49ers Twice as Likely to Win with Shaun Hill than Alex Smith in 2009

the statistical model used to produce the measure is based on a massive sample size, not that each player’s stats is based on a massive sample size.

FO’s stat model used to develop DVOA is based on all NFL plays since 1993. that’s a massive sample size. over that many plays, you know pretty definitively what the “average” performance is, and DVOA is of course defense-adjusted value over average.

burke’s model is based on the past 5-10 years of plays. that’s a massive sample size too.

sorry i wasn’t clear enough that i was making a statistical methodology point. put this together really quickly. but, the basic idea is that you first use some historical data set to develop a measure, then you test how good that measure is at “measuring” what it’s supposed to be measuring (e.g., by looking at correlations w/ important football-related things), and then you apply the model to a current data set. so, in burke’s or FO’s case, the massive sample size i’m talking about is with respect to the “historical data set” they used to create the equation(s) for their stat(s).

Next season is basically put up or shut up time for Alex.

Which is why I don’t put it out of the realm of possibility to draft a QB. Create some competition to who could possibly be the heir apparent if Alex goes down. Hope he doesn’t, but we need some backup, no pun intended. I say we take one in mid to later rounds though. We have bigger fish to fry early on.

I agree with that statement 100 percent.

It isn’t fair to evaluate Smith, when he has had to adjust to so many systems without ever have time to learn them.

Vernon Davis didn’t become a pro bowl starting TE because he threw himself all those TD’s, Smith dropped 3 of his 13 in one half. Smith made Crabtree the second most sucessful 49er rookie WR in franchise history, with no offseason and the first 6 games to work with. This is about timing and knowing the offense, with 5 offensive schemes and 2 years of shoulder injuries, Smith is just starting to show what he is capable of doing. Least ye not forget that 1/3 of his picks came to balls being deflected by his own recievers or guys (VD) not looking for the ball.

Smith is the only QB in the NFL that can throw for 250 yds, 1 TD, 97.5 passer rating, win the game and he still didn’t do enough!

DVOA looks at historical games to determine their statistical metrics. Actually one of my gripes with the entire advanced statistical practice. At some point the historic data becomes too historic but no one has a measure to determine when that point might be.

Will the 49ers really score 18 more points per game on average with Hill than Smith? I find that impossible to believe. Reading that stat made me think of the joke were the accountant is standing before St. Peter at the pearly gates, and Peter asks him what 2+2 equals. The accountant answeres, “Anything you want it to be.”

I remember a certain play with Smith throwing a TD over the heads of two defensive players to Frank Gore in the corner of the endzone. Of course people don’t seem to remember the plays Smith made in games the Niners won. People only remember the losses.

Yeah, arguments can be made that Russell is the worst QB in the league right now. Smith is no where near the worst.

The model is measuring the timeliness of plays. A 2nd and 5 conversion in the 2nd quarter will unlikely decide the outcome of the game, but historical data suggests their is a higher win probability if converted late in 4th.

I went through all the 49ers games and subtracted all the points scored by the defense and special teams. In the 5 1/2 game Hill started, the offense averaged 15.27 points per game. In the 10 1/2 games Smith played, the 49ers offense averaged 20.76 points per game. That’s 5.5 more points per game by the offense when Smith is the QB. But according the stats, the 49ers offense should score 18 points more per game with Hill than Smith. Can someone please explain to me how this works. The 49ers offense scored 5.5 more points per game with Smith, but they really should have scored 18 more points per game with Hill. It doesn’t even make any sense.

Nolan had no clue what he was doing on offense.

I’m not sure that’s what it’s saying. I’ve figured out the expected points stuff thanks to Danny, but I’m still taking some time to consider how the win-probability part is functioning.

And that was known..alot of people didn’t like that..but its true..which points back to the fact that if the offense which early on was predicated on frank gore with predictable shorter passing patterns..instead if it was more creative and opened up from the start then yes shaun hill likely would have led us to the hallowed playoffs.

This is another thing that I’m not comfortable with. Saying, “well, everything being equal….” is kind of disingenuous because, well, no, everything isn’t equal. There are reasons that the offense was so one-dimensional with Hill and not with Smith. Some of those reasons are probably not good reasons, but some of them probably are.

I absolutely understand being critical of the playcalling “hamstringing” the offense with Hill and “freeing” the offense up with Smith, but I’ve never really felt comfortable assuming such a one-sided relationship. It’s well known that Hill doesn’t have the tools to execute some of the more dynamic plays in the playbook. It should also be well known that Hill is a sack and a half waiting to happen on every play.

These are limiting factors. Smith was actually well above average at avoiding sacks, and has the arm to make some of the more dynamic plays feasible.

Now… this is where I get a little crazy… so, I apologize in advance….

I also think – and this is absolutely wild speculation, but bear with me – that Hill might have playbook problems. When Martz was the OC, Hill dropped to #3 behind JTO and a still-injured Smith for almost no other reason than that he was so much farther behind the other two learning the playbook. Granted, JTO had a head start. But think about it. Smith was hurt… and Hill was so far behind the eight ball that he couldn’t beat him out in camp. When Hill finally did get starts with Martz, the playbook was scaled back significantly to accommodate for it. I find it conspicuous that the playbook under Raye was also less dynamic with Hill as the quarterback. When Smith came in, the offense was immediately opened up.

I don’t know… even I don’t really buy that. I’ve been thinking about it for a while. In some ways, it makes sense to me, but in other ways it seems like a trash conspiracy theory.

But ultimately, the fact remains that no, everything is not equal. For whatever reason, the team has used scaled back playbooks with Hill for the last two years. Is this because of his physical tools? Is this a fundamental philosophical coaching issue? Is it a problem with Hill’s actual playbook acuity? There is a reason that it’s been this way. The obvious choice is that Hill’s tools are legitimately limiting and the coaches have overreacted to that.

the WPA explanation article i linked to above. in it, burke specifically says

Stats are tools, and each tool has its own purpose. WPA is what I call a narrative stat. Its purpose is not to be predictive of future play or to measure the true ability of a player or team. It simply measures the impact of each play toward winning and losing.

hope that clears that up. he’s not predicting here. he’s using a stat to tell the story of what happened at Niner QB in 2009. so, this has nothing to do with making a football decision in re starting hill or smith.

…Burke specifically mentions Hill as a QB likely to regress next season. I’m not sure this article says what people seem to think it’s saying (i.e., that Hill is a better QB than Smith).

I read it as Smith is 9x less clutch than Hill because the statistic didn’t indicate Hill was clutch.

…meant to post this in response to howtheyscored.

Well, BABIP is the analogy I’m using in that you assume that players who outperform or underperform historical norms will return to those norms. I’m not sure that “random chance” is what I’m getting at so much right now just because I have no gauge with regard to a player’s ability to control this stat. BABIP is good for this because there is such limited player control over performance fluctuation. You can always expect BABIP to return to a certain “normal” range for a very, very large majority of the players because so few players demonstrate an active ability to control BABIP performance in any way.

Like I said, I don’t know how much control players actually have over WPA. I don’t know how wide the range of regression would need to be, nor do I know how big the sample has to be before you have a reliable measure of a player’s abilities.

I’m just thinking in very general terms. Danny says above that a player having a high WPA isn’t predictive that they will have a high WPA in the future. I was just reminded of BABIP, where a player having a high BABIP isn’t predictive that they will have one in the future. But it does have some predictive power in other ways because it’s such a stable stat. I don’t know how stable this stat is, but I like the analogy on its face at least.

…that was supposed to be in reply to howtheyscored, as well. Sorry for any confusion…

Meet Nate Kaeding’s playoffs performances.

But Al Davis many don’t want to entertain the thought that a backup noodle arm plays on par with a stronger armed, more athletic Jamarcus Russell Alex Smith.

Well, last season at the very least, Hill was not a serviceable backup. He was pretty terrible.

His career tells a different story, though, so it’s not really a nit worth picking.

with the caveat that Hill’s ceiling is as a serviceable backup whereas Smith might be able to turn into a pretty decent QB.