Why not 2-point convert

[Cyberyeti]

i'm not so well versed in advanced stats.

 

is choking statistically significant and who are guilty?

There are certain tennis players particularly who were very well known for it and would lose from winning positions regularly. Laura Robson used to choke regularly but since the Olympics (where she seriously choked in the semis of the mixed but Andy Murray got her through it) she's been much better and has climbed the rankings as a result.

 

While there may not be players who are "clutch" (i'm not sure) I believe there are players who feel the pressure and perform very much below par in high stress moments at the end of games.

 

I'd like to see for example Adam Vinatieri's field goal percentages in games where a FG would cause a change of lead in the last 2 minutes of a game compared with his normal percentages for FGs of the same length. In his heyday, he seemed never to miss in that situation.

 

In cricket, there are some players known for getting runs when their team is in trouble and conditions are difficult (and hence they're under most pressure).

[Aberlour10]

dont't want to hijack this topic but generally... why are there in US sports about 1 trillion statistics for everything under the sport sun. I would not wonder if they even count how many yards make the dudes who bring the drinks to the players http://www.bridgebase.com/forums/public/style_emoticons/default/dry.gif

[jjbrr]

because it gives real sports fans something to debate and discuss beyond the casual "messi just shows more heart than anyone else" type arguments that don't mean anything and aren't based on anything.

 

because statistics can help predict future performance.

 

because seeing various statistical anomalies can provide interesting dialogue.

 

to give an accurate and consistent representation of how X player compared to Y player over a period of time in a meaningful way.

 

if you're willing to study them and understand them they certainly make sports better, not worse.

[jjbrr]

i'm going to shamelessly steal this from another site. and this post is from 2008, so the dates/players are less relevant now, and advanced stats has come a long way since then, so it's gotten even better since then.

 

Due to the recent influx of heathens on the site who constantly bash Sabermetrics/Performance Analysis without knowing the first thing about it' date=' I've decided to make this thread in hopes that it helps educate the non-believers. Input from fellow intelligent posters such as vhawk, DesertCat, Thremp, Jack of Arcades, and others are welcomed and encouraged.

 

[b']What is Performance Analysis / Sabermetrics?[/b]

 

Sabermetrics is the analysis of baseball through objective evidence, especially baseball statistics. Wikipedia's information on Sabermetrics is a great jumping-off point for the foundations of it (Bill James, SABR, Baseball Prospectus) and the influence it has had in the game. Sabermetrics looks to analyze the game of baseball by using statistics to evaluate how runs are created/prevented - subsets of these involve baserunning, hitting, fielding, pitching, home park advantage, steroids/PEDs, and many other fields of analysis.

 

What can be measured with Sabermetrics?

 

In short, everything. Opponents of Sabermetrics will state that "you can't measure heart/clutch," but they're wrong. We will only ask you how you define "clutch" - perhaps hits with runners in scoring position with 2 outs in the 7th inning and later? Due to Retrosheet, we can go back through hundreds of thousands of game logs in the past and collect situational hitting data to see whether or not Derek Jeter or David Ortiz has the "clutch" ability at the plate, or if certain pitchers are really "big game" pitchers and outperform their regular season lines in a meaningful way.

 

What is the role of sample size in baseball?

 

Sample size is simply the number of trials being used in a given analysis. The larger the sample size, the more accurate the analysis will be. A player hitting .300/.440/.600 over 10 ABs means very little, since his true line will be all over the place. However, if this hitter performs at that level over 600 ABs, that's very meaningful - even moreso over 2000 ABs. By pointing to a hitter and saying that he is 0 for his last 30 AB doesn't mean much. To prove this, simply look at a 10 pages of randomly generated numbers from 1 to 1000. Assume that the hitter is a .300 lifetime hitter and nothing would deter him from hitting this rate. If the number on the page is 1-300, the hitter gets a hit. If it's 301-1000, he doesn't. If you go through the randomly generated numbers, you will see long streaks of numbers where he gets no "hits." This is the essence of the mean, or average, of numbers.

 

How can we tell if a pitcher is likely to improve?

 

Statistical analysis of pitchers shows that these three things are most important in posting good ERA/WHIP numbers:

 

1) Strikeout rate. The more, the better.

2) Walk rate. The less, the better.

3) Batter's SLG % against. The lower, the better.

 

That's it. No opponent batting average against, no ERA, no hits allowed, nothing of the sort. GB/FB% plays a large role, but it correlates highly with #3, so we use SLG % against to represent that.

 

Opponents of Sabermetrics will often say that Pitcher X sucks because he has a high ERA or doesn't win any games. The truth is that pitcher Wins are largely irrelevant, as they rely so heavily on run support, which the pitcher cannot contribute to in any meaningful fashion (even in the NL). ERA, while a decent statistic, is flawed for many reasons. Let's take a look at the following pitchers:

 

Pitcher 1: 16-10, 3.83 ERA, 221 IP, 211 H, 27 HR, 99 BB, 151 K

Pitcher 2: 12-10, 4.08 ERA, 169 IP, 146 H, 14 HR, 61 BB, 144 K

 

Standard baseball analysts will say that Pitcher 1 is much better because he won 4 more games and has a lower ERA. Sabermetric analysts will point to Pitcher 1's high walk rate and HR allowed rate and favor Pitcher 2 because of his higher strikeout rate and lower HR allowed rate.

 

Pitcher 1 is 2006 Barry Zito. Pitcher 2 is 2007 Dustin McGowan.

 

When projecting pitchers, you should focus on their peripherals - walks per nine innings, strikeouts per nine innings, and home runs allowed per nine innings. They are the best indicators of who is lucky/unlucky in the ERA department and will give you the best idea of who will be successful in the future. Why? These stats are called Three True Outcomes - situations where the ball is not dependent on anyone but the pitcher and the hitter. When you introduce fielders into the equation, you add an element of randomness and skill that pitchers cannot meaningfully change. Which leads to our next point...

 

Wait - pitchers can't control balls in play?

 

Mostly correct, believe it or not! When a hitter puts a ball in play, the outcome is largely out of control of the pitcher on the mound. This axiom runs counter to common sense, but statistical analysis proves it - Batting Average on Balls in Play (BABIP) is relatively constant regardless of the pitcher. Don't believe me?

 

Pitcher 1: .296

Pitcher 2: .285

Pitcher 3: .292

Pitcher 4: .292

Pitcher 5: .272

 

The only one that is statistically significant is Pitcher 5. Must be an excellent pitcher to allow this type of BABIP rate, right?

 

Pitcher 1 - Jake Peavy

Pitcher 2 - Johan Santana

Pitcher 3 - Brandon Webb

Pitcher 4 - Jamie Moyer

Pitcher 5 - Barry Zito

 

Oops. For more information, Google Defense Independent Pitching Statistics (DIPS).

 

What books should I read to learn more?

 

Good question. The following books are excellent:

 

http://www.drivelinemechanics.com/images/thebook.jpg

 

http://www.drivelinemechanics.com/images/bpro.jpg

 

http://www.drivelinemechanics.com/images/bbtn.jpg

 

Furthermore, you will learn a lot if you read the following sites:

 

Baseball Prospectus

The Hardball Times

Fangraphs

 

If you are interested in how Sabermetrics meets up with traditional pitching/hitting mechanics analysis, you can read the following blogs:

 

Driveline Mechanics (my site)

Saber-Scouting

 

The bottom line is that Sabermetrics/Performance Analysis can analyze things that you might not expect. As a player and coach of the game, things I have learned from the analytical study of baseball directly contradict what I feel is the right course of action to do in a baseball game. That does not make the math wrong. Billy Beane, Sabermetrics's hero and GM of the Oakland Athletics, says that watching baseball is so difficult for him because his objective side is constantly at war with his subjective side. So it will be for you as you learn more about the analytical side of the game.

 

In general, Sabermetrics/Performance Analysis is a way of objectively analyzing the game and studying the reasons behind how the game works. Never take a cliche at face value - "Offense wins games, Defense wins championships," "Pitching is 80% of the game," or "Baseball is about three things: Pitching, fundamentals, and the three run home run." Investigate these claims and see for yourself.

 

If you are unwilling to learn, then kindly stay out of threads where you are ill-equipped to put forth informative arguments about the subject matter. It will only anger the posters in the forum and make you look idiotic. Take the time to research this area of baseball if you really love the game.

[Aberlour10]

because it gives real sports fans something to debate and discuss beyond the casual "messi just shows more heart than anyone else" type arguments that don't mean anything and aren't based on anything.

 

because statistics can help predict future performance.

 

because seeing various statistical anomalies can provide interesting dialogue.

 

to give an accurate and consistent representation of how X player compared to Y player over a period of time in a meaningful way.

 

if you're willing to study them and understand them they certainly make sports better, not worse.

 

 

thanks. you are right we (european fans) are different at this point. we are absolutely not interested in...how long was Messi in ball possession, how many fouls he did, or be fouled, the total length of his passes etc etc.

[jjbrr]

thanks. you are right we (european fans) are different at this point. we are absolutely not interested in...how long was Messi in ball possession, how many fouls he did, or be fouled, the total length of his passes etc etc.

 

but maybe a true soccer fan should be interested in these stats. a quick google search led to an article "Which Goals Matter? The High Value of Transition Play in the Premier League" in which they analyzed win/point probabilities of different types of goals (fast break, corner kicks, set plays, open play) and reached this conclusion: The results are pretty obvious, if you ask me. By far, the most valuable goals in the Premiership are goals from fast breaks.

 

That's kind of interesting to me! And if I can run those sort of analyses myself, I can offer some pretty interesting insight when I'm chatting with friends.

[Cthulhu D]

It difference is baseball has more discreet events, AND the traditional statistics are absolutely ***** useless when examining the game (RBI? Batting average? Really?). It's much harder to analyse football due to the lack of discreet events, and in cricket the traditional stats are actually pretty good, though it shares with baseball a lack of useful fielding estimates.

[billw55]

but maybe a true soccer fan should be interested in these stats. a quick google search led to an article "Which Goals Matter? The High Value of Transition Play in the Premier League" in which they analyzed win/point probabilities of different types of goals (fast break, corner kicks, set plays, open play) and reached this conclusion: The results are pretty obvious, if you ask me. By far, the most valuable goals in the Premiership are goals from fast breaks.

 

That's kind of interesting to me! And if I can run those sort of analyses myself, I can offer some pretty interesting insight when I'm chatting with friends.

hmm, I think all the goals are worth 1, whether open play, set piece, or penalty. Perhaps the teams scoring more often on breaks are just better teams to begin with?

 

traditional statistics are absolutely ***** useless when examining the game (RBI? Batting average? Really?).

Sounds like an overbid. Batting average is certainly worth something: definitely not the best indicator, but not useless. OBP matters, and has been in use a long time, although maybe doesn't meet your definition of "traditional".

[JLOGIC]

hmm, I think all the goals are worth 1, whether open play, set piece, or penalty. Perhaps the teams scoring more often on breaks are just better teams to begin with?

 

 

Sounds like an overbid. Batting average is certainly worth something: definitely not the best indicator, but not useless. OBP matters, and has been in use a long time, although maybe doesn't meet your definition of "traditional".

 

Batting average is worthless in the sense that OBP is just obviously much better and equally simple, and something that incorporates slugging is even more useful.

 

RBIs really are useless though, they are largely opportunity based. Is one guy with the same on base + slugging + HR totals as another guy really better because his team had more people on base for him so he could drive in more runs?

[wyman]

Batting average is worthless in the sense that OBP is just obviously much better and equally simple, and something that incorporates slugging is even more useful.

 

RBIs really are useless though, they are largely opportunity based. Is one guy with the same on base + slugging + HR totals as another guy really better because his team had more people on base for him so he could drive in more runs?

 

The real point is that we want to normalize out things over which we (and the players) have no control. I think it's interesting that you picked HR totals as a good indicator but reject RBIs. HRs are also opportunity based -- opposing pitching plays a role (but that almost certainly levels out over the course of a season, at least within a division), but more importantly a team plays half its games in its home ballpark.

 

Calling a stat useless just means (I think -- I don't want to put words in your mouth) that you feel that you can use some other combination of stats to extract better information (presumably better information that contains all of the information frpm the "useless" stat). I'd imagine that with enough work, we could render almost all existing stats useless. But that's not the same as saying that we have no ability currently to predict in baseball. For example, RBIs might be a weaker predictor of performance than some other predictor, but they are not completely information-free.

[billw55]

The real point is that we want to normalize out things over which we (and the players) have no control. I think it's interesting that you picked HR totals as a good indicator but reject RBIs. HRs are also opportunity based -- opposing pitching plays a role (but that almost certainly levels out over the course of a season, at least within a division), but more importantly a team plays half its games in its home ballpark.

 

Calling a stat useless just means (I think -- I don't want to put words in your mouth) that you feel that you can use some other combination of stats to extract better information (presumably better information that contains all of the information frpm the "useless" stat). I'd imagine that with enough work, we could render almost all existing stats useless. But that's not the same as saying that we have no ability currently to predict in baseball. For example, RBIs might be a weaker predictor of performance than some other predictor, but they are not completely information-free.

More or less what I was thinking. "useless" sounds like "equivalent to doing nothing".

 

Yeah RBI is pretty bad because it is lineup dependent. Even so, if you had murderer's row at the top of the card, but decided to put your pitcher in the 4 hole, he would not then crank out 100 RBI.

 

Also not all of the new stats are as great as they seem. WAR is popular now, but different sources calculate it different ways with different results, which makes it hard to follow. At least everyone agrees what OBP is.

[mycroft]

Yet again, "European" != "Britain". Football/Soccer is a different story, I think, but Cricket statistics - oh, my, cricket statistics...

 

I am interested in the Sabermetrics discussion. It seems to say that "ground-ball pitchers" are irrelevant (except in the HRs allowed, I guess), as they will have low K score compared to strikeout pitchers. That "seems" odd, but it wouldn't be the first time statistics beat feeling - especially *my* feeling re: baseball. I wonder how many Ks/game == 1 HR/season?

[mike777]

thanks. you are right we (european fans) are different at this point. we are absolutely not interested in...how long was Messi in ball possession, how many fouls he did, or be fouled, the total length of his passes etc etc.

 

 

Stats are a big part of the enjoyment of American Baseball compared to the other sports. Perhaps because we all grew up learning how to "score" a game with pencil and paper. We do not do this with our other sports. One reason why baseball records/stats tend to be treasured and debated compared to other sports.

 

Baseball scorekeeping is the practice of recording the details of a baseball game as it unfolds. Professional baseball leagues hire official scorers to keep an official record of each game (from which a box score can be generated), but many fans keep score as well for their own enjoyment.[1] Scorekeeping is usually done on a printed scorecard and, while official scorers must adhere precisely to one of the few different scorekeeping notations, most fans exercise some amount of creativity and adopt their own symbols and styles

 

 

http://en.wikipedia.org/wiki/Baseball_scoring

[jjbrr]

I am interested in the Sabermetrics discussion. It seems to say that "ground-ball pitchers" are irrelevant (except in the HRs allowed, I guess), as they will have low K score compared to strikeout pitchers. That "seems" odd, but it wouldn't be the first time statistics beat feeling - especially *my* feeling re: baseball. I wonder how many Ks/game == 1 HR/season?

 

This is almost certainly the most extreme example they could find, but if you believe the discussion about the role of defense in the primer I posted, it seems that in most cases the pitcher's dependence on the fielders is significant enough that it must be included if you want to isolate the pitcher's true contribution to the team:

Trevor Cahill, the young Oakland Athletics pitcher with a top-tier sinking fastball, has managed a .270 BABIP for his career. Cahill gets plenty of groundballs (56% GB%), and he’s very reliant on his defense. When the Athletics’ defense was ranked atop the American League, Cahill’s BABIP was .236 (last season). Though, when the Athletics defense deteriorated this past year, Cahill’s BABIP ballooned over sixty points to .301 despite his maintaining the same groundball rate (56%), home run rate (.88 HR/9) and a similar K/BB (1.74 vs. 1.87). Cahill’s ERA rose from 2.97 in 2011 to 4.31 this past season, illustrating the effect BABIP- and his fielder’s defensive ability– has on a groundball pitcher’s performance.

 

http://baseballnewshound.com/?page_id=1436

[jjbrr]

And sorry for the baseball derail. We could just as easily talk about advanced NFL stats and forget about baseball, but it was convenient to use baseball as the example given the massive amount of data available in baseball.

[mike777]

The NFL has on field violence as a huge attraction for fans. The NFL also has gambling as the other big attraction. With football you can have your local office pool, fantasy football, hard core gamblers or even numerous prop bets.

 

 

Stats for the NFL are really used by those gamblers, where as in baseball the fan just enjoys the discussions and those few of us who play Rotisserie Baseball.

[Cthulhu D]

Yet again, "European" != "Britain". Football/Soccer is a different story, I think, but Cricket statistics - oh, my, cricket statistics...

 

I am interested in the Sabermetrics discussion. It seems to say that "ground-ball pitchers" are irrelevant (except in the HRs allowed, I guess), as they will have low K score compared to strikeout pitchers. That "seems" odd, but it wouldn't be the first time statistics beat feeling - especially *my* feeling re: baseball. I wonder how many Ks/game == 1 HR/season?

 

Three singles are worth a home run under linear weights. Given that the league BABIP is .300, ~9 strikeouts and one home run has about the same run expectancy as 10 balls in play.

 

However, you do need to remember that the type of pitchers that give up lots of homers have other advantages. Confining the discussion to quality pitchers, consider Robin Roberts. He gave up a ton of home runs because he went in over the plate with a huge percentage of first pitch strike fastballs, but he also got a ton of first pitch swings and a ton of first pitch outs. This means he can pitch a lot of innings because he can keep his pitch count down.

 

Ground ball pitchers though do have another big advantage - you cannot turn two on a strikeout or a long fly ball.

 

Sounds like an overbid. Batting average is certainly worth something: definitely not the best indicator, but not useless. OBP matters, and has been in use a long time, although maybe doesn't meet your definition of "traditional".

 

Sure, but batting average in a baseball context has much less informational content than batting average in test cricket, which was my point.

[mycroft]

Well, seeing the jjbrr link above, one of the metrics includes 13*HR - 2*K (+3*BB) in the calculation. So it looks like 6.5Ks approx == 1 HR (8Ks == 1HR + 1BB). With a 35-appearance season, looks like about 3Ks/game make up for one HR.

[jjbrr]

http://i.imgur.com/uWFgf7U.png

 

Figure 6. (Left) The optimal decisions. (Right) The most popular decision from actual NFL games given a certain fourth down situation.

(Red = punt, Green = go for first, Blue = field goal attempt.) Coaches overuse the punt (instead of going for the first down) on their own side of the field, and perhaps go for the first down too much on the opponents side when they should settle for the field goal in fourth and short situations

 

Very interesting study

[cherdano]

Very interesting study

That's a very strange article, as it seems completely ignorant of the classic:

http://elsa.berkeley.edu/~dromer/papers/JPE_April06.pdf

(which comes to slightly different conclusions).

[jjbrr]

Before I click that, I think I recognize it as an article from 2005

[jjbrr]

derp apr 06

[cherdano]

Lol the article you linked to seems to be unaware that after a made field goal, you are forced to give the ball to the other team (unless you try an onside kick)... Really any single article posted on advancednflstats.com is more knowledgeable.

 

(It assumes the value of a made field goal is 3 points, where in reality it is 3 points minus the average value of the opponent's field position after kickoff.)

 

This article is a bit like a statistician writing an article about bridge in 2013 claiming that you should consider restricted choice! (But then getting all the computations wrong.)

[billw55]

(It assumes the value of a made field goal is 3 points, where in reality it is 3 points minus the average value of the opponent's field position after kickoff.)

The same is true for a touchdown. So the difference in value between a FG and TD is unchanged. Unless of course they subtracted the ops possession in one case but not the other.

[cherdano]

The same is true for a touchdown. So the difference in value between a FG and TD is unchanged. Unless of course they subtracted the ops possession in one case but not the other.

If you consider going for it on 4th down at the 10-yard line, you are comparing an (essentially) 100% field goal with a an uncertain outcome of sometimes turning the ball over on downs, sometimes kicking a field goal, and sometimes a touchdown. So in the latter case, you only have to subtract the value of "opp's possession after kickoff" weighted by some probability.

 

So the two mistakes don't cancel out. Isn't that obvious?