We Need To Talk About Analytics

[Matt Braun]

Now, I really did not want to write this but I thought too much about it and whoops! Here we are.The past few days have seen no less than a million arguments from baseball fans, ex-players, and talking heads on the role of analytics in the sport. All of this was sparked by the decision by Kevin Cash to remove his starter, Blake Snell, in the 6th inning of game 6 of the World Series. Given that we all already have strong opinions on everything as is, this decision specifically sent everyone into a mad frenzy.

 

What were we mad about? Analytics! The bastard had ruined the aesthetic of our game and directly led to the fall of Rome Rays losing in the World Series. Never mind that analytics got them to the World Series in the first place, or that they lost to a team that is also analytically-driven, or that their offense could only score one run against the Dodgers’ myriad of pitchers in game 6. But I digress.

 

This leads me to a crucial question; what does the term “analytics” mean?

 

Anyone?

 

“Analytics” has become sort of a “woo-woo” buzzword among fans and figure heads when it comes to the game of baseball. It describes a sort of magical new-age thinking that can be hard to pin down but is absolutely understood by all who watch the game. Some of this stems from the fact that we don’t know the majority of what teams are looking at because those secrets remain locked behind closed doors. Instead, a lot of what we now understand as “analytics” stems from people like Bill James and Billy Beane through the quite famous story of “Moneyball”.

 

“Analytics” became synonymous with a sort of “different” decision-making. One that ignores traditional baseball logic.

 

While this understanding is partially correct, some of this basis is misguided. "Analytics" in its most basic form is quite simple. For starters, utilizing a platoon would count as analytics. Bringing in a left-handed reliever to face a left-handed batter would also count as analytics. Hell, as long as the lineup wasn’t set by pulling names out of a hat, then analytics played a role in that as well. All of these are choices that managers have been making since the dawn of the game but still fall under the umbrella of “analytics” just like using a hitter’s batted-ball profile in order to shift the defense against him. The only difference is the level of complexity.

 

At the end of the day, analytics is just numbers and data. This data is then seen and interpreted by someone else who then makes a decision. At no point does any analytic choice pass through without any human engagement. Cash had to make the conscious decision to pull Snell. Even if a team decides to only go with the optimal strategy through analytics (which in itself, “optimal” is impossible to define), that decision was still consciously made by a human. Sort of like “if you choose not to decide, you still have made a choice”.

 

Analytics is simply a tool available to anyone who wants to utilize it.

 

Let’s get back to Cash and Snell. Immediately, taking Snell out became an analytic choice in our mind due to teams now being more aware than ever of the fact that starters tend to do worse the 3rd time through the batting order. This was also the reason Cash deferred to during the post-game presser. And yes, this is mostly true. As much as old pitchers and baseball dads like to complain about it, it’s a fact that most starting pitchers perform at a worse level when they face batters for the 3rd time in a game.

 

(Here are two Fangraphs pages. The first consists of the stats of SP going through the order the first two times from 2011-2020 and the second has the same qualifiers but for the third time through the order.)

 

The key word there is “most”. You shouldn’t apply a blanket concept to your players like that. Any decision to pull Snell has to be made with his personal splits in mind. Now, 2020 was a small sample but Snell actually performed a bit better the 3rd time through the order compared to the 2nd time (.379 wOBA allowed vs .407). Both numbers are awful, however.

 

Career numbers paint a different story. Snell has allowed a wOBA of .263 when going through a lineup for the first time, a wOBA .309 when going through that same lineup a second time, and a wOBA of .321 when greeting those hitters for a third time. (For context, Ian Desmond’s career wOBA is .321.)

 

Immediately, there becomes an apparent problem. What numbers do you look at? Cash knows Snell’s historic numbers as well as his recent ones. Which do you put more weight into and how much weight do you put into it? How does Snell’s recent injury history play into it? None of these questions even touch on the specific players Snell was set to face or how his replacement would fare against them as well or how rested his bullpen was or, well, I think you get the point now.

 

This was also a hot topic among Twins fans during the playoffs. Rocco Baldelli pulled both Kenta Maeda and José Berríos after 5 innings each during their respective starts. Some saw this as “over-managing” (and in the case of Berríos, I agree).

 

The point of this isn’t to actually prove whether or not Cash’s or Baldelli’s decisions were “analytical” or not because, well, all decisions in baseball are and always have been analytical. The only difference is that we have different numbers to look at. Putting your big bopper who rakes in RBIs is technically an analytical decision as much as it makes me want to squirm. The manager used data that he had available to make a decision that he thought was best for the team.

 

Keep in mind as well that analytics are never perfect. No system of decision making is. The point of utilizing advanced methods of interpreting information is to improve your odds. If your odds of making a “good” decision go from 55 to 60 percent, then analytics have succeeded. But even that leads to a significant amount of scenarios where failure occurs. That’s normal. It happens in any event where someone must make a decision with random variables.

 

All I want to do is try and steer the conversation in the correct direction. We shouldn’t be talking about whether analytics failed Cash and the Rays but instead whether Cash made appropriate decisions based on the information he had. It appears that the answer is “no” but we can never truly know as the situation only played out once. If Nick Anderson induces a double play and the Rays go on to win, then I doubt the conversation is as aggressive.

 

But who knows? These types of discussions are what drive baseball thought and will be what continues to push the metagame of this beautiful sport.

 

MORE FROM TWINS DAILY

— Latest Twins coverage from our writers

— Recent Twins discussion in our forums

— Follow Twins Daily via Twitter, Facebook or email

 

Click here to view the article

[Doctor Gast]

I believe Cash made the right call. Facing a powerful top of the line-up for the 3rd time w/ the best BP is a no brainer. Snell`s success is a result of Cash`s decisions that have been  made up to that point. If Snell would have stayed in the game & been ripped & lost the game, he`d be the goat. Instead he was praised by Cash for doing his job admirably which he did.

[Shaitan]

A lot of people use (or blame) data without understanding the data. 

[yeahyabetcha]

I believe Cash made the right call. Facing a powerful top of the line-up for the 3rd time w/ the best BP is a no brainer. Snell`s success is a result of Cash`s decisions that have been  made up to that point. If Snell would have stayed in the game & been ripped & lost the game, he`d be the goat. Instead he was praised by Cash for doing his job admirably which he did.

 

The Rays lost game....based on that, Cash made the “wrong” decision

[mnfireman]

Analytics would have had Morris coming out of Game 7 in the 5th inning.......

[Parker Hageman]

 

Analytics would have had Morris coming out of Game 7 in the 5th inning.......

 

In 1991, Morris' best results during the regular season came on pitch 101 and beyond. 

 

He had 113 plate appearances after throwing 100 pitches and held opponents to a .194 average and a .568 OPS. 

 

Maybe TK's analytics team tipped him off.

 

https://www.baseball-reference.com/players/split.fcgi?id=morrija02&year=1991&t=p

 

 

[Darius]

The problem with analytics is it’s trying to analyze a black and white outcome in a world full of grey area.

 

This isn’t the same thing as the likelihood of a certain number coming up on a roll of dice.

 

Clearly, analytics have their place. But, some are getting away from human element, and that’s where it can hurt. There is such a thing as clutch. It matters which batters are coming up in the next inning. It matters how the guy feels, his confidence level, days of rest, etc. There’s a lot of things that aren’t accounted for that cause religious followers of these things to be wrong more than they think they should be.

[Nine of twelve]

 

“Analytics” became synonymous with a sort of “different” decision-making. One that ignores traditional baseball logic.
 

It's become a pejorative term but...

 

Analytics is simply a tool available to anyone who wants to utilize it.

 

The only thing that has changed is that over the years we have greatly improved the depth and quality of the data. That and the term "analytics" has supplanted the term "statistics". All managers have used analytics for decades. I think modern-day managers use analytics more than in the past simply because there is more information that is useful but no manager makes decisions based only on the numbers.

[Twodogs]

In 1991, Morris' best results during the regular season came on pitch 101 and beyond.

 

He had 113 plate appearances after throwing 100 pitches and held opponents to a .194 average and a .568 OPS.

 

Maybe TK's analytics team tipped him off.

 

https://www.baseball-reference.com/players/split.fcgi?id=morrija02&year=1991&t=p

I wonder if there were even 113 plate appearances after pitch 100 for the whole twins team this year? If you never get there then you won't know what those numbers could be? If Morris was pitching today for the twins he wouldn't have ever had the opportunity to pitch past 5 or 6 innings.

[mnfireman]

 

In 1991, Morris' best results during the regular season came on pitch 101 and beyond. 

 

He had 113 plate appearances after throwing 100 pitches and held opponents to a .194 average and a .568 OPS. 

 

Maybe TK's analytics team tipped him off.

 

https://www.baseball-reference.com/players/split.fcgi?id=morrija02&year=1991&t=p

 

Maybe that same analytics team should have told TK not to PH for him in ahead 2-1 in the 6th in Game 4 then....

[Hosken Bombo Disco]

Instead of asking starters to go 7 innings every five days, teams today seem happy to get 5 innings every six or seven days. At least in Twins world that seems to be expectation. The results here in Minnesota speak for themselves.

[mikelink45]

Wow - an analytics apologist!  Numbers don't need to have people explaining and justifying, unless that is they are misused.  When analytics become arbitrary it is wrong.  Anything that is all or nothing is wrong and in this case Cash was wrong and all who try to justify his decision actually make it worse.  Let's just say that I and many many others are not happy and move on.  Every time we try to justify decisions by the numbers we reduce the human factor and those of us who are not analytic bound just get grumpier.  Nice try, but after reading this I just wanted to argue with you instead of letting it go. 

[tony&rodney]

A deeper problem is that we hear that in game decisions are made by upper management via messengers. Whether by electronic or foot traffic this practice is a can of worms. Cheating, gambling, and financially based switches are open season. Keep the game and the decisions on the field. There is enough information in the manager's binders and brains. If Falvine or other executives or employees want to make in game decisions, they should be managers. There are already too many coaches in the dugout. Analytics is just statistics used appropriately and have been a part of baseball forever. If I know a player cannot ever hit a curve, that is enough. We will throw curveballs then. If a pitcher is really slow to the plate, take the bag.

Kevin Cash has general respect around baseball but he and his coaches did not do anything special to advance the team this postseason. Dusty Baker, on the other hand, managed brilliantly throughout the postseason and came up just short with the reviled Astros. It was quite a credit to Cash that his players had his back after the game, but pulling Snell was a bonehead move. Mookie Betts, a pretty good player and also always respectful, noted that the Dodgers had not had any good at bats all day and were at the mercy of whatever Snell was throwing in Game 6. Cash himself would admit this but the word is that the decision was made upstairs. This is an issue.

[T_J]

To pull Snell was correct. To replace him with Anderson was incorrect. Numbers aside, Anderson was in a bad slump. When players are slumping, numbers be damned.

[Hosken Bombo Disco]

What does analytics say about slumps?

[Nine of twelve]

 

A deeper problem is that we hear that in game decisions are made by upper management via messengers. Cash himself would admit this but the word is that the decision was made upstairs. This is an issue.

Who is "we" that hears this? From whom does "we" hear this? Does that source know this first hand? And if so, how?

 

 

[strumdatjaguar]

The end of analytics starts with left handed hitters bunting away from the shift to third base. Fight analytics with anti-analytics.

[Nine of twelve]

 

The end of analytics starts with left handed hitters bunting away from the shift to third base. Fight analytics with anti-analytics.

There's no fight against analytics. It's a fight against the other team. And what you call anti-analytics is just analytics.

[Eris]

Analytics is about long term outcomes with large sample sizes. How did the infield shift work for the Twins vs Astros? The shift doesn’t stop a run from scoring if the hitter is willing an able to execute on an opposite field single.

 

Nick Anderson may have been the Rays best reliever over the entire season (and perhaps analytically he still is) but he had allowed at least one run in 6 straight postseason appearances. This is why teams have managers. Sometimes human judgement and instinct are needed.

[strumdatjaguar]

Every action (analytics) should have an equal and opposite reaction (anti-analytics). Lefties bunting to third and all hitters hitting opposite field singles is the start. Teach how to spray hits to all fields in the minors along with “launch angle” hitting. I’m cheering for anti-analytics, which is simply trying to react to a team which is over-reliant on analytics with good, old-school baseball strategies and a little element of surprise (doing that counter-analytic, unexpected reaction).

[Major League Ready]

 

The Rays lost game....based on that, Cash made the “wrong” decision

 

What if he would have brought in whatever RP they have that had the lowest probability of success but he pitched a scoreless inning. Would you have said that was a poor decision before knowing the results? Pitch hitting Adrianza for Cruz is not a good decision if Adrianza happens to hit a HR. It's a bad decision with a fortunate outcome. If you take a cab home because you are intoxicated, and the cab is involved in an accident, it would not have been a better idea to drive yourself.

 

I would suggest the right decision is the one with the highest probability of success. However, I would allow for overriding the analytics when a guy is in the zone like Snell was in game 6.

[Steve Lein]

Cash made that decision based on an overall philosophy of he didn't wan't his pitcher to face guys three times. His "analytics playbook."

 

Do you know how the next three guys had done against Snell in this game when he was pulled?

 

They were 0-6 with six strikeouts (Betts, Seager, Turner)...

 

You can't use your "analytics playbook" to make a structured decision in a game where nothing hardly ever unfolds the same (e.g.: game flow, situations). I said this in another thread, it can inform you, but you still have to manage individual games with what you see happening.

 

I think you're doing it wrong, especially in a game like this one was, if you take your horse out based on that input instead of considering what has happened in the individual game you're currently playing. 

[SkyBlueWaters]

 

... Lefties bunting to third and all hitters hitting opposite field singles is the start. Teach how to spray hits to all fields in the minors along with “launch angle” hitting. I’m cheering for anti-analytics, which is simply trying to react to a team which is over-reliant on analytics with good, old-school baseball strategies and a little element of surprise (doing that counter-analytic, unexpected reaction).

 

Anyone who has been paying attention to what Tony LaRussa has been saying after he signed on with the Chisox has heard a version of this.

 

A bit surprising to hear a manager say, “It’s actually easier to win now than ever if you’re allowed to use your powers of observation and experience. It really is.” He claims this because of the recent changes in the game that started with what was once called SABRmetrics. (I know, outmoded.)

 

LaRussa says he is not against analytics, he is all for getting as much pertinent information as he can. But after that he makes the point that if your opponent is giving you so much open field, use it. He is all for advancing runners rather than relying on the long ball, when the situation dictates.

 

I've never rooted for any of his teams (to the contrary), but 2021 could be a very interesting season, particularly in the ALC. 

 

How many days to spring training? Wait, what? It's only November ... which?

[Trov]

I think as more data has come available people have used it against the norms of history of baseball.  At times it has been successful, and other times it has failed. Just like any decision sometimes it works sometimes it does not.

 

Look at game 7 of 91 series.  It turned out to be good call to leave blackjack in.  What if we lose because we left him in, it would be second guessed forever saying he should have been pulled.  If he gets pulled and we lose, people would say why pull him he had not given up a run, ride him tell he fails.  Kelly made his call and it worked out.

 

Now a days we have more data to look at and draw from.  Managers act on that data.  Joe Madden was one of first to do this, looking at reverse splits, and pulling starters sooner than later.  Gardy used to never look at numbers like that and had old school, my guy is my guy, not matter the history.  

 

Cash made the call, he needs to live with it.  For all we know it plays out the same way and Snell gives up the lead, we will never know.  Personally, when it comes to Rocco pulling our guys quickly I was a little against it because they were looking good overall.  Maybe it plays out same way if we leave them in, we will never know.

[tony&rodney]

 

Who is "we" that hears this? From whom does "we" hear this? Does that source know this first hand? And if so, how?

Eno Sarris was one, from TheAthletic.com and there were two others from last week - too uninterested to look these up for you, but I do get your skepticism and perhaps I should have copy/pasted those links in to a doc right when I read them over the last few weeks and years. Thing is, like many of us, but not all, I read way too much baseball junk. It was lazy of me to write without proper sourcing and I apologize. 

 

[JadenPoint]

Analytics say this was the wrong move to pull Snell. He had better numbers 3rd time thru the order vs 2nd time, 2 years in a row. Also Betts had a terrible year vs lefties and his ops was awful the 3rd time facing a starter this season.

[#1ShaneMackFan]

 

In 1991, Morris' best results during the regular season came on pitch 101 and beyond. 

 

He had 113 plate appearances after throwing 100 pitches and held opponents to a .194 average and a .568 OPS. 

 

Maybe TK's analytics team tipped him off.

 

https://www.baseball-reference.com/players/split.fcgi?id=morrija02&year=1991&t=p

Great stat. Maybe they are a lot of pitchers who would succeed that way now as well, but we'll never know, because no one is allowed to pitch like that anymore.

[Twodogs]

Great stat. Maybe they are a lot of pitchers who would succeed that way now as well, but we'll never know, because no one is allowed to pitch like that anymore.

I really wonder how many hitters the twins faced this year after pitch #100. I'd bet not many. I know Maeda had a few when he was going for his no hitter, but I can't think of too many.

[ashbury]

I really wonder how many hitters the twins faced this year after pitch #100. I'd bet not many. I know Maeda had a few when he was going for his no hitter, but I can't think of too many.

They keep track of these things, and you are quite correct - the Twins as a team had a season total of 7 PA across 4 games, after pitch #100.

https://www.baseball-reference.com/teams/split.cgi?t=p&team=MIN&year=2020#all_pitco

Maeda had 1 game, 4 PA.

Berrios had 2 games, 2 PA.

Pineda had 1 game, 1 PA.

 

Not much, is it?