There’s gold in them hills. Scoring dynasty rookie drafts

Everyone thinks every pick is great. Some are not.

I recently did some work on the 2016 draft class trying to assign expected scoring against ADP. Although results were fairly interesting the problem was that the overall correlation between ADP and production was fairly low. Early-round disappointments [Laquon Treadwell + Josh Doctson] and pleasant late-round surprises [Dak Prescott + Tyreek Hill] skewed the data. TO understand the relationship I needed more data. So I expanded my dataset.

This study incorporates all players included in MFL’s ADP data from 2014 onwards. Obviously players drafted in 2014 and 2015 also have multiple seasons to also add to the mix given we’re interested in lifetime value — not just rookie performance. In total I’ve included 858 player-seasons here. Although 82 of those rookies have failed to score a single point so far. And one [Brett Hundley] so far has a negative score.

Method

Exactly as the previous version I took rookie ADP data from MFL and made a big table of all players who were taken in rookie drafts 2014–2016 including all IDP players. Next I took points by season by player from one of my favourite soring setup leagues and tallied them. For interest it’s fairly standard offensive scoring with 0.5 PPR and balanced IDP scoring. Solo tackles 1.5, sacks 5 pts. Kick and punt return yards and TDs are included for all players. I split scoring season by season so the 2014 class has 3 seasons of data, 2015 2 seasons and 2016 just rookie data. All pick data is in pure number format rather than round because of the varying number of teams in different leagues.

Now I had a list of where players were taken and how they scored. I then looked at which sort of curve best described the distribution of scores for each season [rookie, year 2 and year 3]. I broke these seasons out separately because of the different expectations on rookies vs veterans.

It turns out logarithmic curves best describe all years. Generally we accept the early picks in rookie drafts are worth far more than late picks — rather a linear relationship. Anyone who’s ever tried trading up to a top3 spot knows this is true.

Once I had those curves defined I could calculate how many points each draft slot was “expected” to score. I’ll call these Predicted Points from now on [PredPs]. As an example the #1 pick should score about 178 points for you. The #2 pick 156 and the #13 just 98. Picking at the bottom of the 7th round in 12 team leagues [#84] PredPs is just 40.

I repeated the same process for year 2 and year 3. I then had expectations on how much picks at each draft position “should” score over their first 3 years in the NFL.

Having established then did a simple calculation of actual pts minus PredPs. This gives us how many pts above or below expectation a certain player scored. I did this for each individual year as well as a career total for each player. In a perfect distribution all of these would be zero but of course that’s not realistic.

How good at drafting are we?

This is actually fairly simple to ascertain in a broad sense. We can simply look at how well the actual pts per player fits the expected distribution. For all data in the set that looks like this:

Average points by draft position 2014–2016

As you’d expect on the whole there are bigger columns on the left. The players picked earlier have scored more points than those picked later. But there are lots of spikes and troughs. Those indicate players picked early who have underperformed or those picked late who over performed.

The degree of accuracy can be expressed as an R-squared value. The higher the R2 the better the correlation. A figure of 1.00 would indicate a perfect match whilst a figure of 0.00 would indicate no relationship whatsoever. With the curve and R2 added it looks like this:

There is a definable drop-off rate

It’s clear there is a relationship but the R2 isn’t amazing. Broadly we can pick good players but we still draft plenty of players too early or late.

I’d like the correlation to be higher but it’s plenty strong enough to use here. And it’d be a sort article if it wasn’t the case.

So what?

Ultimately the aim here is to make us better at drafting. Where do we make mistakes and why? And how can we avoid them? To start doing that we need to understand where our biggest mistakes lie. The further away from the baseline a column is the bigger the overall discrepancy [averaged across 3 players for each spot]. This is the average +/- PredPts for the whole dataset by draft spot:

Variance from predicted points by draft spot

There are a few interesting things here I think.

There is a negative score from every spot between #1 and #6 inclusive. Over 3 seasons those players selected have on average been disappointing. Here’s the players in the set and there scoring:

Higher blue bars than red bars indicate good performance.

So Zeke, Amari Cooper and Brandin Cooks have beat the target but that’s it. Several have come close but the abject misses [Watkins, Sankey, Doctson, Ebron, Treadwell and Agholor] counterbalance them.

No doubt plenty of you will say Sammy Watkins and the rest are still hits and they certainly have lots of dynasty value but in terms of pure scoring they have not lived up to expectations.

There are two big positive spikes in the data at #8 and #18. The players taken in those spots in the study were:

Are #8 and #18 magically great spots?

Similarly there’s a big trough at #19. The players taken there were:

Is spot #19 cursed?

This is just coincidence of course but if it makes you feel better feel free to trade up from #19 up to #18.

Should we be targeting players from certain NFL teams?

This is data from the whole set in total points scored broken down by position and ranked by team:

Jags rookies have scored way more FF points than NFL ones

So young Jaguars players have been impressive in the last 3 seasons. However that doesn’t tell is us those Jags were good value or not. So we can look at the dame data for +/1 PredPts:

The Cards have been efficient.

All of a sudden Arizona shoots to the top spot. Here are the players they selected and their +/- PredPts:

One of the best RBs in the league going late helps out with efficiency.

David Johnson was drafted at #18. Bucannon at #47. John Brown at #50.

The worst team for selecting players from in the study is the Rams. Here’s their players:

Why did Jeff Fisher get sacked again? That’s a LOT of players under the line.

A couple of IDPs who did really well does not balance out lots of poor offensive players. Even though dynasty players tend to think they’ve down-weighted Rams players over the period it’s not true. We still took them too early. Interestingly Todd Gurley has an overall negative score. His rookie season was about expected [+7.46 PredPts] but his 2nd year dragged him down [-21.3 PredPts].

Also noteworthy are Dallas, Tennessee and Green Bay who all had some very successful players [Marcus Mariota, Haha Clinton-Dix, Dak Prescott] and some very disappointing ones [David Cobb, John Crockett, Randy Gregory].

What positions are most and least valuable?

We can also break down +/-PredPts by position. Before we discuss this it’s worth noting that scarcity will of course alter how useful this is. Some positions contribute far more to fantasy teams and are harder to fill so FF players will accept more failures there in the pursuit of good picks. If we throw a lot of darts we’re bound to hit a bullseye. On top of that some positions have lots of good-scoring options but because plenty of alternatives exist on the waiver wire those positions are worth less and therefore get taken later in drafts. Everyone knows how undervalued QBs are in fantasy football and the same is true of some IDP positions.

Having said that points are a good currency to work against. On the most basic level if you have more players that score more points you’ll win more fantasy matches. That means we can compare players across positions but we need to take it with a pinch of salt. Across the study positions break down like this in terms of +/- PredPts:

Why do people even bother drafting rookie kickers?

Corners tend to do much better than their draft spot. Some of this is the well-known “rookie corner” syndrome. They tend to get picked on by QBs so they score a lot of points. Some of it is because corners tend to be undervalued in FF. You can always get another one from the waiver wire so people don’t invest early draft capital in them.

Safeties and linebackers are interesting here. There are also plenty of options for those two positions two but the top ones at those positions tend to be very good players [in most IDP formats at least]. This data suggests that they are good investments for picks. Here are the best value picks in the study for those positions:

Best value LB picks in the study

The number on the bars is the ADP slot they were taken at. Pretty much all difference-making players who were taken outside the first couple of rounds. A lot of FF players will no doubt say they’d rather take a shot on RBs and WRs in the first round and pick IDPs up later but given the failures we’ve already seen it might be time to rethink that strategy. Of course there have been plenty of disasters at those positions too:

Who the hell even is Lamin Barrow?

Jadaveon Clowney is a good player but certainly not worth the #16 pick. Jaylon Smith was way overdrafted even accounting for his injury status. And Paul Dawson was a flat out waste of a pick for everyone who took him.

Going back to the overall figures by position it’s also worth noting that TE is the worst position to draft on average. Accepted wisdom is of course that they take 3 years to turn into productive options and the data here backs that up:

I’ve excluded all players who scored beneath 20 points in a single season here so we’re only looking at viable players. It’s a really low base but notice how TE is the only position that increases scoring year-on-year.

Across the study here are the top TEs:

It’s not great when Richard Rodgers is the best option you could have taken

Yes. Those are the top ones. And only 3 have really been positive investments so far. Try asking Richard Rodgers and Crockett Gilmore owners if they’re happy with those players as starting options.

So what should you change?

I think there’s plenty of interesting information and food for thought here but does any of this help? Ultimately does it help you be better at drafting? I think it can do but you need to be ruthless. Don’t assume you can scout talent better than anyone else — it’s very unlikely to be true.

1. Think about taking IDPs earlier than you currently do

2. Fade skill position players from bad offenses as you normally would — and then fade them some more.

3. Be very, very careful drafting between #1 and #6. The smart play is quite often trading down because in some classes there’s no player that is worth that investment of capital.

4. Every spot from about #50 to #130 is pretty much worth the same. If you can trade down there for additional assets you should do it.

5. Don’t draft a TE. Just don’t do it.

Of course this is a relatively small dataset and there are a couple of issues [mainly the positional scarcity I mentioned earlier] but hopefully this gives you some ability to start grading your drafts a bit more objectively. If you ask dynasty owners how good a certain draft was then 9 out of 10 will claim it was better than average. Even the bad players they’re certain are just about to get good. I don’t think that’s right. I think that it’s far more valuable to understand where we made mistakes and try to learn from them. Good luck everyone with your 2017 drafting.

@TomDegenerate

One clap, two clap, three clap, forty?

By clapping more or less, you can signal to us which stories really stand out.