Looking For Market Inefficiencies: TE Return On Investment

No matter which type of person you are. No matter which type of fantasy owner profile you fit into. No matter what, you're looking for market inefficiencies. Why pay extra for something that ultimately will yield the same results?

The question you might be asking yourself, though, is the one regarding how and where to find market inefficiencies to take advantage of? And that's the hard one to answer. There must be one, that's for sure. The problem, of course, is finding it.

We can look at correlations between different stats, try to find what goes cheap for its production. One way of measuring how good our draft was in terms of price/production, that is, value or what we come to call Return On Investment (ROI), is to just take two data points in consideration: ADP (where we draft players) and season-end rank (where the player ends the year ranked at in fantasy points). Just using those two values we can easily calculate how valuable our picks were. The concept of ROI got me thinking about how one year's values relate to the next one, and how they impact future ADPs in fantasy football.

Return On Investment Values

When it comes to ROI, let's keep things simple. Take a player's ADP entering the season and divide it by his season-end final ranking among all players in the league. That's because every player is available in drafts and therefore every player should be taken into consideration for the ranking.

Any player with a ROI at or over 1.0 yielded a positive value and turned into a valuable play. Any player with a ROI under 1.0 finished the year in a position lower than where they were drafted. Although there is virtually no limit in how large (positively or negative) a ROI mark can be, we can assume the lowest value is 0.001 (ADP 1, rank over 500) and the largest 500 (ADP 500, rank 1).

For this research, though, we're using a dataset containing every tight end season from 2000 to 2018 (474 in total), with ROIs ranging from 0.07 to 11.55 (in both cases Rob Gronkowski). It only includes players from which their ADP in years N and N+1 and their ROI marks for years N and N+1 too.

This is how ADP and ROI correlate in the same season N.

The correlation is almost nonexistent with an R-Squared value of 0.03 It makes sense, considering at the point of drafting we don't really know whether a player will be good or not. We make our best guesses, and are mostly right when it comes to ADP and final rank (the R-Squared there goes up to 0.17), but the relationship between draft position and ROI is totally random.

Year-to-Year ROI Stability

If the relation between ADP and ROI is barely existent, the relationship of ROI marks from one year to the next one should be expected to be absolutely random too...

...and that is precisely the case. The R-Squared value here drops even more down to 0.0004. Don't give this relation even a split of a second of your free time, as you'd be basically throwing it away.

My first takeaway from this was that, if ROI isn't predictable at all, it must be related to the fact that a player exceeding his level in year N would translate into a higher ADP (meaning a more expensive draft position) in year N+1, thus lowering his potential ROI no matter what (the more expensive the player, the lower his ROI becomes). I had no real knowledge of the correctness of that thought, but I had the data to try and back it up.

Year-to-Year ROI to ADP Correlation

What I wanted to test was a pretty simple idea: if a player exceeds his value in year N, we can assume he will become more expensive in year N+1 and therefore he would be less valuable in terms of potential ROI. Think of Eric Ebron. He entered the 2018 season with an ADP of 168.8, yet he finished ranked as the 50th-best player that year and the TE4. Obviously, his ADP in 2019 went all the way up to 95.0, almost 75 spots more expensive! No wonder his ROI in 2018 was a great 3.3 but it dropped to a horrific 0.5 (making him an overvalued/overpaid player) in 2019.

This is how the relationship between the season-end rank in year N and the ADP in year N+1 has gone through the last couple of decades.

There is, in fact, a positive correlation up to an R-Squared value of 0.04 between both variables. The better a tight end has done in year N, the more expensive he has gone the following season.

With all we know by now, we should expect a similar relationship to exist between the ROI in year N and the ADP in year N+1. We should assume a player beating his expected value would be drafted higher the next year.

Absolutely correct. The correlation here is positive again, yet it doubles the strength of the last one with an R-Squared value of 0.25 this time.

This means one thing: fantasy owners focus more heavily on final raw results rather than the value returned by the players given the paid price.

While that is nothing unreasonable (we're bumping up the prices and paying more for the best performers), it is not the best way to tackle the market. It is an inefficiency. It's a good strategy, but it is not the smartest one nor the one most beneficial. That's why there is still a window there to go grab the best possible values that are not yet inflated.

What History Tells Of Changes In ADP And ROI

Since the 2000 season, and looking only at tight end season from players of whom we know their year-to-year changes in ADP, Rank and ROI, this is how the numbers look like:

• 68 players became more expensive while improving their ROI (acceptable investment)
• 151 players became more expensive while lowering their ROI (worst investment)
• 100 players became cheaper while lowering their ROI (acceptable investment)
• 151 players became cheaper while improving their ROI (best investment)

In percentages, we can say that 35.8 percent of players remained in the balance, 32.1 percent became worse plays from year N to year N+1, and an equal 32.1 percent became better plays. Those are three very evenly split numbers, but if we add together the first and the last ones we get to 67.9 percent of players at least retaining their ROI values from one season to the next one.

What we should try to identify are the commonalities among those in the remaining group of players in order to try and avoid them. I tried to find some similar numbers and traits repeating themselves in their profiles to get to a sound conclusion.

Avoiding ROI-Fallers

The 32.1 percent of players becoming worse plays from year N to year N+1 (that is, more expensive in terms of ADP while providing worst ROI-marks at the end of the season) make for 416 players in my data set ranging from 2000 to 2018.

There is a boatload of data to unpack there, so let's go step by step.

• The majority of players were at or under 26 years of age.
• Virtually "every" player came from playing a full 16-game season.
• The vast majority of players logged between 60 and 95 targets.
• Most of the players logged between 40 and 75 receptions.
• The greatest number of players fell in between 465 and 830 receiving yards by the end of the season.
• Most players scored between three and six touchdowns.
• The peak on PPG was in the 7.5-to-13 clip, with most players finishing the year averaging between 10.5 and 13.

Here are the players from 2018 that would have fit or gotten closer to that profile at the year's end, and how they did in 2019.

Three of the four players highlighted had valuable seasons in 2018 in terms of ROI except for Evan Engram (0.5), but all of them had worse ROI marks in 2019 than they did the year before (ranging from drops of -0.1 with Engram to -2.8 with Ebron). Only Austin Hooper improved his overall ranking in PPR leagues from 94th to 81th, but given his price's (ADP) rise from 177.3 in 2018 to 111.3 in 2019 the ROI was lower (going from 1.9 to 1.4).

Finding ROI-Risers

The same process can be followed to try to identify traits present in all of the historical ROI-Risers in order to find what has repeated over the years in their profiles to take advantage of it going forward. This is how all of the players in the data set that became cheaper but better ROI-values are distributed in different stats.

And some of the shared similarities:

• The majority of players were between 24 and 27 years of age.
• Most players came from playing between 14 and 16 games.
• The vast majority of players logged between 40 and 75 targets.
• Most of the players logged between 25 and 40 receptions.
• The greatest number of players fell in between 150 and 445 receiving yards by the end of the season.
• Most players scored between 0 and three touchdowns.
• The peak on PPG was in the 5-9 clip, with most players finishing the year averaging between seven and nine.

Here are the players from 2018 that would have fit that profile at the year's end, and how they did in 2019.

Of the four players found, all but Jesse James improved or remained on the same ROI marks. Geoff Swaim, though, wasn't very valuable in any case finishing the year with a paltry 368th overall rank. Blake Jarwin and Gerald Everett were better plays and usable in deeper leagues, definitely worth investing in late in drafts or after being acquired as free agents in 2019 after going under the radar given their 2018 seasons.

Potentially Great ROI-Plays for the 2020 Season

Now that we have identified stats that fit the model for both good and bad "next-year ROIs", we can try and apply it to the current season trying to take advantage of our knowledge to build the best possible roster in 2020. Here are some 2019 players that fit the profile of the average ROI-Riser.

Here is a little breakdown of the group of players the model spit out:

• Blake Jarwin is an RFA this offseason, but Dallas has offered him an extension and plans to keep him. With Jason Witten in Las Vegas, it will be Jarwin taking the most opportunities in the TE position. He was a value in 2019 already, and he very well could keep that up next season.
• Jonnu Smith will be the starter for the Titans next year and will have TE-needy Ryan Tannehill throwing passes to him from Week 1. He already broke the 100-PPR barrier last season, but his ADP isn't rising a lot, still under the 100th spot.
• Jordan Akins will be the No. 1 tight end for the Texans. He excelled in low-usage in 2019 so a bump in his opportunities should only do him good.
• Gerald Everett had to share the field, snaps and targets with Tyler Higbee, but his production was notable on a much lower usage. If he bumps it up a bit next season (and given his ADP won't go up as his situation hasn't changed) we could be in front of a very valuable player at the position.
• After having a pretty disappointing 2019 given his 2018 monster year, Eric Ebron will enter 2020 on the low and flying absolutely under the radar. He landed with the Pittsburgh Steelers, and he could be a sneaky value late in drafts.

Here are the actual ADP values of the aforementioned players in best-ball leagues as of this writing.

None of them are getting off the board earlier than at the 100th spot (Jonnu Smith) and even that amounts to more than eight rounds of picks. The value to extract from any of those players is really high, and given their historical comps, the odds are all of them have more than valuable seasons in 2020.

More Fantasy Football Analysis