This is the 3rd in a series of articles I am writing to estimate the value over replacement (VoR) for various positions of need for the Colts. The last installment was published all the way back when Josh McDaniels was our head coach. So, clearly it is time for an update and this time around I looked at linebackers (LB).
Since 2000, LBs have comprised about 10% of all 1st round picks, which is one of the lowest of all positions(1). That number peaks in round 2 with 15% of picks and it holds steady through round 3 with about 14% of selections. The point of this is that teams don’t seem to prioritize LBs as a first round pick, preferring to wait until the 2nd round or later to spend their draft capital.
I again used average AV(2) earned each year after the draft to create a performance decay curve by position to provide a comparison of lifetime value for players at different positions.
For the first 3 years LBs have a similar growth period to DEs, which is steeper and longer than RBs. But after year 3, LB performance declines similar to RBs, which means that of the three, DEs retain the most long-term value.
Breaking the LB decay curve apart by draft round shows a clear performance gap between 1st and 2nd round draftees. It also displays a similar, albeit smaller, gap between rounds 2 and 3. This is the first hint that maybe the early round VoR for LBs is high, but I can’t make that call yet.
AV is fine to compare across disparate positions but to truly compare draftees within a given position, play-specific metrics are required.
When I analyzed DEs, I was able to extract a handful of metrics like Sacks, QB hits, Tackles for Loss etc. Since then, I have run across a composite metric for defensive players that was previously unknown to me.
Defeats is a measure created by Football Outsiders that utilizes a lot of the same metrics that I had used for my DE analysis, but it consolidates them in a simpler and more sophisticated way than I did. FO defines a defensive player defeat as a play where the defender performs any one of the following:
- Tackle for Loss (TFL)
- Interception (Int)
- Forced fumble resulting in a turnover (Fum_TO)
- Defending a pass that results in an interception or defending a pass on 3rd or 4th down which prevents a 1st down conversion (DP_i34)
- A tackle on 3rd or 4th down preventing a 1st down conversion (Tkl_34)
Defeats is the sum total of those plays made by a defender and that is what I used for my VoR metric. However, I also wanted to track the individual component stats to visualize what drives the total.
This chart shows each metric that makes up defeats normalized to a common scale (z-scores). These are per game measures and they show a very smooth exponential decay in performance from round to round.
That smooth decay is to be expected but it isn’t informative at all. This is just confirming that starters play more than back-ups.
There is a large disparity of snaps per game between players from different draft rounds. For example, 1st round players had over 52 snaps per game to accumulate their game totals while the 7th rounders only had about 8. So it isn’t surprising that the previous graph shows such a volume difference.
Converting the data to a per snap basis provided a common comparison(3).
From rounds 1 - 4, there is a pretty clear downward trend in performance, but the later rounds have some crazy spikes to them. That is similar to what I saw with RBs and DEs and a portion of this volatility is driven by the dramatically fewer snaps from later round picks.
But that is OK as I really only care about the earlier rounds anyway. I’m not going to make any arguments about the value of delaying a pick until the 5th round.
The last step to the VoR curve was to combine the data into total defeats, which I calculated for both LBs and DEs (forgoing my previous DE measures). I also trimmed the uncertain later rounds to limit the visualization to just rounds 1-4.
Interestingly, the LB VoR curve actually increases from rounds 1 & 2, although not dramatically. After that, however, there is a steep production decline that continues to the 4th round.
As I stated before, I am not trying to build a analytical model to identify which position to draft when. All this merely does is provide a comparison between positions that can be seen as an input into player selection.
When it came to DEs, the data showed an “urgency” to draft early. For LBs that same urgency isn’t there until round 2. And that fits with the behavior seen in the draft. Teams tend to hold off until round 2 to start spending their capital on LBs.
1) Based on a per player on the field allocation. LB assumes 1.75 players on the field on any given play (3.5 for defense, 0 for offense). This is merely a non-data based methodology to allow a rough comparison of pick volume among different positions.
2) AV data extracted from Pro Football Reference roster data
3) The snap count data was extracted from Football Outsiders and limited to 2012 - 2016