Week 1 is in the books, and clearly it’s time to sell Drake London, buy Khalif Raymond, and get whatever you can for Ja’Marr Chase before his value bottoms out completely.
Of course I’m not serious about any of those takes, but they do raise an interesting question. How do you actually *act* on all the new information we have to process week after week as the NFL season progresses? How do you absorb all the new data quickly, and get value from your league mates without being reactionary?
Among all the dynamics that make dynasty fantasy football such a rich and immersive strategy game, the most challenging is the dilemma of how to get value before the consensus catches on, but without shooting yourself in the foot. If you never, ever make a move until the consensus agrees with you, you might miss out on the best opportunities. However, overreacting is usually far worse than underreacting.
In this two-part series, I will first discuss how to process new information and update your dynasty values without being overly reactionary. Then next week I’ll discuss how to trade on value changes responsibly yet decisively.
The Over-Information Age
After every single week of NFL football, we have to process an overwhelming amount of new information. Not only does every game’s box score contain the potential to move dynasty values, but we live in a golden age of analysts and content creators who produce a never-ending stream of metrics, models, and splits. All of these tidbits may or may not matter more than the box scores themselves, and as owners we have to put the pieces together while ignoring the fluff.
Today, the goal is to organize our thoughts into a system to productively react to each subsequent headline and datum as the season progresses.
Results vs Underlying Data
Nearly all experienced dynasty players are familiar with these concepts, so I won’t belabor the point.
Broadly, the new information produced by each week of NFL football can be categorized as either a result, meaning information that appears in box scores and that directly contributes to fantasy scoring, or as underlying data that hints at a player or team’s process. By “process,” I mean the underlying player performance and context within a player’s control that strips out the luck and variance of results.
Every piece of information, whether relating to results or process, operates on a spectrum of relevance and usefulness. Results are highly relevant, but they’re often not useful for a couple of reasons. Underlying data is often irrelevant, but when it’s not, it can be extremely useful.
Results
On its face, it can be somewhat paradoxical that the most relevant data is often the least useful. After all, among all the dozens of models and metrics that the fantasy community’s brightest minds publish year after year, the single metric with the highest R2 (https://en.wikipedia.org/wiki/Coefficient_of_determination) with a player’s yearly fantasy points is, you guessed it, last year’s fantasy points.
Relevance notwithstanding, box score results and fantasy points are highly likely to be priced in by your league mates, even if the other owners are less active than you. By “priced in”, I mean your league mates know the same information as you and are using it the same way you are. This means if you’re updating your fantasy values based solely on the results, you’re already behind the curve.
While this may sound obvious so far, I cannot stress the next point enough. Results are shaping your opinions, biases, and emotional responses all the time, no matter how enlightened you think you are about the underlying data. We’re humans, and we never, ever will be able to fully ignore the data that’s staring us in the face.
You will finish reading this entire article, and you will still get the itch to acquire a player who is priced at his ceiling based on a recent hot streak, or add an underperforming player to your “Never Again” list. I’m not even saying that’s a bad thing, just that you should acknowledge this and notice when you’re doing it, so you can judge if it’s helping you or not.
Underlying Data
Moving past the surface-level results, every single box score can be distilled into a menagerie of metrics and splits. Referring back to the spectrum of relevance and usefulness, most of these are irrelevant.
While previously I warned you not to reactively chase box scores, you also cannot move your player valuations around every single time you see a tweet or GIF thread that’s bullish on a specific player.
An analyst can build a strong-looking case for literally any relevant fantasy football player if they cherry-pick the metric that makes them look the best, and particularly if they over-emphasize how important that metric is.
Some irrelevant data is easy to spot, such as most TikTok content creators’ cherry-picked rankings lists. However, some is more insidious, or it’s promoted by a large website in an official-looking format. For example, while proving this is an article for another day, I’m becoming increasingly convinced the slew of “separation” metrics all over fantasy Twitter have made us collectively less sharp as a community.
To figure out what to follow and what to ignore, you should ask yourself a couple of questions.
Does This Metric Have Evidence For it's Relevance, Or a Track Record?
Metrics such as yards per route run, first downs per route run, target rate, red zone targets, running back opportunities, and expected fantasy points are battle-tested and have been demonstrated to predict future fantasy performance.
If Not, Did the Author Make a Compelling Case Demonstrating This Metric Actually Matters?
If this metric isn’t one of the classic powerhouses, did the author actually prove to you that we should care about this metric? For example, when Ryan Heath, one of the best in the business, first wrote about first downs per route run, he made a compelling case for why it actually improves on yards per route run, and backed it up with robust findings.
Of note: Content creators showing that some combination of metrics puts a player in elite company is extremely unreliable. You can always tweak the thresholds until some mediocre player fits in with far superior players.
Is This Metric Measuring Something we Already Measure Better, Or is it More Convoluted Than Another Approach?
For example, separation metrics are at least correlated to the underlying skill level of a wide receiver, but in my opinion they’re trying to accomplish almost exactly the same thing as yards per route run. Granted, in theory, separation metrics can measure if somebody gets open but is not getting targeted, but to accomplish this they’re much more convoluted and less transparent.
For example, separation metrics are often just measuring what types of routes a player is asked to run and what coverages they’re facing. But not everybody knows that because separation metrics tend to be proprietary and not overly transparent. In the meantime, as far as I know, nobody has yet proven that separation metrics add predictive power to the existing metrics we already have.
Regression vs Momentum
Now that we’ve had our disclaimer on not blindly following each new metric that fantasy Twitter churns out, we can discuss how to incorporate these metrics into our decision-making.
Whenever the underlying metrics tell a different story than the results, they’re pointing to either regression or momentum. Regression means that the results we’re seeing on Sundays are not sustainable, usually due to luck or contextual factors that are subject to change.
Momentum means a signal is likely to increase in strength if you spotted it soon enough. A few classic examples include rookie snap shares, post-bye bumps, and players ramping back up after injuries.
Generally, trading opportunities arise when the market is in denial about the likelihood of regression, or is too slow to price in growing signs of momentum. In general, I would argue the dynasty is very familiar with the concept of regression, but to the point that it can become a blind spot.
Young players who perform better than we expected are often dismissed because everybody is so familiar with the concept of regression, but in practice unsustainable efficiency can create long-term changes in playing time and priority in an offense.
Take De’Von Achane’s rookie season, which was clearly unsustainable from an efficiency standpoint. Achane sellers were missing the point, which is that he showed his team he was an ascending stud, and that he deserved the playing time and role to sustainably produce.
Pattern Recognition
Certain signals are notoriously noisy, while others are reliably sticky. One of the biggest perils for dynasty owners is acting on emotional responses to noisy signals. When adjusting your player values, consider if your data fits into a pattern that is historically unreliable.
A major example is when players have Week 1 performances that contradict their track records. Wide receiver performances in Week 1 are extremely noisy, with Sammy Watkins, Quentin Johnston, and more recently Khalif Raymond looking like high-end players for a week or two before fading into obscurity.
Another example of pattern recognition is discounting preseason performances. You can sometimes glean a real signal from playing time trends, for example, a rookie playing dangerously late into preseason games (hello, Kaleb Johnson), but generally these can be safely ignored.
I’m old enough to remember Ja’Marr Chase having an awful preseason his rookie year, Kenny Pickett looking like an elite passer, and everything in between.
After Week 1, if you’re about to change your opinion on a dynasty asset in a way that contradicts their track record, the underlying data better be strongly on your side. Below are some reliable signals.
Reliable Week 1 Signals
Snap rate, and a useful resource for it. Naturally for rookies and players coming off injuries, you can generally expect these to increase. Take George Kittle, Malik Nabers, and Tucker Kraft, for example. They were involved in their offenses in much lower snap shares than they’ll be seeing by midseason. https://www.footballguys.com/stats/snap-counts/teams
Route participation - this can be extremely important for tight ends and running backs, who may not necessarily run routes every pass play, and who may sub in and out in different personnel groupings. I usually have to search these on Twitter.
Yards per route run, targets per route run. Found here https://sumersports.com/players/wide-receiver/
Opportunity share - useful for running backs. You can usually look this up or just take a peek at a box score
Humility
Humans are really bad at predicting exactly how complex situations will unfold. Compounding this issue is that we tend to conflate possibility with probability. Just because we can vividly picture a dynasty player having a specific fantasy outcome (e.g. how easy it was for us all to picture TreVeyon Henderson dusting Rhamondre Stevenson last year, or becoming a pass catching demon), doesn’t necessarily mean it’s the most likely outcome to happen.
At an emotional level, we’re extremely likely to overestimate the probability of a specific event once we’ve imagined it. To fight against this bias, we need to ask ourselves a few key questions.
How Grounded Was My Previous Opinion?
How Convoluted Is My New Prediction?
You can be much bolder about updating your previous opinion if it was based on a hunch, and you should be very conservative if you have a hunch contradicting an established pattern. For example, if your valuation for a specific player is based on their past results and supported by their current underlying data, you need a high burden of proof to act against that trend.
On the other hand, if your original opinion was based primarily on guesswork, you can update very quickly once the data starts rolling in. One major example is with rookies. As a community, we tend to be overconfident in our original rookie valuations despite coaches showing us what they really think of these players once games start. Going back to the idea of pattern recognition, this also doesn’t mean you should jettison any rookie who doesn’t get playing time immediately. The more important signal is that their responsibilities increase over the course of the year.
Another example would be offseason hype for impending breakouts. These are usually an educated guess at best and a hunch at worst, and if the snap rates and route participation sharply contradict that prediction once the season starts, you’d better update your valuations quickly.
Related to this topic is the idea of simple vs convoluted predictions. Fantasy seasons aren’t binary good or bad. They’re an infinite number of forking paths representing all the possible outcomes. If you need an extremely specific scenario to play out for your valuation to be correct, you’re fighting an uphill battle.
Instead, if all you require is a single specific variable to fall into place, that’s a highly parsimonious prediction that can justify an aggressive valuation. To make this more concrete, my favorite example of a parsimonious prediction is finding already efficient players with paths to increased playing time.
Examples of convoluted predictions include:
Highly specific roles (Sean Payton’s “joker”, or “the Amon-Ra role”, and so on).
A new playcaller will be a major improvement. This can matter a lot, but as outsiders, we’re terrible at predicting which specific playcallers will catalyze which specific player’s breakout.
Players being earlier in the progression or receiving more layup targets
Players improving drastically
These can all end up being correct, but they’re complex outcomes that humans are bad at nailing, and whose probabilities they’re likely to overestimate.
Consider Counterfactuals
Another key part of humility with your player valuations is thinking about the most likely ways you can be wrong. Suppose you're trying to buy low on a player, but you’re wrong and their value doesn’t increase. Can you live with that? For example, maybe that player would still be solid depth for you, or has value insulation.
Another example of considering counterfactuals is when drafting rookies. Generally, we aren’t as good at scouting as we think we are. This means dynasty players need to be very cautious about making expensive trade-ups in the draft. Additionally, dynasty owners should consider occasionally drafting players they don’t personally believe in, so long as that player is set up to smash if they’re better than we realize. One recent example of this would be Jaxson Dart going in the late first and early second of Superflex drafts despite his promising scramble rate.
Conclusion
We’re finally ready to face the deluge of Week 1 takeaways with clarity and purpose. Whether you’re trying to buy low, sell high, or abandon ship on a distressed asset, you now have a framework for keeping overreactions in check.
Consider whether the underlying data supports or contradicts the results, what it means for regression or momentum, and whether player price movements fall into established patterns from past dynasty seasons. Then, before you act on these value changes, take a moment to think about whether your waiver pickups, trade offers, and lineups are backed by evidence. Or, do they require you to nail a convoluted guess in order to pay off?
Next week I’ll talk in more detail about specific approaches to trading based on new information, but until then, it’s on to Week 2.