/isolate an estimate of the luck and non-luck involved between players and teams.)
 

Anyway, that median projection of +1 win with Gonzalez assumes no luck, or neutral luck. And Brock's question (correct me if I am wrong) is asking whether it properly captures Gonzalez's value in a bad-luck scenario, like an extended injury or performance collapse, that normally involves replacement players.

 

Gonzalez can definitely help as a replacement in some bad luck situations, especially with his versatility. But not in all bad luck situations -- most obviously, bad luck on the pitching staff won't have too much to do with Gonzalez (beyond what he contributes with his teammates on defense). Bad luck at catcher is pretty unrelated to Gonzalez too (beyond maybe the Astudillo utility factor, if we'd really use him in these roles). Even with bad luck at other positions, he doesn't help equally across all positions -- an injury to Buxton in CF is only covered by Gonzalez indirectly, by pushing a corner player to CF. And even where he could cover them directly, like at 2B/SS/3B, there is still some uncertainty in Gonzalez's own performance record about how he might perform in truly extended action at those positions. This is no knock on Gonzalez, it's really just a reflection of limitations on any one player's effect on a team.

 

All of this is to say: the more bad luck we have, the more likely we are to get a bad luck situation that Gonzalez can best address. But the more bad luck we have, we're also moving further away from the median and toward the floor on a spectrum of projection outcomes. We can all come up with scenarios where Gonzalez successfully mitigates some isolated patch of bad luck while the rest of the team enjoys good luck near our ceiling projection -- but definitionally, that's going to be a fairly rare combination of events. It shouldn't move a projection needle much.

 

So to me, anyway, it makes some intuitive sense that Gonzalez, projected as a 1.5 WAR player, might have value at mitigating bad-luck situations that could be represented as +1 win at our median projection. And that value could be higher (+1.5 wins?) as we move towards our projection floor, and lower (+0.5 wins?) as we move toward our ceiling. (Another way to look at it would be percentage odds of each outcome. I think adding 1 win to the median projection, bumping it up to 83 wins, can also be interpreted as increasing our odds of winning 82, increasing our odds further of winning at least 75 or so, but probably giving a smaller increase in our odds of winning as many as 90.)

 

Of course, none of this is gospel, it's just Fangraphs estimate. Projections based on B-Ref's value estimates might give Gonzalez a +0.5 projected WAR boost for 2019. Plenty of folks probably disagree with the 82 win pre-Gonzalez Twins baseline projection too. But relative to that, I hope I've laid out how +1 median win is probably fairly encapsulating Gonzalez's replacement value.

[KirbyDome89]

 

Eh, I just think that term is loaded with meaning for people.  And it just doesn't apply to Gonzalez.

 

The ceiling/floor argument doesn't do much for me.  I want Kimbrel signed, but I'm not sure he's going to radically increase our ceiling either.  Not for wins any way.

No, likely not, but he would fill a much greater need. I think it's more likely that the bullpen will cost this team wins than the 10th man on the bench, so that's where I'm coming from with tamping down a bit of the hype surrounding the Gonzalez signing.

 

Look, the Twins upgraded at a position. I'll always be happy about that and I'd love to see similar moves in the future. My only point was that the upgrade came at a position that was further down the list of needs, and so the gap between the production they could've expected vs. the production they can expect moving forward very likely doesn't move them to a position where they're .500 at worst. 

[yarnivek1972]

It does increase the win projection -- just like it did at Fangraphs. It's just that it increases it more towards the floor (where Brock's scenario of needing replacements is more common) than it does at the median or towards the ceiling.

 

 

This is a separate issue, and not really a reflection of reality either. The Twins weren't playing Morrison because they had to roster Motter. They played Morrison because they were invested in him and felt he was the best bet; they rostered Motter as a reserve because they felt they had the space available to do so. There's not a competent front office in the world that would keep starting Morrison if they felt they had better options, just to keep Taylor Motter on the bench.

 

I know Vargas was pretty bad in AAA last year but his MLB track record is still pretty decent. As poorly as Morrison played last year it still makes me wonder what might have happened if Vargas had come up.

[Brock Beauchamp]

 

I was just thinking, it might be instructive to look at Brock's question in terms of luck. (Pardon the long post here, and don't take it as some long diatribe against Gonzalez - it's really just exploring the nature of projections and things!)
 

The projection I shared from Fangraphs is a median projection, assuming no luck or neutral luck. (An aside: the projection is not *predicting* that there will be no luck or neutral luck -- it's simply attempting to quantify/isolate an estimate of the luck and non-luck involved between players and teams.)
 

Anyway, that median projection of +1 win with Gonzalez assumes no luck, or neutral luck. And Brock's question (correct me if I am wrong) is asking whether it properly captures Gonzalez's value in a bad-luck scenario, like an extended injury or performance collapse, that normally involves replacement players.

 

Gonzalez can definitely help as a replacement in some bad luck situations, especially with his versatility. But not in all bad luck situations -- most obviously, bad luck on the pitching staff won't have too much to do with Gonzalez (beyond what he contributes with his teammates on defense). Bad luck at catcher is pretty unrelated to Gonzalez too (beyond maybe the Astudillo utility factor, if we'd really use him in these roles). Even with bad luck at other positions, he doesn't help equally across all positions -- an injury to Buxton in CF is only covered by Gonzalez indirectly, by pushing a corner player to CF. And even where he could cover them directly, like at 2B/SS/3B, there is still some uncertainty in Gonzalez's own performance record about how he might perform in truly extended action at those positions. This is no knock on Gonzalez, it's really just a reflection of limitations on any one player's effect on a team.

 

All of this is to say: the more bad luck we have, the more likely we are to get a bad luck situation that Gonzalez can best address. But the more bad luck we have, we're also moving further away from the median and toward the floor on a spectrum of projection outcomes. We can all come up with scenarios where Gonzalez successfully mitigates some isolated patch of bad luck while the rest of the team enjoys good luck near our ceiling projection -- but definitionally, that's going to be a fairly rare combination of events. It shouldn't move a projection needle much.

 

So to me, anyway, it makes some intuitive sense that Gonzalez, projected as a 1.5 WAR player, might have value at mitigating bad-luck situations that could be represented as +1 win at our median projection. And that value could be higher (+1.5 wins?) as we move towards our projection floor, and lower (+0.5 wins?) as we move toward our ceiling. (Another way to look at it would be percentage odds of each outcome. I think adding 1 win to the median projection, bumping it up to 83 wins, can also be interpreted as increasing our odds of winning 82, increasing our odds further of winning at least 75 or so, but probably giving a smaller increase in our odds of winning as many as 90.)

 

Of course, none of this is gospel, it's just Fangraphs estimate. Projections based on B-Ref's value estimates might give Gonzalez a +0.5 projected WAR boost for 2019. Plenty of folks probably disagree with the 82 win pre-Gonzalez Twins baseline projection too. But relative to that, I hope I've laid out how +1 median win is probably fairly encapsulating Gonzalez's replacement value.

I mostly agree and see where you're coming from but disagree on "bad luck". I consider it an inevitability in today's game.

 

Someone will get injured. Gonzalez will take their place. Where Gonzalez separates himself from someone like, say, Adrianza, is that he can play anywhere. In your scenario, I'd probably agree that it requires a fair amount of "bad luck" to get Adrianza 400 PAs, as it means your middle infield was torn to shreds over the season.

 

But in the case of Gonzalez, it could come in small bursts all over the diamond: 40 PAs in left, 60 PAs at short, 80 PAs at third, etc. Because of Gonzalez' versatility, I don't believe it requires bad luck for him to get a lot of playing time all over the field, all the while replacing someone at neutral or negative WAR value. Starters will get dinged up over the season and having a guy already on the roster to fill in for them can lead to a lot of PAs, all the while keeping your roster more stable in the process and not risking some pretty awful performances from replacement players, even if their scope of damage is limited to only a few plate appearances.

[diehardtwinsfan]

not to make a silly point here, but the replacement player is of itself a somewhat mythical creature. It's a construct. It's not like our AAA farm is loaded with them. Guys will either out perform the replacement player or underperform. Very few will actually be a 0 WAR guy. 

 

Gonzalez, as I see it, allows us to not give a ton of at bats to guys who are much closer to replaceent player, or even below it. I like that. 

[Otto von Ballpark]

 

I mostly agree and see where you're coming from but disagree on "bad luck". I consider it an inevitability in today's game.

 

First of all, I don't know if you were implying this here, but the median projection isn't assuming no "bad luck", particularly at the team level. It's really assuming neutral luck -- part of which will be the inevitable bad luck as you note, but part of which should be offset by some similarly inevitable good luck too. Projections aren't dismissing the effects of bad luck, or under-rating them, but rather saying the net effects of luck -- good and bad -- are a bit unpredictable, and it's worth isolating that net effect to better estimate other factors of player and team performance.

 

You and I and others are correct in noting that Gonzalez, as a versatile, competent replacement helps mitigate the effects of bad luck. Others have noted that Gonzalez gives us a slightly better chance at another 4 WAR player too on the good luck side. Both of these factors effectively increase our neutral luck, like the projection shows with +1 win. But as I understand, you were skeptical of whether the projection reflects the benefit enough. I think it does, and hopefully I can explain better below.

 

Someone will get injured. Gonzalez will take their place. Where Gonzalez separates himself from someone like, say, Adrianza, is that he can play anywhere. In your scenario, I'd probably agree that it requires a fair amount of "bad luck" to get Adrianza 400 PAs, as it means your middle infield was torn to shreds over the season.

 

But in the case of Gonzalez, it could come in small bursts all over the diamond: 40 PAs in left, 60 PAs at short, 80 PAs at third, etc. Because of Gonzalez' versatility, I don't believe it requires bad luck for him to get a lot of playing time all over the field, all the while replacing someone at neutral or negative WAR value. Starters will get dinged up over the season and having a guy already on the roster to fill in for them can lead to a lot of PAs, all the while keeping your roster more stable in the process and not risking some pretty awful performances from replacement players, even if their scope of damage is limited to only a few plate appearances.

This is all true to some extent, but it's important to keep in mind that Gonzales is only a projected 1.5-2 WAR player himself. It appears to be a fairly solid 1.5-2 WAR, but there's a limit to how much you can expect to gain with him.

 

Even in an extreme scenario, if Gonzalez spent every single one of his ~500 PA replacing a 0 WAR performance -- Gonzalez is still only contributing 1.5-2 WAR. And we shouldn't expect that to simply add +1.5-2 wins on top of our median, neutral luck projection, because we don't expect to have 500 PA of 0 WAR performance in neutral luck. And in order for us to get to a situation where we actually might need 0 WAR replacement level players for 500+ PA in a season if we didn't have Gonzalez, we're likely losing some PAs from players with higher projected WAR totals than Gonzalez too, like Polanco (2.5 WAR), Rosario (2.7 WAR), and Kepler (2.7). The downgrade from them to Gonzalez is going to offset some of the gains that Gonzalez provides over other replacement options, and limit how much Gonzalez can improve the median neutral luck projection. And obviously, if the example is less extreme, and Gonzalez only spends half of his PAs replacing 0 WAR performance, that cuts this benefit in half too.

 

And while replacement level is not necessarily a super-predictable, exact science, the odds that Gonzalez would be displacing net 0 WAR performance over 500 PA is fairly low, given our roster, and the odds of him displacing notably net negative WAR performance over 500 PA is extremely low. That's not to say there wouldn't be a negative WAR guy in the mix if we had to turn to those replacements -- there probably would be -- but we'd be unlikely to *net* worse than, say, -1 WAR performance overall, among all of the replacements, in a reasonable worst case over those 500 PA. Looking at Fangraphs, Adrianza has played in parts of 6 MLB seasons now, and has never posted a negative WAR -- he's actually +1.1 WAR per 500 PA. Torreyes has 4 MLB seasons, again never posting a negative WAR, and he's +1.4 WAR per 500 PA. It's certainly possible they don't reach those levels in small samples in 2019, and especially over extended action -- but they're probably not posting significantly negative WAR either. They don't have the equity to accumulate much more than -0.3 WAR or so before they get replaced themselves. The same is true for our other replacement options (Cave, Austin, Granite, Reed, even Duda, heaven forbid!), and even a few our starters like Cron and Schoop. And simple odds tell us that it's highly unlikely we'd get enough of these -0.3 WAR performances in a row, without the occasional 0 or even +0.3 WAR, to wind up cumulatively worse than -1 WAR among this group. That's not to say it's impossible, I could flip a coin 10 times in a row and get heads each time too, but the odds of that series of events are low enough that we can't really assign much value to Gonzalez now for potentially mitigating such an unlikely occurrence.

 

Considering all this, it certainly seems to me like a median projection of +1 win, from adding the projected 1.5-2 WAR Gonzalez to this roster, is accurately capturing his median projected value as a versatile replacement. That's not to say that Gonzalez can't help more than that -- we can all come up with scenarios demonstrating otherwise -- just that the Twins can't really count on it being much more than +1 win right now, even allowing for the inevitability of injuries etc.

 

I hope that's clear? Let me know if something doesn't ring true to you. It was an interesting question, and it's been fun for me to ponder.

[Brock Beauchamp]

 

Considering all this, it certainly seems to me like a median projection of +1 win, from adding the projected 1.5-2 WAR Gonzalez to this roster, is accurately capturing his median projected value as a versatile replacement. That's not to say that Gonzalez can't help more than that -- we can all come up with scenarios demonstrating otherwise -- just that the Twins can't really count on it being much more than +1 win right now, even allowing for the inevitability of injuries etc.

I actually agree with most of this, which is why I haven't really refuted Gonzalez' +1 projection... as a median.

 

At its root, my questionable confidence in this kind of projection is two-fold:

 

1. That replacement players are actually available. We have a pretty long track record of watching some truly abysmal players get time on the field, though it should be noted a lot of that happened under the previous front office (which makes it as much of a management mistake as player).

 

2. That WAR itself - which is used for the projections - is even worth a damn when you get in SSS of

 

I'm not even saying the projection is wrong. It very well could be spot-on but when we start dealing with tiny numbers and lots of them, my faith in advanced metrics begin to waver (and, essentially, Gonzalez' replacements would be lots of tiny numbers).

[Otto von Ballpark]

 

not to make a silly point here, but the replacement player is of itself a somewhat mythical creature. It's a construct. It's not like our AAA farm is loaded with them. Guys will either out perform the replacement player or underperform. Very few will actually be a 0 WAR guy. 

Absolutely true. But there's enough of them out there, and they are fungible enough, that collectively they shouldn't be much worse than 0 WAR. Adrianza, Torreyes, Astudillo -- these guys were very recently freely available, and while perhaps not starters, they seem comfortably above 0 WAR to many of us now. Johnny Field, Ryan LaMarre, Gregorio Petit. Palka too. Cave and Austin didn't cost much, if you want to extend the spectrum above "zero cost" a bit.

 

Virtually every team stocks enough of these guys in the high minors every year, or has the resources to regularly grab them from someone else's bench or high minors, that their collective performance shouldn't be much worse than 0 WAR.

 

Heck, I've picked on the 2011 Twins for being unprepared and grabbing catchers off the street when Mauer went down -- but even those catchers off the street weren't necessarily that bad. Steve Holm was -0.2 bWAR in just 18 PA, then Rene Rivera was -0.1 bWAR in 114 PA. The big mistake was trusting Drew Butera so much (-1.6 bWAR in 254 PA).

 

 

Gonzalez, as I see it, allows us to not give a ton of at bats to guys who are much closer to replaceent player, or even below it. I like that. 

True, but not many teams unexpectedly wind up giving a ton of at bats to guys who are replacement level, at least not where signing a guy like Gonzalez before the season would be the only better option.

[Otto von Ballpark]

 

1. That replacement players are actually available. We have a pretty long track record of watching some truly abysmal players get time on the field, though it should be noted a lot of that happened under the previous front office (which makes it as much of a management mistake as player).

What examples would you cite for this? Seems to me that most such instances are where we're committed to playing a guy for other reasons (i.e. Morrison), not where we're simply cycling through replacement options seeking a 0-1 WAR band-aid.

 

 

2. That WAR itself - which is used for the projections - is even worth a damn when you get in SSS of <100 plate appearances.

I agree there's no extreme precision to WAR measurement -- although that cuts both ways, positive and negative. And an extreme SSS limits the extent of the potential damage in actual WAR.

 

Note that when I say 0 measurable WAR, I'm really thinking of a range from -0.5 to +0.5 actual WAR, at least (maybe more for defense). In any case, Gonzalez doesn't move the needle much differently between those values -- either in small sample, because Gonzalez isn't generating much actual WAR himself in a small sample, and over a larger sample because -0.5 actual WAR is obvious enough over a small sample that it isn't allowed to increase over a larger sample!

 

The Twins could be desperate enough sign a guy off the street again like Steve Holm, and I'm pretty confident he won't be worth more than -0.5 actual WAR before we've seen enough to move on from him to the next option. And for the 95% of the time when MLB teams aren't that desperate, when they're choosing between guys who aren't simply signed off the street -- Adrianza, Torreyes, even Palka or Petit or LaMarre -- I'm very confident they won't be worse than -0.5 actual WAR, and most will probably be 0 or better.

[Otto von Ballpark]

 

I actually agree with most of this, which is why I haven't really refuted Gonzalez' +1 projection... as a median.

Worth remembering that a median projection is really just an average of weighted scenarios. We can all come up with scenarios where Gonzalez helps more than +1 win -- like your 0 WAR or negative WAR replacement scenarios -- but we can also easily come up with scenarios where he helps less than +1 win. Even scenarios where Gonzalez proves to be an absolutely useful, valuable contributor, he's not necessarily improving our overall projection.

 

As an example, note that the median projection has Sano and Buxton worth ~2.3 WAR at their respective positions for 2019. After 2018, it's not hard to imagine a scenario where one of those guys contributes 0, either by performance or health, and Gonzalez spending most of his PAs directly or indirectly covering those spots. That scenario -- which sounds like an absolutely useful, valuable deployment of Gonzalez -- still doesn't represents a net increase from our pre-Gonzalez median projection, because Gonzalez himself only projects to 1.5 WAR:

 

82 wins pre-Gonzalez projection

minus 2.3 WAR from Buxton or Sano

plus 1.5 WAR from Gonzales

equals 81.2 wins

 

The scenario would actually need at least +1.8 WAR of good luck (Kepler breaking out, etc.) to get to where Fangraphs is right now, at +1 win post-Gonzalez (83 wins). And that would be fair, to project at least 1.8 WAR good luck to balance out -2.3 of bad luck. Rounding might be a factor here too -- you could easily refer to Fangraphs +1 win as +1.5 wins, perhaps even as high as +1.9, without changing my point, or constituting a blind spot in the projections -- but hopefully you can see how hard it is to move the needle significantly with these weighted scenarios, without assuming more than neutral luck.

 

Ultimately, it's worth remembering that Gonzalez doesn't bring us more than neutral luck -- which could represent many wins; rather, he just shifts our neutral luck point upward, which appears more modest (but can still be useful and valuable, and again, doesn't necessarily indicate a blind spot in the projections).

 

(And note I'm not claiming any precision to measurable WAR or accuracy in these particular projections, just using these figures for illustration purposes of actual value.)

[Otto von Ballpark]

Just want to say, I have enjoyed this discussion and all participants!

 

My overly long posts are not screeds of negativity. Not about Gonzalez, and certainly not about any viewpoints here. Just exploration.

[yarnivek1972]

WAR is the enemy of all mankind. The thought of WAR just blows my mind...