How did our model do at the deadline?

 

Starting with the Aaron Civale trade in early July, here’s the scorecard:

 

Number of trades: 72

Number accepted by the model: 60 

Acceptance rate: 83.3%

Average value variance per deal: 3.7

 

Not bad, but we’ve done better. 

 

Wins: Significant trades on which our model was very close – aka “fair deals”:

 

Hunter Harvey, Austin Slater, A.J. Puk, Randy Arozarena, James Paxton, Austin Hays/Seranthony Dominguez, Ryne Stanek, Nate Pearson, Nick Mears, Danny Jansen, Jesse Winker, Carson Kelly, Michael Lorenzen, Ty France, Quinn Priester/Nick Yorke, Justin Turner, Jalen Beeks, Frankie Montas, Amed Rosario, Lucas Sims, Josh Bell, Alex Cobb, Caleb Ferguson, Lucas Erceg, Paul DeJong, JT Chargois, Paul Blackburn, Trevor Richards, Andrew Chafin, Bryan De La Cruz, Isiah Kiner-Falefa, Tanner Banks, Kevin Kiermaier/Ryan Yarbrough, Dylan Carlson/Shaun Armstrong, Reilly/Cook, Mark Canha, Martin Perez, Austin Slater (yes, he was traded twice), Dylan Floro, Enyel De Los Santos

 

Borderline calls: Significant trades marginally accepted by our model as overpays within reason:

 

Zach Eflin, Yimi Garcia, Carlos Estevez, Huascar Brazoban, Eloy Jimenez (depending on how much cash is being sent to Baltimore, this could be in the above category), Luis Garcia, Gregory Soto

 

Misses: Significant trades rejected by our model as unreasonable overpays:

 

Aaron Civale, Jazz Chisholm Jr., Jason Adam, Isaac Paredes, Tommy Edman/Erick Fedde 3-team, Lane Thomas, Yusei Kikuchi, Jorge Soler, Mark Leiter Jr., Trevor Rogers, Tanner Scott, Jack Flaherty

 

So while 83% is a decent enough hit rate, which compares favorably to previous deadlines, this year was marked by too many higher-profile misses. 

And that’s because, as was obvious to everyone, it was an extreme seller’s market. There were 11 sellers and 19 buyers. That imbalance of supply and demand led to more overpays than usual, a few on the extreme side. 

One of our longtime users asked, what if this is the new norm? Given the recent changes to the playoff structure, where more teams have a shot mid-season, will we see this imbalance every year? And if so, should we adjust our model for it?

Maybe. We knew the day before the deadline that the top players left on the market, such as Jack Flaherty and Tanner Scott, were going to command big overpays, based on the patterns we were seeing. The precedent had been set by the Padres in the Jason Adam deal, and the Astros in the Yusei Kikuchi deal. We left our numbers unchanged because, at that point, we wanted to be consistent with hewing to fair-market value lines we typically set. But we are starting to wonder if we should factor that in for future deadlines, if indeed this is the new normal.

 

Common themes

 

Players with question marks

 

Another unusual factor at this deadline was that players with large question marks were traded. Consider:

• Tommy Edman hadn’t played all year
• Erick Fedde pitched well all year, but there was an iffy track record before this
• Jazz Chisholm, Jr. had well-documented “personality” issues

All three of these presented modeling challenges. How much, if any, would Edman’s injuries and long layoff affect his future performance?

Is Fedde’s performance this year indicative of his new normal, or is it a blip? We went with the former, based on the fact that he had accumulated enough of a sample size this year to suggest that he’s figured out what works for him, and the fact that several pitchers who had gone to Asia to remake themselves had transitioned back to the U.S. successfully (Merrill Kelly, Miles Mikolas, among others).

Personality is impossible to measure. We knew Jazz’s market might be limited because of that issue, but we also knew some teams didn’t care, and could easily absorb it (the fact that, so far, he’s transitioning well to his new clubhouse in the Bronx supports that). All we can do as modelers is stick to the facts, which were enough to justify a higher price tag than what he was traded for.

 

Statcast matters – unless it doesn’t

 

One of the more confusing themes we saw was: which teams were using more advanced metrics to value their players? The Rays/Cubs trade of Isaac Paredes may provide a glimpse into this. As is well known, Paredes has great surface stats, but very poor Statcast numbers – he doesn’t hit the ball hard, and gets by with pulling everything to left. Christopher Morel is the opposite – not great surface numbers, but hits the ball hard everywhere. On that basis, the Rays maybe got the better player. Morel’s expected stats are better, Paredes’ are worse.

But most teams use more established versions of WAR (there are three: Fangraphs, Baseball Reference, and Baseball Prospectus), which fundamentally use more common performance stats, but weight them differently. 

So are the Rays reconstructing their own version of WAR using expected stats? We suspect they are.

In our model, we use expected stats to modify the performance data, which we think gets us pretty close. 

But what if a team isn’t using expected stats? This trade suggests the Cubs are not. They then traded Mark Leiter, Jr. – whose Baseball Savant page is very red, meaning his expected stats are excellent – for two low-rated prospects. Leiter is worth a lot more than what they got, and the Cubs may end up getting the short end of the stick in both deals over time.

 

Prospect valuations, part 1: The Rogers trade

 

We got a lot of flak on Twitter from Orioles fans who were surprised that the Trevor Rogers acquisition was an underpay by Baltimore in our model. We don’t usually see this. Why be angry that your favorite team got a good deal?  

To be fair, we have a lot of new Twitter/X followers who may not be as familiar with how this works. Many look at surface stats, and think Rogers is meh. Many also look at outdated prospect rankings on MLB Pipeline and believe their prospects are worth more than they actually are. 

So to be clear: the package that Baltimore gave up is nothing special. Connor Norby is rated by Baseball America as a 50H, which means he has the upside of a future regular – not a star, mind you – but with high risk of achieving that status. Fangraphs rates him as a 40, which is a marginal bench bat. 

Kyle Stowers had been passed over several times by the Orioles, which we have found is a clear indication of value. If fans think Stowers is so valuable, why wasn’t his own team playing him? He had three years to win a job, and could not do so, even when other outfielders were struggling. That matters. He is most likely a bench bat going forward.

Rogers, meanwhile, has two years of control after this, has been performing well lately, and is very inexpensive in salary terms. As a result, he has a fair amount of surplus value, and the market for starters was hot. This was a very good deal for Baltimore, who need him more than they do a bench bat and a guy who is blocked at 2B and may never amount to much.

 

Prospect valuations, part 2: Use the right data

 

Interestingly, there were few top prospects traded this deadline. Thayron Liranzo was the prominent one – he was dealt to Detroit as the lead piece in the Flaherty package. Jake Bloss is a borderline Top 100 arm, and led the package to Toronto for Yusei Kikuchi. The Padres dealt two former top pitching prospects in deals for relievers – Dylan Lesko in the Jason Adam package, and Robby Snelling in the Tanner Scott package. Both had been struggling, and both had been downgraded heavily, out of the Top 100, by Baseball America. So those two were not as valuable as they might have been perceived to be. 

When evaluating deals, it’s important to use updated prospect data. Most casual fans check MLB Pipeline, which unfortunately is not helpful, as they haven’t updated their ratings in quite some time, and therefore do not reflect current realities. Further, prospect valuation is less about rankings than ratings if you check team lists. The #3 prospect on a team with a strong farm (e.g., Liranzo on the Dodgers) is much more valuable than the #3 prospect on a team with a weak farm (e.g., Snelling on the Padres). Look at how they’re rated, not how they’re ranked.

 

Prospect valuations, part 3: Expendability

 

A.J. Preller is an outlier, as the Padres’ President of Baseball Operations will trade most of his farm for a reliever.

But what if the guys he traded don’t amount to much? Notably, he kept his two top prospects, Ethan Salas and Leodalis De Vries. Those two project to be future stars. Anyone below that line was fair game, and he didn’t seem to care how many of them he sold off. Could that be because they have a lower probability of ever making much of a big-league impact? 

The same could be said of the Astros, who traded three prospects for two months of Kikuchi. But what if Jake Bloss is nothing more than a fifth starter, Loperfido ends up being a bench player, and Wagner never makes the show? Does that really matter? All of those, then, it could be argued, are easily replaceable. 

That implies that teams are less concerned about selling off anyone rated under, say, a 50, because they’re the future bench players and marginal pitchers who are more easily replaced. You keep the keepers, because those are your future stars and regulars; everyone else is disposable. 

Granted, we’ve known most of this for years, but what we’re seeing is more of a willingness than before to trade off the tier below 50 and use them for trade currency. Quantity doesn’t matter much because, in theory, none of them will move the needle as much.

 

In any event, we will fine-tune our model, as always, to stay abreast of trends and correlate to an ever-changing market.