Can someone help me understand DVOA please?

**The Excellector** · 12-25-2016, 01:39 PM

Now, most here know well enough that I am not an advanced metrics guy at all. We have some posters who speak of this particular one as if it's gospel (Otherwise I wouldn't debate it so heavily). I provided an article in another thread about it, with quotes from one of the heads of one of these advanced metrics sites, yet it's still been declared by some that I misunderstand it. And you know what, honestly, perhaps I do. So I honestly figured it'd be a good idea to start this thread, so that everyone on the board can gain a better understanding of it. So, for those of you who have a much better grasp of DVOA, can you explain it to the rest of us please?

**Ravenswintitle** · 12-25-2016, 03:51 PM

start here:http://www.footballoutsiders.com/info/methods

**The Excellector** · 12-25-2016, 03:57 PM

Originally Posted by Ravenswintitle

start here:http://www.footballoutsiders.com/info/methods

I was actually checking it out as you posted this. I'm sure I'll have some questions though. Thank you for the lead.

**Bawlamer** · 12-25-2016, 04:14 PM

Originally Posted by Ravenswintitle

start here:http://www.footballoutsiders.com/info/methods

You have got to be kidding me. From what I read it's a perfect way to resolve nothing.

**Ravenswintitle** · 12-25-2016, 04:38 PM

Originally Posted by Bawlamer

You have got to be kidding me. From what I read it's a perfect way to resolve nothing.

Get off my lawn!

**Bawlamer** · 12-25-2016, 04:53 PM

Fine.... But put up a fence! I just stepped in dog crap!

Sent from my iPad using Tapatalk

**The Excellector** · 12-25-2016, 04:53 PM

Originally Posted by Bawlamer

You have got to be kidding me. From what I read it's a perfect way to resolve nothing.

That's what I thought as well. That was my gut instinct, but I just wanted to give this thing a fair shake, as far as really trying to understand it. I guess I'd be less against it if it were just treated like any other stat. However, not only on this board, but everywhere now it seems as if this DVOA is spoken as if it's gospel.

**BcRaven** · 12-25-2016, 05:46 PM

It's really so easy a cave man can do it :

What's the problem? ... Bc

P.S. - Does DVOA mean "Dimwitted Voice Of Authority?"

**RavensPPark** · 12-26-2016, 10:31 AM

This is a good introduction: http://www.fieldgulls.com/seahawks-n...ball-outsiders

On a very basic level, DVOA attempts to do a couple of things. First, to take into account the difference between running for 3 yards on 1st and 10 on the first play of the game and running for 3 yards on 3rd and 3 in a tie game late in the 4Q. They have figured out the expected value of every play in all possible combinations - there are points assigned to actual results versus expected value.

They also then compare what that defense has done versus the same plays. Finally, they do adjustments for strength of schedule. So to take my analogy further, it is more valuable to run for five yards against the Ravens 2000 defense at any point in a close game than to run for eight yards against the Browns this year.

There are lots of complaints about the system: it's a black box (no one else can figure out the numbers because the system is proprietary); doesn't separate units (i.e., doesn't credit OL versus RB); people find their OL rankings suspect (I trust Ken's film study better but that's just me); doesn't deal with injuries very AFAIK - but it's way better than using counting stats and it's probably the least imperfect metric out there.

Hope this helps.

**GaTechRavens** · 12-26-2016, 12:27 PM

Originally Posted by RavensPPark

This is a good introduction: http://www.fieldgulls.com/seahawks-n...ball-outsiders

On a very basic level, DVOA attempts to do a couple of things. First, to take into account the difference between running for 3 yards on 1st and 10 on the first play of the game and running for 3 yards on 3rd and 3 in a tie game late in the 4Q. They have figured out the expected value of every play in all possible combinations - there are points assigned to actual results versus expected value.

They also then compare what that defense has done versus the same plays. Finally, they do adjustments for strength of schedule. So to take my analogy further, it is more valuable to run for five yards against the Ravens 2000 defense at any point in a close game than to run for eight yards against the Browns this year.

There are lots of complaints about the system: it's a black box (no one else can figure out the numbers because the system is proprietary); doesn't separate units (i.e., doesn't credit OL versus RB); people find their OL rankings suspect (I trust Ken's film study better but that's just me); doesn't deal with injuries very AFAIK - but it's way better than using counting stats and it's probably the least imperfect metric out there.

Hope this helps.

Depending on how flawed their secret formula is then it'll probably end up being worse than counting stats, actually. Any attempt to put numbers in context runs a severe risk of overcorrecting. Look at their all time defensive rankings. The 2000 and 2006 Ravens aren't even in the top twenty, yet two other Ravens defenses somehow are. The 1985 Bears are just behind the 2012 Bears. Those would be cases where playoff performance vindicated their counting stats moreso than DVOA. DVOA has lots of suspicious results that need to be questioned and explained deeper, but can't because we don't know what the formula is. That doesn't mean it's entirely worthless or anything, but I'm very reluctant to give it any kind of real authority when it comes to current/all-time rankings - although I will say their offensive and even overall team rankings seem much more normal.

And that's just for overall units. Their DVOA position rankings (especially WR) are a disaster when it comes to measuring individual ability, though they do seem willing to acknowledge that. DYAR does a better job in that regard, though I'm still far from a huge fan.

**The Excellector** · 12-27-2016, 12:50 AM

Well, one thing came to mind for me, so anyone feel free to help me out with this. Let's say the average yardage on a first down run to the right is 3 yards. Let's say a run to the right is called, but a lineman meets the back three yards in the back field, for what should be a 3 yard loss. However, the back breaks the tackle, and then breaks another, before shaking and baking for a 1 yard gain. If I'm reading things correctly, that play would still count as a negative for that back, even though watching it it's a very good play by the back, because he didn't reach the average yardage gained on that play.

**Ravenstatsman** · 12-27-2016, 01:06 AM

Originally Posted by The Excellector

Well, one thing came to mind for me, so anyone feel free to help me out with this. Let's say the average yardage on a first down run to the right is 3 yards. Let's say a run to the right is called, but a lineman meets the back three yards in the back field, for what should be a 3 yard loss. However, the back breaks the tackle, and then breaks another, before shaking and baking for a 1 yard gain. If I'm reading things correctly, that play would still count as a negative for that back, even though watching it it's a very good play by the back, because he didn't reach the average yardage gained on that play.

DVOA makes no claim to evaluate a player outside his context. It isn't a measure of a player, it is a measure of a play. With your example, it is evaluating how efficient a rush is by an offense. So it is accounting for the poor quality of offensive line play that allows free lineman in the backfield, but also the skill of the running back for eluding them and gaining back 4 yards.