Run the Runs: Timing Picks in Fantasy Football Drafts
You know the feeling: you’re sitting with the 8th pick in the third round and considering two players at two different positions (for concreteness, imagine an RB and a WR). While you might prefer the RB, there could be a good chance that he will fall to you after the turn; you might be anticipating a ‘run on wide receivers’ as you get into the fourth, which would make getting your wideout unlikely.
These sorts of decisions are ubiquitous across drafts. Your greatest preparation lies in knowing your league, and paying attention to other picks; this is the best way to anticipate which manager is targeting which position. The Fantasy Footballers have even promoted a skeleton key, or a mock draft that you fill out entirely before your league actually gets together.
However, while the devil is in the details, there are broad trends in drafts that we can glean tips from. Specifically, I’ll be looking at real-life mock draft data from FantasyPros; I just downloaded the fifteen most recent drafts (12-team Half-PPR) at the time of this writing. Naturally, this is different from straight-up ADP because teams have different roster construction needs: a manager might reach for a WR, for example, because his roster is thin at that position. ADP is just an aggregate ranking that doesn’t take into account individual draft tendencies.
Anyways, Christian McCaffrey went 1.01 in every single case, which gives us a bit of confidence that these drafts were at least reasonable! Let’s get into the data.
While every draft is different, there is often a lot in common across leagues. To chart this, we can look pick-by-pick at each position (QB/RB/WR/TE) and chart the probability that the position is selected with that draft pick. For example, in this dataset, the first pick is used on a running back 100% of the time. The ‘smooth’ lines are just trend lines to give us a sense of the pattern; the original data is the faded, spiky lines in the foreground (naturally, it’s noisy). Vertical dotted lines separate the different rounds.
- The most defining characteristic is an RB-heavy first round. We’ve often seen how valuable it can be to get a top-tier, consistent, bell-cow running back, so this pattern makes sense. Still, this is pretty significant: naively (not actually considering how the specific draft is going) an RB is the most likely position at every single pick in the first round.
- WRs begin to rise in probability towards the end of the first (where players like Davante Adams and Tyreek Hill will often be picked), and cross RBs as the most likely position to be selected – a lead they don’t relinquish for the rest of this chart – in the mid-second round.
- Quarterbacks gradually begin to rise in probability as the draft progresses. The TE chart is interesting: a small hump in the first few rounds for the likes of Travis Kelce, Darren Waller, and George Kittle, a valley in Rounds 3-4, and a slope upward into the later rounds. This is all intuitive given the Fantasy Footballers’ standard advice: if you can’t get an elite TE early, punt until later rounds for another option.
A simpler chart just gives us the average number of positional players drafted in each round. For example, in the first round we see about 8-or-9 RBs go off the board, about two WRs (likely Davante Adams and Tyreek Hill), and one TE (Travis Kelce). The amount of RBs and WRs intersects in Round 2 (5 of each) and WRs lead the positions from there. By Round 6, all of the positions are being selected at close to the same rate (RBs and WRs are being picked far less than early/mid rounds).
As referred to in the introduction, a common draft strategy is to try and ‘play the runs’. That is, you don’t want to be stuck holding the bag, forced to pick a running back after a full slate of RBs went off the board.
We can quantify this effect by considering the ADP of players picked at the end of a run; here, a run is just defined as players from the same position picked consecutively. If we look at the average draft pick of players picked at the end of a run – the last pick in a positional run – vs. their ADP, or overall average draft position, we see a definite negative relationship with WRs:
The RB chart, which doesn’t show much of a relationship, is likely affected by that common ‘first-round RB run’, where nearly every RB is picked at around their ADP; variance starts to pick up as we get later into drafts, which is where WRs are more common.
Anyways, the takeaway for wideouts (and likely running backs outside of the first round) is to try to avoid being in a situation that forces you to pick at the end of a run. Specifically, if you grab a player at the end of a run, you are on average picking them 0.25 spots earlier than they are usually picked; this is a result for all players, not just wide receivers. That is, you are paying a higher price for players than you would normally pay!
It can be pretty valuable, then, to have a general sense of when a run is coming. Again, a run is just defined as consecutive picks where the same position is picked in a row.
Here’s a chart that gives us the length of a run (y-axis) combined with where the run started (i.e., the pick number) for RBs and WRs:
- You can clearly see the ‘first-round RB run’ pattern here: at a ‘Run Start’ of 1 (the first pick), we have RB runs (y-axis) varying in length from three in a row to eight in a row. There actually was one draft that saw 14 RBs go off the board to kick things off (Josh Jacobs was the last pick of that run, which is suspect, I would certainly prefer rostering Kelce) but I excluded it here for visual purposes.
- RB runs start to decrease in length as the draft continues onward; after Round 2 (pick 24) there was only one instance of an RB run that was longer than two picks! This is in contrast to WR runs, which start to pick up and get longer as we get into later rounds. This agrees with what we saw before: running backs fly off the board to start, while WRs start to become the more popular position at about Round 2.
To dive even deeper, we can build a regression model that tries to predict the number of picks at each position for a given round. Our inputs will be simple: the number of picks at that position made in the entire draft so far, the number of picks at that position made in the previous round, and the actual round that we are picking in. Let’s explore the results:
- Round: there is, as to be expected, a strong ’round effect’. With each passing round, we expect 0.72 fewer RBs, .80 fewer WRs, .70 more QBs, and .48 more TEs to be picked in a round; all of these coefficients are highly significant. This makes sense: the skill positions of RB and WR go early (RBs especially early) but start to fade by Round 6. At this point, the less important positions of QB and TE are being picked more.
- Runs: for wideouts, every WR pick in the previous round means there is an additional 0.5 WR picks expected in the current round (this coefficient is significant). Intuitively, this means that if a bunch of WRs went in Round 4, we expect a bunch of WRs to go in Round 5 as managers scramble to fill their roster while solid-tier WR options are still available. We see the same effect for RBs, albeit with a smaller coefficient (0.18 more expected RB picks in the current round for every RB pick in the previous round), but it’s unfortunately not very significant, likely thanks to the ‘first-round effect’ (tons of RBs go off in the first round, far more than any other round). QBs are interesting: the effect is the opposite, and every QB pick in the previous round decreases the expected QB picks in this round by 0.42 picks. This is likely because the tiers are so thin at quarterback: there are a couple of studs worth taking early (Patrick Mahomes, Kyler Murray, etc.) and, if these options get picked, you might punt on the position until later.
- Positional Depth: all of these positions have an inverse relationship to the average rate that they’ve been picked in the draft so far. In English, that means that, if lots of RBs have been picked in the entire draft so far, fewer RBs are expected to be picked in the current round. The intuition here is being ‘post-run’: once many running backs have been taken, managers might focus on other positions with higher-tier options for a bit until the tier of running back available better matches the necessary draft capital. Interestingly, this effect is largest (and most significant) with tight ends: if a lot of tight ends have gone in the entire draft so far, significantly fewer tight ends are expected in the current round. This is similar to the QB effect we discussed in the earlier point: if the top options of Kelce, Waller, and Kittle have gone, you’re likely to wait for a late-round flier.
What did we learn?
Let’s close where we opened. The most important preparation you can do when learning about draft patterns is to know your league and, on draft day, pay attention to who picked whom. However, there are some general trends that you can use to your advantage:
- It’s important to know common draft dynamics. Naturally, the biggest ‘runs’ happen in the first round: drafts usually start with multiple RBs going off the board. In turn, RBs peak in the 1st, WRs peak in the 3rd/4th rounds. QBs and TEs climb all the way through the 6th, unsurprisingly (we only looked closely at the top-6 rounds here because they are the most important and behavior gets sporadic after that).
- All else equal, it doesn’t behoove you to pick the player at the end of a positional run: intuitively, this will mean you are usually reaching for that player. To ‘avoid’ these runs, remember that, especially for WRs, if there has been a recent uptick in picks for that position (i.e., the last round) you can expect a ‘run’ to continue (i.e., into the current round) which could leave someone holding the bag. On the flip side, if, over the entire draft, a position has already been ‘run on’ and is currently sparse, you can expect managers to turn elsewhere for the time being.
- All of these dynamics hold strongest in the early and mid rounds; in the 6th round and beyond, after the ‘team foundation’ has been built, it’s much more of a free-for-all!
Did I miss anything? Curious about other draft patterns? Message me on Twitter.