The Z Files: Winning Tendencies, Part Two

To say fantasy baseball is all about profit is a bit myopic. Sure, a positive return on investment is obligatory, but profit alone doesn't win championships. Profit yields more stats to work with, no doubt. However, it's the nature of the stats and their efficient management into rotisserie points that drives championships. Still, it all begins with profit.

This installment of the Winning Tendencies series will focus on where teams derived profit in a general sense. The next part of the series will hone in on specific players and types of players. By means of review, data from the 2019 National Fantasy Baseball Championship Main Event will fuel the research. This is a contest featuring 38 leagues with 15 teams each. All 570 drafts were downloaded and utilized in the study.

Profit from Opening Day Rosters

Drafts aren't usually thought about in terms of player prices, but using historical data from conventional auction valuation, the expected return for each draft pick can be quantified. Here's a grid displaying typical expected earning per draft slot for 15-team leagues:

Each player drafted was assigned their expected earning based on their pick. They all have actual earnings based on final stats. Valuation is inherently flawed, even for known production, but the return on investment for each player can still be determined, using the assumption the player was active for the entire season. The overall profit/loss (p/l) can be summed for each team. What follows is the average opening day lineup profit/loss per standings position.

Yes, even the league champion's draft, on average, yielded a negative return on investment from their draft. In fact, of the 570 drafts, only 11 showed profit. While it's a good sign the top teams lost less than the others, working with data of this nature is tenuous at best. Convincing you a winning strategy can be based on drafting less of a loss is an uphill battle. There must be a better way.

Adjusted Profit from Opening Day Rosters

What if, instead of simply comparing a player's actual earnings to his expected earnings, it was compared to the actual earnings of everyone drafted at that same spot? For example, let's say Mike Trout was drafted 1.1 in every single league. It doesn't matter how well or poorly he performed; he helped/hurt each team the same. His p/l would be $0, right? Well, not quite.

Within each league, Trout still aided or damaged his team. The adjusted p/l can't solely consider all 38 players drafted at that specific spot. It's necessary to compare his impact to that league context.

One way to deal with this isn't to do a spot-for-spot comparison, but rather to average the individual p/l from each spot for players drafted just before and just afterwards. For example, the adjusted expectation for the 20th player selected would be the average of the expectations for the 15th through 25th players (five spots either side). The way to handle the first five players taken is to use the expectation for the top overall player in hypothetical slots "above" 1.1. For example, the math for the fourth overall pick would count the top player three times, so there are five "before" data points (1st, 1st, 1st, 2nd, 3rd), then 4th through 9th to get the requisite 11 players to average out for each specific slot.

Making this adjustment renders the following:

This is more practical and passes the sniff test as 292 clubs, just over half, drafted to a profit. This provides a realistic foundation for follow-up studies. Even so, note how the returns on investment don't perfectly line up in descending order. This speaks to the in-season management contributions discussed last week, as well as the need to efficiently manage stats into points. That said, recall that on average the second-place finisher added, by far, the most stats of any standings place. As can be seen above, this was often necessary to leapfrog the average third-place team's more profitable draft.

Fantasy MVP and Cy Young

Before moving onto further studies, there's an interesting repercussion from this method of determining return on investment. According to his ADP, Mike Trout's expected earnings should have been low to mid $40s. He returned just $31. However, based on the adjusted expectations, looking at everyone drafted at or near his spot, he broke even. Similarly, Mookie Betts fell short of expectations, but after the adjustment, he actually earned about $1 relative to other players drafted at or near his spot. This is true of most of the players in the first round, reinforcing the notion of safety and reliability early.

Here's a list of the Top 50 earners relative to their draft spot.

At the end of each season, I'm usually asked for my fantasy MVP and Cy Young. While I'm not sure it's worth downloading 570 rosters into Excel just to answer that specific question, based on the above, LeMahieu was the 2019 MVP while Kate Upton will be thrilled to know her hubby captured the fantasy Cy Young.

There is a practical use for this list, as by definition all the players listed performed well above expectations. While it doesn't take the downloading of 570 roster to identify players to put under the microscope for 2020, the ranking serves to prioritize which ones deserve the most scrutiny as you dig deep, discerning whose production is sustainable and those who likely have a date with the regression monster.

Profit: Early, Middle and Late

Knowing when competitive teams logged a profit could help in roster construction and risk management. Here's a table exhibiting the P/L for each average finish with the draft divided into thirds:

It's interesting to note the average first place team's chief profit advantage came early. Chances are, this was a combination of picks with a positive return and injury avoidance, which negates profit. The results conflict with conventional wisdom suggesting leagues are won in the later rounds. At least according to this data, competitive teams are consistently strong throughout the draft. That said, keep in mind the profit for late picks is skewed, since many of those players were released and replaced by reserves or in-season free agent acquisitions. Still, the notion you can't win the league early is at least up for debate, if not debunked. In the aggregate, league champions crushed early choices.

Average Draft Position Correlation

Every spring, there are a cornucopia of missives waxing poetic on the proper manner to utilize ADP (Average Draft Position). Having 17,100 draft picks spread over 38 leagues provides a great backdrop to investigate how closely winning teams follow the market.

The first study is a simple exercise: correlate the ADP of each team's selections to their draft spot. The closer to 1.000, the more a team stuck to the ADP.

Please keep in mind, this is average data, with each finishing spot the average of 38 different teams. There is no doubt some successful clubs strayed from ADP, though the results suggest the best teams didn't waiver from the market as much as the bottom dwellers.

Jumping ADP vs. Value Picks

An obvious deficiency of this correlation exercise is the uncertainty whether the deviations were the result of players taken earlier than their ADP or later. Here some subjective analysis is necessary. How should jumping the ADP or making a "value" pick be quantified? Before broaching that, my disdain for "value" is well documented. However, in this instance, value reflects a selection made relative to the market, and isn't a straight assessment of profit. That will be addressed shortly.

Three different levels of quantification will be deployed. Teams will be measured how far they wandered from ADP in terms of percentage: 10, 20 and 30 percent. Here's an example using an ADP of 10.

Twenty percent more than 10 is 12. For a selection to be considered a value pick using a 20 percent range, the player must be chosen 13th or later. Conversely, eight is 20 percent less than 10. The ADP is considered jumped if the player is drafted 7th or sooner.

To show how the range widens deeper into the draft, 20 percent of 200 is 40. Therefore, to jump an ADP of 200, the player needs to be taken by pick 160. Anything after 240 is considered a value pick in this scenario.

Here's how each league finish drafted using the three cutoffs:

Jump Picks

There's a notion successful teams know who they want and go get them. While that may be true for some, it's not reflected in the aggregate.

While having a feel for how often teams jump the ADP is useful, knowing what they do with these picks is more telling. Here's a table displaying P/L per jump pick at each finish. That is, it's not the total P/L, but the average P/L per jump pick made.

When venturing just a little from ADP, winners made solid picks. However, when they really jumped, they had less success. To be honest, this isn't actionable since it's just one year of data and thus can't be deemed a trend. The more relevant point is that sage jumping of a player can be profitable. Keep that in mind the next time you're chided for a reach.

Before presenting the value pick data, what does your gut say? Will they truly be values in the P/L sense of the word?

Value Picks

Spoiler alert: the lower finishing teams drafted more perceived values. Here's the P/L per finish, again exhibited per pick.

Surprise! What everyone lauds as a great pick was more than likely a poor selection. In the aggregate, only the top two teams in each league enjoyed a modicum of profit when a "value" fell in their lap, and that's only when the player fell at least 20 percent from their ADP. The next time you're drafting and have the opportunity to grab a guy sliding well past his ADP, make sure you aren't missing something; there's likely a reason he's available. You're probably in a sharp room.

Breaking the jump and value pick data by sections of the draft yields expected results. Big reaches occur more frequently in the first third of the draft, likely a combination of urgency to get your guy and smaller ranges in terms of number of picks needed to jump. Small jumps and value picks were also more prevalent early, but they remained present throughout the draft, albeit dwindling as the picks moved down the snake.

Summary and Conclusions

1. Profit is paramount, though there are indications roster construction and category management are crucial as well. These factors will be examined in upcoming studies.

2. Straight profit/loss determination is misleading as all but 11 of the 570 teams in the 2019 NFBC Main Event showed a loss on their drafted ledger. However, adjusting expectation to account for how each player performed relative to those drafted around him rendered a practical foundation for further investigation.

3. There's strong evidence you can indeed win the league early, as the average league champion lapped the field in teams of profit through the first eight rounds, sustaining their dominance throughout the festivities.

3. A side benefit of the adjusted P/L method is ranking players by performance relative to their draft slot.

4. While some adept drafters undoubtedly possess the ability to eschew the market and get their guys, the results show sticking closer to the vest is the prudent course of action. That said, when the champions planted their flag on an early reach, they were quite successful.

5. Be wary of a player falling past ADP. The odds are overwhelming he'll end the season with a negative return on investment. Big picture, only league winners were able to identify the legitimate "value picks".

6. Teams finishing in the bottom portion of the standings made fewer jump picks while grabbing more so-called value picks.

7. Winning teams demonstrated the best balance between trusting the market yet knowing when to stray in either direction. That is, there's a fine line between being a slave to ADP and knowing when to trust one's own analysis. Consistent jumping of the ADP may work for some, but the data strongly points towards selective aggressiveness. This may seem obvious, but many believe it's necessary to take chances to win. The data disagrees. A consistently solid approach is the winning formula.

Next time: How champions approached pitchers, catchers and multiple eligibility players