Interesting that the one small college player on your top ten list is also by far the clearest example on that list of the type of player who gets heavily over-rated by start-counting. CC Brown is a dismal, dismal safety who has no business starting on an NFL team. He can't really tackle and he really can't cover. But he's spent his entire career playing for a team with persistently dismal talent at safety (and one of the worst secondaries in the league as a result) so he's got a lot of playing time. In 2007 he was also assisted by a rash of preseason injuries to Texans safeties, without which he might not have started. In a few years, his starts/years eligible will be much less impressive.

Nice article, in all.

I came in prepared to make a joke about how the way to find value on Day 2 was to draft SEC players. Then I read the article and saw that it concludes that the way to find value on Day 2 is to... draft SEC players.

Score one for conventional wisdom.

Neat article, but I think to compare conferences, more weight should be placed on more important positions. I mean, getting a starting QB in Day 2 of the draft is a lot more impressive (and valuable) than getting a starting left guard there.

re: 3
I think the end of the article addresses that. Yes, New England got a whale of a deal with Brady, but he's an exception. You're probably better off finding your francise QB in the first two rounds and using later rounds to fill your other needs.

One question I would have is where RB's fit in the stats. Aren't there a number of later round picks that have turned out well? Or is it possible the position is so fungible, that GSAA works out lower because of turnover at the position?

Very good article, one minor nit-pick. Wasn't Jason Taylor a second round pick?

So how do teams get valued from round 2. Is it guys that had a pecieved flaw, guys who had limited experience, or guys who improved a lot?

Zack Thomas was too small ( flaw) etc.

You’re probably better off finding your francise QB in the first two rounds and using later rounds to fill your other needs.

Yeah, but if we're trying to evaluate conferences, we should give more value to more important players. If Conference A has lots of Day 2 picks that become starting Guards, while Conference B has slightly fewer Day 2 picks become starters, but at positions like QB, DE, and WR, then Conference B should probably have a higher rating, even though this method would rate it lower. That's what I mean.

I like this comment on the ESPN site:

BIG BAD BILL BARNWELL,

I wanna draft you on Day 2 all night long

<3

Joe

Shouldn't the positional breakdown table be adjusted for the number of eligible starting positions? To say that offensive linemen and linebackers tend to be the most successful players ignores the fact that there are more eligible starting positions, than say QB. If you adjust each position for the number of starting positions, the effects was out. OL drafted in the 5th round are about as likely as a QB to start at a particular roster spot like LG (2.89 / 5 = 0.57) as a QB is likely to start in the only QB position (0.48). Adjusting DL (2.41 / 4 = 0.7) and LB (1.58 / 3.5 = 0.45) yields similar effects. This analysis just says, "yeah draft a lineman because you need to start 9 of them and this kid is likely to get some playing time".

The Major Conference effect is however very striking.

The major conference effect to me is a reflection of the fact that players from major conferences are more likely to be NFL-ready from a coaching and experience standpoint whereas players from smaller schools are more likely to be "project players".

Re: 9

Well, if you need more OLs, DLs, and LBs than QBs, doesn't that mean you should draft more of them? Their odds of being good might not be higher than for QBs, but their odds of being useful are higher, since they have more opportunities to contribute.

#11 You're making my point for me. The word "successful" is a misnomer. It implies that the player is good. A more appropriate description is useful as you point out. Any player taken in the 4th round is a "flyer taken for naught" - not just QBs.

Re: Indeed. I think that could be a bit of bias introduced by looking all the way up to 2007. There may be small school players who take a couple years and then become starters. Conversely big school players might start if the position is a team weakness, but flame out or be replaced by a higher draft pick or free agent.

The numbers are probably similar, but I think it might have been more valuable to look at 96-05 instead of 98-07. It might not really matter though.

Ideally you'd want to look at longevity as well and look at 10 draft years, and assume a 10 year career for each player. You need a 20 year span though, so how you're looking at 87-98.

12. If your arguement is that it's easier for an O-Lineman to find a roster spot than a QB, then you need to take into account the number of players taken at each position. So if there are indeed 5 times as many OL taken in the 4th round as QBs, then they're fighting for the same amount of playing time as the QBs. In other words, just dividing by 5 won't give you your answer. 25 O-Lineman averaging 4 games/season obviously have displayed more value than 5 QBs averaging 1 game/season.

Of course, if you're talking about value you need to also take into account relative value of a position (I'd rather draft a QB with say a 5% chance of success than draft a Guard with a 20% chance of success, roughly speaking).

13. Actually I wasn't pointing that out as a weakness of the system, although obviously some of the later years would be slightly biased. Rather, I meant it to state a reason why we see the results we see.

But yes, as you say, maybe a team with a strong D-Line will take a small school pass rusher who they need to develop for a few years (and thus hurt his GS/S) but will ultimately, in their judgment, be a better player than the SEC guy who would start right away but have a lower ceiling.

#8: I like this comment better: "Bill must have got with KC Joyner on this one....metrics..Pppf ft "

Re: 15 I think KGB would be an excellent example of that. His GSAA would have been negative at the beginning of his career, get progressively higher, and is currently sliding the other direction now that he is no longer a starter.

From about halfway through the 3rd round, it's essentially a crapshoot. There is very, very little value of a 4th round pick above a 6th round pick. They are all very heavy longshots to become starters and most will top out as special teams guys.

For example, at RB, the average years as primary starter by round are:

1st rd - 4.1
2nd rd - 2.7
3rd rd - 1.5
4th rd - 0.7
5th rd - 0.8
6th rd - 0.6
7th rd - 0.5

#13

A way to attempt to find that info is go to pro-football-reference.com and use their draft querier. It is a godsend that I try to use everyday.

for instance in the 1996-2005 time period, the SEC has had 207 players drafted in the 4th-7th round. 81 of them has started at least 1 season. 81 of them has played 15 or less games. There has been 3 Pro Bowlers: Stephen Davis, Rudi Johnson, and Fred Beasley (all from Auburn).

The MAC conference (teams that were in the MAC between 1996-2005) had 41 players drafted. 15 started at least 1 season. 18 played 15 or less games. There has been 1 Pro Bowler: Asante Samuel.

The '"Tom Brady" spike in sixth-round games started for quarterbacks' suggests that variance needs to be considered in the data. If one guy can change the data, then the analysis that follows would seem suspect.

Not related, but I just wanted to say "Hooray!" for getting an email from Amazon saying that PFP 08 is shipping earlier than expected. Good job, guys!