Photo: Imagn Images

Most everyone understands that effective trading means buying low and selling high. The tricky part -- and what goes a long way towards separating the wheat from the chaff fantas leagues -- is the ability to accurately identify buy-low/sell-high candidates. No doubt, this skill is both an art and a science. In this new weekly column, I'll be leaning towards the science side, using statistics to estimate the number of fantasy points by which a player has overachieved or underachieved so far this season.

True fantasy points explained (BRIEFLY)

Now, when we talk about overachieving and underachieving, the question becomes "compared to what?" One obvious example would be a player's preseason projection. For instance, Aaron Rodgers is curently averaging 22.1 points per game compared to David Dodds' final projection of 24.1, and so underachieving by 2.0 points per game means we should think about buying Rodgers low.

The system I'm going to use for identifying underachievers and overachievers is a tad more complicated than that, but its fundamental building block is regression to the mean (RTM). In fantasy football, the popular understanding of RTM is that a player who's scoring far above average is likely to score closer to average going forward, and vice versa. Underlying RTM is a deeper idea, though: A player's fantasy points result from some combination of his "true" scoring ability and dumb luck; RTM only occurs because of that "dumb luck" component. Using some fancy math, we can quantify it, and subtract it out, thereby getting an estimate of a player's "true" fantasy scoring ability. It's this ability estimate, which I call True Fantasy Points (TFP), that answers the "compared to what?" question in my system.

If desired, you can read all of the gory methodological details of TFP elsewhere, but the abridged version is as follows:

1. We know a player's current fantasy point total.
2. Using reliability analysis, we can estimate the number of touches it takes before a given stat represents half-skill/half-luck.
3. Using this estimate alongside a player's current number of touches, we can calculate how much RTM to expect for him going forward.
4. Accounting for RTM gives us an estimate of a player's TFP.
5. The difference between a player's TFP and his current point total tells us the extent to which he's overachieving or underachieving.

TFP through week 5: quarterbacks

Below is a table showing current TFPs for the Top 36 QBs in our Top 200 Forward projections using standard FBG scoring:

Before we start trying to use these stats to make trade offers, let's go through a quick example of how to read the table. Take a look at the row associated with Andrew Luck. The first stats column, "Obs FBG," indicates that he's scored 113.3 fantasy points in the passing game through Week 4. The next two stats columns tell us Luck's TFP from two different perspectives of RTM. "TFP14" assumes that only his attempts in 2014 should count toward calculating his "true" fantasy scoring ability, while "TFP" assumes that all of Luck's attempts with his current team -- in this case, Indianapolis -- should count. This distinction has important implications that I'll come back to later. In any event, the table indicates that Luck's TFP is either 91.8 points or 84.6 points, depending on which assumption we make. The final two stats columns are simply the differences between "Obs FBG" and the two versions of TFP. "+/- 14" tells us that Luck's overachieved by 21.4 points assuming his TFP should only be based on this year's pass attempts, while "+/-" tells us that Luck's overachieved by 28.6 points assuming passes he's thrown for the Colts in 2012 and 2013 should also influence TFP.

With that out of the way, let's start identifying the best buy-low/sell-high candidates heading into Week 5.

buy-low candidate #1: Nick foles

Either way you slice the "only 2014 or all Eagles attempts" pie, Foles has scored wildly below what one would expect from his 166 pass attempts. Applying RTM to his current stats, one finds the following: His 7.07 yards per attempt is likely to improve to somewhere between 7.19 and 7.55, his 3.61% TD rate should improve to somewhere between 4.50% and 4.88%, and his 2.41% INT rate should drop to somewhere between 2.09% and 2.37%. In the parlance of reliability analysis, Foles' current point total is likely to be far below his "true" point-scoring ability. Further helping matters is that he's coming off of a 7.8-point dud against San Francisco. If there ever was a time to pull the trigger on Foles, it's right now. (Of note: Despite said dud, he still ranks ninth in terms of fantasy points from the passing game.)

buy-low candidate #2: drew brees

As it stands right now, Brees has the top-ranked TFP among QBs. That's because his previous 4,990 attempts with New Orleans suggests that his current 7.47 yards per attempt should be closer to 7.72 and his 4.35% TD rate should be closer to 5.46%. And although Brees himself isn't coming off of a dud, popular perception after a blowout loss to Dallas indicates that the Saints have a distinct odor of underachievement, which is likely to stick to Brees in fantasy trade negotations.

buy-low candidate #3: tom brady

Yes, the Patriots offense in 2014 is an insult to its recent ancestors. Yes, Brady doesn't look too good out there right now, especially in terms of his once-legendary throwing accuracy. Yes, he can't hit WRs deep anymore in the rare event that he even tries to. And yes, his underwhelming receiving targets wouldn't help out much even if he could hit them deep. With all of that said, I still think Brady's a good buy-low candidate.

Here's why. As previously mentioned, I give you two options for TFP, depending on whether you think pre-2014 pass attempts for the same team are relevant for gauging a QB's "true" fantasy scoring ability. The vast majority of the time, I would personally rely on TFP rather than TFP14 because prior attempts for the same team do matter by and large. However, the offensive situation in New England, coupled with his age, heavily suggests that the Tom Brady of 2005-2012 isn't walking through that door. And yet, even if we ignore his entire Patriots career, Brady's still underperforming his "true" 2014 ability by 14.7 points, or 3.7 points per game. That's more than enough to make me consider him a QB2 bye-week target that I can acquire on the cheap -- especially after "The Debacle for the Decibels" we saw on Monday night.

sell-high candidate #1: andrew luck

Luck is currently the No. 1 QB in fantasy football, with a margin of 3.8 passing points per game separating him from second-ranked Matt Ryan. The relevant question for trading purposes isn't "will he remain No. 1 for the rest of the season;" it's "will he remain No. 1 by that large of a margin." Even accounting for RTM, Luck could still finish 2014 at the top. But if other QBs close the gap, then it probably means they had a higher per-game average than him over the final 13 weeks. And if you're able to trade Luck for one of those QBs, alongside a shiny new toy at another position in a package deal, then the fact that he's No. 1 right now is neither here nor there: It's likely to be a positive expected value trade.

Now, about that RTM. So far in his Colts career, Luck has thrown 1,364 passes, averaging 6.97 yards per attempt, with a 4.33% TD rate. Given the number of passes it takes for these stats to stabilize, our estimate of Luck's "true" ability is 7.03 yards per attempt, with a 4.46% TD rate. Based on his current stats in 2014, this means that Luck's overperforming his yardage ability by 0.78 yards per attempt and outperforming his TD ability by nearly double. A 7.78% TD rate is a stratospheric level sustained over a full season only by QBs like Peyton Manning, Aaron Rodgers, and Tom Brady in their primes. If you believe Luck suddenly turned into those guys when the calendar flipped, hold on to him by all means. However, if you think that his performance over 1,197 attempts prior to this season tells us something about his "true" fantasy scoring ability, then make an effort to sell Luck high.

And like Foles and Brees, the general public's recency bias is working in your favor, what with his season-high 34.7 points against Tennessee in Week 4.

(Full disclosure: I'm a Luck owner in the FBG staff league, and I'm seriously thinking about selling high on him, especially given the sorry state of my RB corps. In other words, I'm not sitting here advising you to do something I wouldn't consider doing myself.)

sell-high candidate #2: eli manning

Manning is another QB who ows his lofty position in the fantasy rankings to an out-of-character TD rate. Thus far in 2014, Manning has thrown 9 TDs in 139 attempts, for a 6.45% TD rate. His career-high for a season was 5.75% in 2010, and his "true" TD ability given all of his pass attempts with the Giants is 4.63%. Even if you ignore all attempts prior to 2014, his "true" TD ability is still only 4.85%. Now, combine that with the fact that his current INT rate of 3.60% is not that far from his 3.14% "true" INT ability, and you've got a recipe for fewer TDs, but with a similar number of INTs, going forward.

tfp disclaimers

In the interest of professional integrity, it's important that I'm up front with you about TFP's blind spots, of which there are at least five:

1. It doesn't know anything about the strength of a player's future fantasy schedule. Eli Manning's next three games are against Atlanta, Philadelphia, and Dallas. Andrew Luck's next three opponents are much tougher (Baltimore, Houston, and Cincinnati). I've listed both of them as sell-high candidates, but if you think that schedule difference makes Luck even better trade bait than he already is, go for it.
2. It doesn't know anything about team game scripts and pass identities. NFL offenses tend to pass early in the game and when they're behind late; they tend to run later in the game, especially while they're ahead. Similar to TFP's strength-of-schedule blind spot, its underlying RTM has nothing to do with whether a player's offense is likely to be ahead most of the time or behind most of the time. For instance, Blake Bortles may be a weak sell-high candidate according to TFP, but TFP doesn't know that Jacksonville's offense has been -- and is likely to continue to be -- playing catch-up during the bulk of their time on the field. If you think Bortles won't regress as much to the mean because he'll be throwing 40-50 passes per game going forward, feel free to hold onto him. (And obviously, I'm only talking about redraft leagues here.)
3. It's less trustworthy with younger players, especially rookies. Speaking of Bortles, having zero pass attempts in Jacksonville prior to 2014, and only 61 attempts so far this year, means that his TFP is heavily weighted towards the league-average QB. Is Bortles destined to be a league-average QB? Early signs suggest not. If you think that makes him less likely to experience RTM going forward, TFP shouldn't stop you from holding on to him.
4. It doesn't know anything about injuries or suspensions to the player himself, his teammates, or his opponents, be they past, present, or future. Jake Locker has a bum wrist, and Teddy Bridgewater has a bum ankle. Jay Cutler has two chronically banged-up wide receivers. Brian Hoyer (or Johnny Manziel) will have Josh Gordon back as a receiving target for the stretch run. If such information makes you avoid buy-low candidates like Locker and Bridgewater, hold onto a sell-high candidate like Cutler, or get even more gung-ho about acquiring Hoyer (or Manziel), there's nothing about TFP that should stop you.
5. Although it knows how long a player's been on his current team, it doesn't know how long his team has been running its current offensive scheme. Eli Manning, Andrew Luck, Philip Rivers, and the Washington QBs have all experienced recent wholesale changes to their offensive schemes despite still playing for the same team. It stands to reason, then, that their pass attempts in previous schemes shouldn't count towards TFP. If you think the "true" fantasy scoring ability for these QBs is much different in their new systems, then feel free to mentally adjust their TFPs -- and implied trade values -- accordingly.

Follow @IR_DannyT