# What is the Best Draft Strategy for My Mental Health?

While I had a great many things to be thankful for this holiday season, luck from the fantasy gods did not feel like one of them. It was a deeply frustrating season that was terrible for my mental health, as my dreams of the fantasy playoffs were dashed for the majority of my teams.

If I can take any solace from all this failure, it’s from the great variety of ways in which I managed to fail. My early-round wide receivers, like **Alshon Jeffery**, busted, and my late-round WRs, like **Allen Hurns**, did too. On one team, I took **Doug Martin** in the middle rounds, and on another, I took his backup**, Charles Sims**, later on. Somehow, they both managed to bust this year, which is particularly impressive given that both were top 16 running backs last year. I’m especially displeased by my Zero RB team, where neither of my top two RBs, **Dion Lewis** and **Danny Woodhead**, managed to stay healthy. However, let me take a moment to congratulate those who had the wherewithal to draft Woodhead’s handcuff, **Melvin Gordon**, a round or two earlier!^{1}

Rather than continuing to wallow in self pity, though, I decided to take a more constructive route: looking back at my draft strategy to figure out what I could have done better. In particular, I wanted to figure out what approach would have improved my mental health by giving me a roster whose weekly lineups — despite byes, injuries, and busts — would leave me consistently confident going into each game.

While there are many ways to look at that problem, I went with the most naive one: test out every draft order. Thanks to a free trial offer for the Google Compute Engine, where I could use up to 32 computers simultaneously for a month, that was easily achievable.

## Methodology

A certain amount of volatility is inevitable in this game. Even the most coveted fantasy players like **Odell Beckham**, **Julio Jones**, and **A.J. Green** have all scored under six fantasy points at least once this season. However, starting those players every week still leaves me confident because, going into each game, they’re always projected to put up big point totals. Generalizing that idea, I wanted to look for the draft strategies that yield lineups whose weekly *projected* fantasy points were as consistent as possible.

In order to model projected fantasy points, I started by calculating the average points per game of each player taken in drafts from 2000 to 2014. Here’s what that looks like after some smoothing:

Since this is PPR, WRs outscore RBs of the same position rank all along the curve.

Here’s the other part of that calculation, the probability of actually playing in the game:

Once again, the curve for RBs lies entirely below that of WRs. This reflects the fact that RBs get injured more often and also lose their roster spot more often (since fewer RBs than WRs typically suit up on game day).

The last part of the model is an estimate of weekly standard deviation in projected points. I calculated this using the weekly projections from 4for4 Football last season. Once again, WRs end up more consistent than RBs, with a standard deviation of 1.8 points versus 2.7 for RBs. That could be, for example, because RB performance is more dependent on matchup than WR performance.

Those components are enough to allow us to simulate constructing a weekly lineup by generating a random projected point total for each player and then choosing the best options. I used a lineup with 1 QB, 2 RB, 3 WR, 1 TE, 1 K, 1 DEF, and 1 RB/WR/TE flex position. To relate that to draft strategy, I used PPR ADP from Fantasy Football Calculator to determine which position rank would be available for each position in each round. Each draft I considered was 16 rounds and included 2 QBs, 4-6 RBs, 4-6 WRs, 2 TEs, 1 K, and 1 DEF.

For each roster, it took about 150,000 simulations to get a good estimate of the roster’s average projected points and standard deviation. Doing all of those simulations for each of the more than 2.4 million draft orders, required over 360 billion simulations. Using 32 machines on Google Compute Engine, this took around eight hours to complete.

## Results

As expected, the week-to-week volatility is almost completely explained by where RBs are taken in the draft. However, the relationship is not what I expected.

The relationship between standard deviation and the average round in which the RBs are taken is almost perfectly linear (R-squared of 0.96). However, the line is sloped the wrong way: we get less volatility by drafting RBs earlier!

The relationship is even clearer if we look at where WRs are taken rather than RBs. As with RBs, the relationship between volatility and the average round in which WRs were taken is almost perfectly linear (R-squared of 0.96). However, with WRs, we also get linear relationships to where the first or last WR was taken (R-squared of 0.83 for both).

Once again, the line is sloped the opposite way of what I would have expected. Taking WRs earlier leads to more volatility, not less. The results indicate that, if you want to optimize for your mental health, you should draft WRs late and load up on the other positions earlier on.

What is going on here?

I think the key to understanding this is appreciating how bad late round RBs are, at least on average. They have more week-to-week-volatility than WRs, and they miss more games. (Plus, even when they do play, they score fewer points than late-round WRs.) These results are telling me that I probably should have been preparing for an emotional rollercoaster when my first two drafted RBs were Dion Lewis and Danny Woodhead. From that perspective, those taking Melvin Gordon a little earlier in the draft were actually being more sensible.

This problem is not just true of late-round RBs, however. Late-round TEs and QBs also have a lower probability of playing each week. Late-round TEs (though not QBs) also have more week-to-week volatility in projected points. This explains why the relationship between WR draft position and volatility is stronger; those that take WRs later avoid not only late-round RBs but also potentially late-round TEs and late-round QBs.

## Conclusion

My main takeaway from this study is that it’s too easy to forget that you can’t just focus on which positions you are taking early in the draft. Skipping a position early means taking it later on, and, even though all of the positions get pretty bad late in the draft, some are worse than others.

There are also a couple of caveats to this study that are worth mentioning.

First, it assumed an average drafter. The points per game and play probability models are based on average results for players drafted near that position in past seasons. It is certainly possible there are inefficiencies in the draft market that allow some to draft much better than average. There is a lot of anecdotal evidence, for example, that pass-catching RBs were undervalued in PPR leagues in the past. That was probably not the case in my leagues this season, but depending on how ADP for RBs changes next year, those inefficiencies could reemerge.

Second, this study ignored the waiver wire. Now, it’s safe to assume that players who end up on the waiver wire, by going undrafted, have even worse preseason prospects than those drafted late. However, those prospects can change during the season due to injuries, trades, and so on. If you can routinely spot changes in the prospects of waiver wire players before others, then you probably want to change your draft strategy to account for that. I don’t have a lot of confidence that I can do so in my own leagues — my opponents tend to hover over the waiver wire like vultures — but this is a real inefficiency in many leagues.

In the end, are these results going to change the way that I draft? If you ask me today, I’d say yes. The idea of having **David Johnson** or **Ezekiel Elliot** on my team right now sounds really nice. That said, despite how traumatic this season has been, something tells me that, next season, I’ll have forgotten all about this misery. Instead, I’ll probably start the season convinced, once again, that I know exactly which late round players are going to hugely outperform their ADP, so look forward to part two of this sad story in one year’s time.

### Subscribe for a constant stream of league-beating articles available only with a Premium Pass.

- Why, yes, I am still bitter about that. Thank you for noticing. (back)

Glad to see someone here bucking the trend of zero RB. I've felt there were things not being considered when picking the WRs first, and this has at least partially explained a potential advantage of going RB early. I hope to see more debate or discussion on this.