This is the 138th installment of The Labyrinthian, a series dedicated to exploring random fields of knowledge in order to give you unordinary theoretical, philosophical, strategic, and/or often rambling guidance on daily fantasy sports. Consult the introductory piece to the series for further explanation.
This ultimately is a piece about how to make non-horrible trends. First, though, I’m going to talk about something else . . . right after I talk about something else.
This is the Last Time I’ll Mention Stephen Hill
I started out in the industry by writing at RotoViz. I’m still tight with those guys — #NASCAR — and in my opinion Fantasy Douche is a giant.
At one point, though, I wasn’t thrilled with him. Still basking in the glorious success of my 3,000-word two-part debut (which detailed why T.Y. Hilton is the greatest wide receiver of all time and was read by literally dozens of people), I submitted to him my second RotoViz piece: A treatise breaking down just how historically disappointing Stephen Hill was as a rookie, especially after he turned six targets into five receptions, 89 yards, and two touchdowns in his first NFL game.
Douche rejected my piece. He said that it was too long for the subject matter and that I was using skewed samples to make my case. He was right on both counts. At the time I thought he was being unreasonable and stubborn, but I’m eternally grateful now that he put his editorial foot down when he did. Eventually I revised and resubmitted the piece. It was published.
Although I still write articles that are way too long — you can take the boy out of the Ph.D. program in Renaissance literature, but you can’t take the yada yada yada — I’m happy to say that from that experience I learned two important lessons.
- Treat data in an unbiased way.
- Don’t write an article about Stephen F*cking Hill.
Just to be clear: This is not an article about Hill. This is a section of an article about an article about Hill. Those two are in no way similar.
The Daily Fantasy Edge, Episode 131
Earlier today I appeared on the 131st episode of The Daily Fantasy Edge podcast (a.k.a. The Hedge). If you’d like to listen to the episode, here you go:
I talked with GOAT host Adam Levitan about the 2017 rookie running backs and their DFS potential for the coming season (which, by the way, is about four months away, so this is an incredibly timely and actionable resource).
A couple of these runners I highlight in my two rookie breakdowns:
- Five Fourth-Round NFL Rookies Who Could Crush in 2017
- Five Late-Round NFL Rookies Who Could Crush in 2017
Unrelatedly, FantasyLabs is looking to hire someone full time to title our articles. The Editor-in-Chief who does it now has no knack for it. (Side note: That was a joke. Don’t spam Jonathan Bales with emails inquiring about “the position mentioned in the article about Stephen F*cking Hill.”)
In my conversation with Levitan, he asked me how I got into NFL prospect evaluation, which is what I spend all of my time doing when I’m not writing and editing articles, listening to podcasts, reading books, pretending to work out, trolling the ignorati on Twitter, and skin-to-furing my dog. And watching Star Wars movies on repeat on May the Fourth and Revenge of the Fifth. Obviously.
Basically, I started to get into NFL prospect evaluation a decade ago, when I decided that I was going to dominate my dynasty league. By the way, I just won my third championship in five years, it’s no big deal, I’m totally humble about it, suck it losers.
How I evaluate players is (I think) sophisticated in a straightforward way. It’s not unlike the process we use at FantasyLabs to create ceiling and floor projections. I collect data for a variety of factors — college production (raw and market share), college competition, age, experience, size, speed, agility, and explosiveness — and then I create cohorts of past prospects who collectively were comparable to the players under consideration now. Naturally, after players are selected by teams I add draft position to the equation and update my assumptions like a good Bayesian. Using the cohorts of comparable players, I then create a range of potential outcomes for each prospect.
I don’t weight all factors equally across positions. In fact, I don’t even weight all factors equally within positions, since for different types of players some factors matter more than they do for others. Even though I try to be as scientific as possible (for a self-trained hack), I understand that NFL prospect evaluation at this point still requires the creativity of art. The process moves forward by trial and error. It must be flexible.
I mention all of this because, in a general sense, the way I evaluate NFL prospects is similar to the way I research with our Trends tool.
How to Build FantasyLabs Trends That Suck Less
This article is entitled “How to Build Dominant FantasyLabs Trends” — but let’s be honest: If I had titled the article “How to Build FantasyLabs Trends That Suck Less,” would you be reading this sentence right now?
Also, I can’t promise that I can show you how to build trends that dominate. Who do I look like? — Peter Jennings? I am, though, (98.7 percent-ish) sure your trends will suck less after you read this piece.
I have no data to back up my next statement, but my sense is that one of the primary factors separating positive expected value subscribers from those less satisfied with their DFS experiences is their superior ability to research with our Trends tool, which (I believe) is the most valuable resource of its kind in the industry.
If you’re someone who wants more out of DFS, let me tell you the main problem I tend to see: You subscribe to Labs and then start experimenting with the Trends tool without thinking much about what you’re doing. You start researching for slates with our Player Models and see some players who really seem to have an unusual confluence of factors. Maybe a pitcher is $5,000 with a high K Prediction. Maybe a batter has elite Statcast data but a horrible Plus/Minus and Consistency Rating over the last month. It could really be anything — but something about a player catches your eye.
And then you go to the Trends tool and attempt to answer these questions:
- Have there ever been any other players with this extreme mix of characteristics?
- If so, how have they done?
- How can I skew this trend to give this guy a comparable cohort of past players with the highest Plus/Minus and Consistency and Upside Ratings and lowest ownership percentage possible?
We’ve all been there. Seriously, you know you do it.
How NOT to Build a Trend
Let’s say that on a team implied for 4.5 runs there’s a No. 6 hitter who costs $3,500 on DraftKings and has a high recent batted ball distance of 250 feet. How do you build this trend? You go to the Trends tool, you see that on average the 128,779 batters in our database have a -0.01 DraftKings Plus/Minus, and you say, “This guy is going to crush that number.” And then you start adding filters:
- Runs: You set 4.5 runs as a minimum with an uncapped ceiling. The Plus/Minus jumps to +0.45.
- Lineup Order: You select batters hitting first through sixth. Now the Plus/Minus is +0.73.
- Salary: You set $3,500 as the maximum with no minimum floor. The Plus/Minus climbs to +1.18.
- Recent Batted Ball Distance: You set 250 feet as a minimum with an uncapped ceiling. This is the money shot.
Once you regain consciousness, you check the numbers again just to make sure the trend is actually correct. Yep, it is:
And then you squeeze this player into as many rosters as you can with our Lineup Builder — because a player with a 29 percent Upside Rating must be used a lot, right? — and then . . . you wonder why you’re not getting the results you want.
There are two main reasons:
- You’re assuming that players who match for such trends will hit or exceed the average. You’re not taking into account the certainty that within the cohorts of past players are many performances that fall short of salary-based expectations.
- You’re building unrepresentative threshold trends. The cohorts of past players aren’t highly comparable to the players around whom you’re constructing these trends in the first place.
Basically, when you build these trends you’re engaging in intellectual dishonesty. You’re asleep and dreaming on a bed of bullsh*t.
How to Build a Trend
Don’t delude yourself by building trends that structurally make your target player an outlier. Instead, construct balanced trends with your player at the center of the filters. Let’s return to the No. 6 hitter with an implied team total of 4.5 runs, salary of $3,500, and recent batted ball distance of 250 feet. Here’s one possible and reasonable way to build a trend around that hypothetical player:
- Runs: 4.0 to 5.0 runs
- Lineup Order: 5 to 7
- Salary: $3,000 to $4,000
- Recent Batted Ball Distance: 225 to 275 feet
Here’s the trend:
Is this No. 6 hitter valuable? Yes. Is he one of the rarest No. 6 hitters ever to exist in the DFS universe? Probably not. He’s just a No. 6 hitter on a hot streak with a good matchup and pretty standard salary.
You don’t assume that every minorly attractive person you meet is marriage material, do you?
Thinking About Thinking
DFS is less about sports and more about the process of thinking. That’s why it’s important to think about how you think and why so many of my Labyrinthians are about thinking (about thinking).
There are a lot of people who know a lot about sports, but they underperform as DFS players because they make basic mental mistakes, like contextualizing player data in unbalanced ways.
Don’t make that mistake anymore.
——
The Labyrinthian: 2017.43, 138
Previous installments can be accessed via my author page or the series archive.
