RSOB Chapter 2: How Psychology is Out to Get You

New to Ron Shandler’s Other Book? Read the Introduction. 

---

For the sake of argument, let’s say that you buy into everything I wrote in the last chapter. (I won’t delude myself into thinking that everyone is drinking my Kool-Aid.) Let’s say that you agree that player projections are garbage… um, flawed. Still, our brain plays its own tricks on us. Even if we could believe the data, there are psychological pitfalls that also do us harm.

We base decisions on small sample sizes.

When Billy Beane signed Rich Hill to a one-year contract, I thought, “$6 million for one year? What is it in his skill set that made him worth that much? Four great starts in September?”

Admittedly, those starts were very, very good. He threw 23 innings, gave up only 10 hits and three runs, and struck out 30 while walking only two. Only one of those starts was at home in Fenway Park, against the Orioles. The other three were at Tampa, Toronto and in the Bronx. He faced some good and bad lineups. Tampa was a pushover; his outings against Toronto and New York were not as good as his line scores would indicate. However, he owned the Orioles. But $6 million?

How much are you going to pay for Hill in 2016 auctions? More than $5? More than $10?

In what round will he go? Earlier than the 20th? Earlier than the 15th?

For me, Hill is nothing more than a $1, end-game 23rd round player. Odds are he won’t last that deep into a draft except in the shallowest of leagues. (His current ADP in the NFBC is #269, or late 18th round.) But what evidence is there that a 35-year-old with a 4.54 career ERA in 500 innings over 11 seasons, who has not seen even 60 innings in a year since 2009, is really worth more than an end-game flyer?

How do 23 innings against tired batters at the end of September trump his other 477 career innings? He moved to the other side of the rubber and adjusted his arm angle? Where is the proof that he will be able to maintain that advantage and stay healthy over a full season?

But man, he looked so good!

But four starts! Over a 4-game stretch early last year, Corey Kluber posted an ERA of 7.43. Over a 4-game stretch in early May last year, Clayton Kershaw posted an ERA of 5.00. Would you have considered cutting bait on those pitchers after those four starts?

Of course not. But they had a track record of much better.

Exactly. And Hill has a track record of much worse. It works both ways, and even moreso in Hill’s case, because – in the inimitable words of colleague Lawr Michaels –”baseball is hard.”

Four starts is a ridiculously small sample size from which to draw conclusions, but even larger sample sizes can be just as suspect.

I always found it odd that major roster decisions were made based on spring training performance. Four weeks of lackadaisical games against an mixture of veterans getting their rust off, marginals working on a new pitch or batting approach, and minor leaguers playing like minor leaguers. Congratulations, you just won a spot in a major league rotation by posting a 2.50 ERA in four games against Crash Davis and Pedro Cerrano. Really?

Better yet, some teams make final roster cuts based on a player’s last outing in the spring. The difference between success and failure – a major league paycheck or a bus trip back to Reno – comes down to a sample size of four at-bats or six innings pitched? Yet we will chase those players on Draft Day because they “made the cut.”

That all said, there are times when a small sample is enough to make an informed decision… um, no – an educated speculation – but it’s not often. A high-skilled minor leaguer who gets promoted and immediately succeeds is a decent bet to at least have a high floor. (Note: 1st round is not a floor.) A player coming off the disabled list who struggles mightily in his first few outings is a good bet to not be completely healthy.

But if Daniel Murphy goes for even a dollar more than he should based on his post-season heroics, I want to be in that league. If a fellow owner starts shopping Josh Donaldson if he’s batting .220 after four weeks, I want to be in that league. And if Rich Hill goes any earlier than Round 20, I want to be in that league.

We try to ferret out patterns within statistical noise.

Humans (including you and I) are hard-wired to try to find patterns. In its grandest sense, we do this to survive. The world is full of chaos – even in non-election years – and it’s the way our brains attempt to create order.

Baseball analysis is similarly all about finding patterns in data. We see a batter hitting 8, 10 and 12 home runs in successive years, and we immediately label that as a growth trend. Maybe it is.

But research back in 2010 by Ed DeCaria showed that the odds of the next data point in that series being 14 are small. In fact, the greatest odds are that the next point regresses back to 10, or even 9.

As described in the last chapter, since that we don’t even know how real 8, 10 and 12 are, it’s difficult to conclude that there is any trend at all. That 8-HR year could have been 13 if five of his doubles had traveled another 5 feet. That 12-HR year might have been 9 if not for those three nights when the wind was blowing out.

We fantasy leaguers need to find patterns. That’s the starting point for the entire forecasting process. But when the data itself is suspect – obscured in great measure by noise – maybe it’s better not to be looking for something that might not exist. Like better sentence structure.

Let’s play a little game.

Oo, I like games!

Good! Here is a short series of data points representing one player’s Rotisserie earnings during his first three years in the majors: $7, $15, $18. Tell me what you think he earns in year #4.

Well… it seems like growth, but you warned me against assuming that. I’ll take the bait. I’ll say that he earns $16 in year #4.

That’s a very reasonable guess. Any of $14, $15 or $16 would take an appropriate level of regression into account. In year #4, this player actually earned $23.

What? You tricked me!

I didn’t trick you. This is an actual player. So, now you’re faced with a 4-year trend: $7, $15, $18, $23. What does this player earn in year #5?

Okay, now you’re screwing with me. Logic dictates that I say $19 or $20, but you’ve already primed me to expect the unexpected. I’ll say $25.

Another good guess. Most analysts would probably have stuck with some type of regressed value, and I can tell you that the Forecaster projected this player to earn $22 in year #5. But he actually earned $28.

Of course. Four straight years of increasing earnings – is this a real player? Should I believe you?

You can choose what to believe. But let’s keep going. We’re now at $7, $15, $18, $23, $28. What does he do in year #6?

There is no way this can keep going. I’m going to say $24. That’s my final answer.

And that is the correct play. Regression is always the correct play. The Forecaster projected $26. But he actually earned $32.

You’re playing me. You clearly picked an outlier… if he actually exists at all.

Well, that’s one thing you got right. A player with this consistent a 5-year trend is clearly an outlier. Do you want to keep going?

Sure, why not? It’s only a guessing game at this point.

Okay. $7, $15, $18, $23, $28, $32. What’s next?

Regression is always the correct play… even when it isn’t. I’ll say $29.

Remember that Matt Cederholm said, “Players who earn $30 in a season are only a 34 percent bet to repeat or improve the following season.” Given that, it would seem that the odds of him continuing to improve, or even holding steady, are low. In the next section, I’ll show you how that skews our expectations, but for now… in year #7, he earned…

Wait for it…

$28.

Hooray! The planets finally align! Does it keep going?

For sure. Give it one more shot. This last data point coming up is 2015.

$7, $15, $18, $23, $28, $32, $28. It’s no less tricky now. Was 2014 an outlier? Does he rebound in 2015? Or does the downward trend continue? I’ll give you one hint: he was 29 years old in 2015.

That’s still young. I’d have to say he rebounds a bit. I’ll peg his earnings at $30.

Yeah, but no. He only earned $19 last year. Forecasting is a tough game.

More like a sucker’s game. Who was the player? Was he real?

Adam Jones is very real. And as much as this exercise was frustrating, a look at Jones’ career provides a pretty slick bell curve: $7, $15, $18, $23, $28, $32, $28, $19. We would be so lucky if every player’s career followed as fine a trend as this. They’d be a cinch to project each year (oh, the irony!).

FWIW, the Forecaster projects him to earn $27 in 2016. You want to bet the over or under?

Forget it. I’m out of this game.

Me too. I’m getting out of this game as well. That’s why I’m writing this book.

Here are two other interesting players:

A 28-year-old: $16, $27, $25, $39, $40, $35, $28. Somewhat similar to Jones, this is a high-end player who seems to be fading prematurely. But most drafters will continue to spend over $30 and a first round pick on Andrew McCutchen with barely a second thought. They’re smart.

Another 28-year-old: $2, $13, $23, $10, $25, $14. I have no clue what to project for this guy in 2016. But as I noted in the last chapter, Alcides Escobar’s underlying skill set is pretty consistent.

Wait a minute, wait a minute. Is any of this data valid? Can we even use Rotisserie earnings to evaluate players? Isn’t this the same argument you made against using OPS?

Nice job. That’s why all of these data points are suspect. Adam Jones’ bell curve is probably not nearly as consistent as it seems. Escobar’s values are probably not as erratic as they seem. Still, there are two areas where Rotisserie dollars can have some value.

1. I wouldn’t use past Roto earnings to project next year’s dollar value, but they do have an advantage over other metrics. This is because the dollar calculation normalizes statistics to the level of offense and pitching each year. So a 30-HR performance in a high offense season would earn fewer dollars than that same 30-HR performance in a low offense season. The above data sets are fine to evaluate within the limitation of the imprecise inputs.

2. Sharp changes in performance are reflected pretty accurately, even if the precise dollar values are inexact. So we can use roto dollars to suggest the magnitude of a breakout or breakdown performance.

If there’s one thing that I’ve learned about breakouts, it’s that they don’t typically arise in a straight line out of a trackable growth trend. Most folks perceive a breakout player’s dollar values to look something like this: $8, $10, $13, $25. But the reality is, most breakouts look more like this: $8, $13, $10, $38 – a massive, unexpected spike. Here are a few examples:

$-3, $5, $0, $2, $31, $36, $12, $19, $32, $26

Jose Bautista shuttled between full-time and part-time work his first four years in the majors before exploding in 2010. If we had focused on the skill and viewed his playing time as a variable risk, we might have been able to see something coming. His performances since then have fit no discernible pattern.

$23, $20, $10, $40

A rookie tease and then two years of waiting until Bryce Harper finally broke out. The skill was always there, but virtually nobody took the risk to project 2015 as anything more than a growth year.

$14, $13, $30, $25

Similar to Harper, mediocre early returns preceded the breakout for Todd Frazier.

$-7, $-5, $-15, $-5, $19

This data set is through 2014. What would you have projected for this player coming into 2015? The Forecaster believed that his improvement was real and projected $18. But Jake Arrieta earned $44. Projecting 2016 will be fun. That’s why we need to find a better way.

In the next chapter, we’ll start looking at players as entities that possess assets and liabilities. By evaluating each separately, we can sometimes detect the breakouts before they occur.

One last thing. This quest to draw conclusions about performance trends extends to teams as well.

Last spring, I made an out-on-a-limb projection that the Cubs and Astros would meet in the World Series. At the time, it was an outrageous speculation meant only to help everyone keep their eyes open. After all, the Cubs were coming off five consecutive losing seasons and the Astros were just a year removed from three consecutive 100-loss seasons. We all acknowledged that they were teams about to turn the corner, but a World Series appearance? Ridiculous.

Just like breakout players, teams don’t always advance or decline in a straight line. With teams, there are so many moving pieces, and so many opportunities over six months to tweak, that it’s tough to predict performance from one season to the next. Entertaining the possibility that the Cubs and Astros would even make the play-offs defied our quest for logical patterns.

Every season starts as a blank slate; last year’s won-loss record is not the starting point for this season’s results. It works the same way for players.

We look at research results based on aggregate data and draw finite conclusions about individual players.

I’ve done a ton of research over the past 30 years as have the analysts who’ve written for me at BaseballHQ.com. Most of this stuff is incredibly insightful and the findings really help us understand the components of true skill.

The problem is that these results reflect tendencies on a macro level. None of them produce a percentage play that’s good enough to make micro player decisions with any confidence.

A standard fantasy roster with 23 players is way too small a sample size for any of this to matter. (There’s that line again.) You are not going to be able to leverage miniscule percentage differences with so few chances to be right or wrong. Those 23 players are just not enough opportunities to cover your risk.

Here are three widely-used variables that are almost always a waste of time to worry about.

Age: Research shows that players’ skills peak at a certain age – 26, 27, 28, 23, 31 – pick a number. But those are just rough averages. Not every player is going to peak at a given age. So targeting 28-year-olds in your draft will only pay off if you’re in about 30 leagues. And even then, you might end up passing on a 21-year-old Carlos Correa (pictured above) who hits the ground running or an Alex Rodriguez who has a huge rebound season at age 39.

With only 23 chances, the odds of rostering an outlier are not much different from the odds of rostering a player that fits your target.

However… there are a few times when the odds are high enough to pursue. Eventually, players age out of rosterable skills. That age is different for every player, but the older they get, the higher the odds. So, if a player has a career year in his mid-to-late 30s, bet against a repeat. If a player has a crappy year in his late 30s, bet against a rebound. Those are higher percentage plays and are pretty much the only ones worth chasing.

Park effects: I know from experience that most touts go through a painstaking conversion process every time a player switches teams.

Umm… switching teams?

No, not that type of conversion.

I’ve come to find the exercise of adjusting projections for park effects mostly a waste of time. In recent years, we’ve seen players like Brian McCann and Evan Gattis move to new parks that should have turned them into 30-plus HR monsters. In both cases, any change in power skill was far short of expectation. Even extreme ballpark changes are inconclusive. Yes, Michael Cuddyer’s batting average tanked in moving from Coors Field to CitiField, but was it park effects or the fact that he was hurt for part of the year? Wasn’t Nelson Cruz’s power supposed to disappear moving from Baltimore to Seattle? And everyone is going to be upgrading Brett Lawrie moving from spacious Oakland to hitter-friendly U.S. Cellular Field, but he saw his power actually improve when he came to Oakland from the just-as-hitter-friendly Rogers Centre.

That brings up a bigger question: how do you know that an increase or decrease in a player’s output is really park-related?

If a 30-HR hitter moves to a park that increases power by 20 percent – which is huge – then we could expect him to now be a 33-HR hitter (the percentage only affects home games). But a 3-HR increase is well within the limits of normal statistical variance. How do we know that normal skills growth didn’t drive the increase in home runs? Or simple statistical volatility? Or a trio of well-timed gusts of wind? If the 26-year-old Lawrie has a big year in 2016, I’d more likely attribute it to skills growth than ballpark.

It’s even more fuzzy with the ratio gauges. David Price’s move to Fenway Park would be expected to add a chunk of ERA to his projection. But even if there was a 10 percent swing in run-scoring from Detroit/Toronto to Fenway (there’s not; it’s only around seven percent), Price’s ERA would increase about 0.25 of a run over a full season. That’s one extra run per month. It’s just not a projectable level.

However… if you are going to use it at all, focus on the margins. The noticeable impacts are only going to come from a hitter moving from one of the best hitters parks to one of the worst, or vice versa. The inverse goes for pitchers, obviously. I have given up calculating anything in between. Todd Frazier moving from Cincinnati to Chicago? Over 550 AB, the impact of park effects is going to be nothing more than a rounding error.

Team: If you have two players of comparable skill, but one plays in Kansas City and the other in Milwaukee, you’ll almost always opt for the Royal over the Brewer. Team environment matters, right? More runs and RBIs, more wins and saves.

Unless you invested in the Red Sox and Tigers last year. Seattle and Washington were supposed to play in October as well. Instead, last year’s fantasy winners were the ones who resigned themselves to drafting a few random Astros and Mets. Or maybe they tacked a few lowly Twins or Cubs onto the tail end of their roster. We failed to correctly predict team environment for those clubs that had a huge impact.

Even picking the right team is no guarantee of success. You could have invested in the most stable of the Cardinals’ arms, but you would have gotten only 13 wins out of John Lackey and only 12 out of Lance Lynn. The Dodgers would have been a prime source of stats, but nobody behind Adrian Gonzalez amassed more than 60 RBIs.

In Chapter 4, I am going to compare Dallas Keuchel to Sonny Gray. If I can prove to you that they are essentially the same pitcher, you might still opt for Keuchel because he’s on a better team. Maybe he is; maybe he’s not. But the odds of there being a significant difference in their eventual stats are probably not high enough to matter.

As a tie-breaker when everything else is equal? Sure. But I’m willing to bet you can find some other variable that will have more of an impact.

We are largely driven by recency bias.

We live in a world where we’re inundated in information. It’s far too much to process so we have to rely on smaller chunks that are easier to remember. And the easiest pieces of data to remember are those closest to the surface of our consciousness. Ask me what I had for breakfast this morning but forget about me remembering what I had for dinner two nights ago.

(“Fish and chips at that seafood restaurant.” – the wife)

The effects of recency bias on managing our fantasy teams have grown over time as the amount of information we’ve had to process has grown. Part of it is just the endless quest to grab at whatever we can. I’ve already talked about small sample sizes – that’s part of it – but these days, even a full season of aberrant performance often trumps a 10-year career of consistency.

How else can you explain why Miguel Cabrera is not still a Top 10 lock?

Recency bias drives each year’s ADPs. The quickest way to earn a first round ranking is to post first-round earnings the previous year. A.J. Pollock, Manny Machado, Nolan Arenado and Anthony Rizzo are pushing down first round stalwarts like Cabrera, Andrew McCutchen and Carlos Gomez after just one season of uncharacteristic production. This class of player that has supplanted the vets could well be the next wave of star talent, but are we passing judgment after just one season? After all, outliers run both ways.

It’s like we completely ignore one of the very first tenets of baseball prognosticating: Don’t project a player based on one season’s stats. After 30 years, have we learned nothing?

Is the oft-injured Machado really a better bet to return high first round value than the consistent track record of McCutchen, a player who has racked up a grand total of 15 days on the DL over the past seven years?

Is Jake Arrieta, who posted a second half for the ages and is about a 99.9 percent bet to regress – and significantly – a better bet to return elite value than Felix Hernandez or Chris Sale, long-term members of the Elite?

With the exception of Clayton Kershaw, the historical track record shows that pitchers earning first round value in one season almost never repeat the feat in consecutive years. Volatile pitching stats and the changing composition of the talent pool drive that phenomenon. But Arrieta and Zack Greinke are still likely going to get drafted ahead of Hernandez, and maybe even Sale.

 Finally, is it not ludicrous to include Carlos Correa’s name among 2016 first-rounders after 417 major league plate appearances? Doesn’t anybody remember Eric Hosmer (followed a .293 rookie year with .232) or Danny Salazar (followed a 3.12 ERA mid-season call-up with 4.25)?

Those are not Correa-caliber players. What about Mike Trout?

Is Correa really another once-in-a-generation player? Maybe he is, but are you going to bet on it by committing a core roster spot to a speculation of guaranteed greatness?

Well, maybe I will. I don’t want to miss out.

We make decisions based on the fear of missing out.

I get it that you don’t want to be the guy who misses out on the next Hall-of-Famer. But are you really, really absolutely certain that this is a can’t-miss player? After 417 plate appearances? Enough to risk that all-important first round pick?

Last year was a great example of what happens when you buy into the Fear of Missing Out. Everyone was convinced that Kris Bryant would get an early call and be That Guy, so he was drafted at inflated prices. But the teams that won leagues last year were not those that owned Bryant, because he was purchased at nearly full value. There was no advantage to paying that much; there was only the risk that an unproven player would fail. The winners were the ones that owned Machado, Pollock or Keuchel, or had Correa or Miguel Sano tucked on their reserve lists. Those were the monster profit machines.

When you draft Correa in the first round, there is far more downside than upside. If he is fully productive, you’ve set a very high bar for him to return par value (forget about profit; that percentage play is minute). Perhaps he has a higher floor than others, so your downside is mitigated. But we simply don’t know what that range is. Here is my completely unscientific take on Correa’s odds as a first rounder:

Profit         1%
Par value     20%
Some loss     60%
Major loss    19%

You can quibble with the percentages, but the general conclusion has to be the same: what are you chasing? I’d probably attach similar percentages to some of the other first-rounders too, but I’ll talk more about that in a later chapter.

If you’re overpaying for a speculation at the draft, you’re also potentially passing up on profit opportunities later on. As much as you think you can find profit in every player, you only get 23 chances, and there are at least a dozen other guys in your league, all thinking the same way.

This is particularly dangerous in the early rounds where we’ve shown that our overall track record is terrible. Here are a few interesting players of note:

                   # years drafted in 1st Rd     # years earned
Player             for Fear of Missing Out       1st Rd value
----------------   -------------------------     ---------------
Troy Tulowitzki               4                        0
Evan Longoria                 3                        0
Carlos Gonzalez               4                        1
Prince Fielder                4                        1

If you’re still craving Correa, I’ll show you how to quell that fix later in the book.

We base decisions on NOW.

There is a subconscious part of us that actually agrees with the fact that you can’t predict the future. If our decision-making process was fully conscious and deliberate, we might take an objective look at each situation with an eye towards tomorrow. Instead, we tend to take the easy way out and just view what is happening right now as a fixed reality.

But reality is not fixed. It is fluid. One decision begets uncertain outcomes, which beget other decisions.

English, please. At least give me an example.

Okay. “Once upon a time (early last season), there was a closer for the Seattle Mariners named Fernando Rodney. He had a volatile career – some very good years and some very bad ones – and despite there being some question about his ability to hold down a closer’s role, International Expert (and Man of Intrigue) Ron Shandler spent full-price closer dollars for him in Tout Wars ($16). Shandler reasoned that, despite Rodney’s erratic track record, he was the closer NOW.

As it would turn out, it didn’t take long for Rodney to turn into a pumpkin, wiping out Shandler’s investment (and relegating him to last place in saves for the rest of the season). When Carson Smith innocuously slid into the closer’s role, he immediately became the NOW guy, and fantasy leaguers around the world proceeded to exhaust a significant part of their free agent acquisition resources on a pitcher with far better skills than the deposed Rodney. Because, better skills and NOW.

These NOW investments also come with an inherent expectation of longevity – we expect the pitcher will hold the role for the rest of the year. But when it comes to closers, they hold that role until they don’t, and sometimes the in-season lifespan for that role is weeks, or days.

Smith’s ninth inning lifespan was about two and a half months. He started losing games and blowing saves in late July, and was supplanted by Tom Wilhelmson by mid-August. Wilhelmson’s skill set paled in comparison to Smith’s (and once Smith lost the role, he did not give up a run for the rest of the season) but that’s not what reality is about. Wilhelmson was now the NOW guy drawing whatever meager free agent resources were still left.

After the season was over, the Mariners responded to all this by tossing last year’s NOW guys to the curb and starting over with a bunch of new NOW guys.

And they all lived happily ever after.

Except for Shandler.”

Some stories don’t have happy endings.

But watch… NOW is going to come into play in many of our future conversations.

Nice story. I assume you didn’t win Tout Wars.

Um, no. Here are other ways that our decision-making processes are influenced by NOW:

As mentioned earlier, there are some players who lock down roles at the very end of spring training. We treat those NOW guys as fixed realities, bidding them up to full value on Draft Day as if “winning a job” is the only prerequisite to full-season success. (Yay, Dalton Pompey!) This also goes back to the small sample size discussion.

Your #4 starting pitcher gets off to a ridiculously good start. Despite the fact that his skills have not changed substantially and his recent success is against weak competition, you refuse to entertain trade offers, because he is doing well NOW. What if he keeps it up? Are you contracting an acute case of Fear of Missing Out?

Many of these psychological potholes are interrelated. They are all obstacles to success.

But enough pain, for now. It’s time to begin the construction process.

NEXT: The BABS Concept

Follow @RonShandler