Playing the Long Game: Tracking Fantasy Value Over Time
A key tenet of Dynasty and Keeper fantasy formats is gauging player value over time. Great players generally stay great, although nothing lasts forever: even the most dominant fantasy players end up on the bench eventually.
Todd Gurley II is a prime example. After a 10-TD rookie season, he rocketed up to the first-round ADP (Average Draft Position) echelon. He was an early-round pick for years, ultimately peaking as the near-consensus 1.01 in 2018 drafts. Knee troubles and a stint with the Falcons, though, have seen Gurley’s draft capital fall to the tenth round in 2021.
Managers, then, have to constantly consider how value decays (or, hopefully, appreciates) over time. Is it worthwhile to trade an RB2 for a first-round pick next year? How about trading a first this year for a second and a third next year? How much value can you expect from a stud WR1 four years from now?
To study this question, let’s analyze the last seven years of fantasy redraft ADP data at the start of a season (PPR scoring), available via The Fantasy Football Calculator and FantasyPros. Specifically, I can aggregate this data and calculate the transition probabilities of player ADP over the years.
For example, imagine looking at all of the second-round picks across the dataset. What is the probability that a player worth a second-round pick today (according to ADP) is still worth a second-round pick a year from now? What is the probability that they drop to the third round, or even rise to the first-round status?
The matrix below gives the layout of these probabilities:
This is a lot to consider. The y-axis tells us the round (according to ADP) of a player at a given point in time, and the x-axis tells us the round (according to ADP) of a player the next year. The number in the box tells us the probability of making this transition. For example, a player with a third-round ADP has, on average, a 15% chance of being a fifth-round ADP player in the next year (box in the third row, the fifth column).
Right off the bat, there are a couple of interesting takeaways:
- Aside from rounds 10+, first-round talents are the easiest to predict over the next year. According to this data, a first-rounder today has a 53% chance of still being a first-rounder next year and an 88% chance of being in the top three rounds next year. This makes sense: known fantasy studs are extremely valuable because of their dominance over time. This should inform any of your trades that include first-round talent/picks: that level of draft capital is worth significant value.
- The third and fourth rounds are ‘call your shot’ rounds. Players drafted in these rounds have a similar chance of landing as high as the second round or as low as the sixth round next year. It might be surprising how spread out this distribution of outcomes is over just one year: third-round talent/picks might be more of a gamble than you realize.
- Low picks usually stay low: players drafted in the 10th round or later (or that go undrafted) stay there with 88% probability. This is partly influenced by kickers, defenses, and many quarterbacks, as well as the fact that multiple rounds are bucketed into this category. Still, it’s important to stay realistic when targeting late-round sleepers!
Now, I’m modeling a player’s ADP over time via this ‘transition matrix’ above; the technical term is that I am constructing a ‘Markov Chain’*. This essentially means that the player’s ADP ‘bounces around’ over time according to the probabilities established in the above matrix.
What’s interesting about a Markov Chain is that I can easily look many steps forward. I do this by taking the above transition matrix to a power. For example, I can take the matrix to the power of two (multiply the matrix by itself, not square every number) to get the two-year transition probabilities:
This chart tells us, for example, that the probability of a first-rounder still being a first-rounder in two years is 34% (and the probability of a first-rounder being a second-rounder in two years is 18%, etc.).
Intuitively, much of the ‘density’ in the chart above has shifted to the right: the longer horizon I use, the greater chance that highly drafted players will decay in value. For example, the probability that a first-rounder goes in the tenth round or later (or undrafted) in two years is 12%, up from 3% in one year.
I see a similar phenomenon when looking across three and five years:
In both of these cases, the outlook for top picks gets darker. If we’re looking five years ahead, a first-rounder only has a 13% chance of still being a first-rounder and has a 45% chance of being a bottom-dweller (10th round or later, or undrafted). Given the shelf life of running backs – who usually dominate the first round – this is unsurprising.
Altogether, you can use these charts to help you value potential trade options. Even though Dynasty and Keeper managers will find this the most useful, it still can help managers in redraft leagues: what’s the probability that your first-round stud will actually perform as expected? What sort of variance can you expect out of your third-rounder?
I’ll be following this up with a transition breakdown by position. Stay tuned!
Note: Some readers may note that this process might not be Markov, since conditioning on the filtration up to time t – 1 might not be sufficient (i.e., a player that has been rising to the 3rd round is different than a player who has just fallen to the third round; a big lurking variable here is age). Ultimately, though, these dependencies are weak enough to allow us to use this model safely (within reason).
Really interesting – thanks for posting.
Would be great to see this by skill position if possible?