rluzinski

This does show something which does not make sense to me, that a 2B is worth .78 runs but a single + SB is worth .77 runs. Your value for a SB looks a little high. These are the values I typically use: http://www.tangotiger.net/RE9902event.html 1B + SB= .474 + .195 = .669 runs 2B = .764 runs Roughly speaking, the "moving runners over" portion of a double is worth about .1 runs greater than a single. That sounds reasonable. Keep in mind, however, these run values are still only estimates and are dependent on the sample used. It would be interesting to see how much the "moving runners over" run values differ by batter type. A slugger's average single is going to score more runners on second than speedy bunter with no power. Probably not too much.

I don't understand that logic at all. A team that is desperate for starting pitching is going to cut Bush? You'd have to think he's not worth what he'll likely get in arbitration. If that were the case, he would have negative trade value. CHONE has Bush projected for a 4.78 ERA in 170 innings. If he get's more than $5 mil in arbitration, he might start having negative value. I'm assuming he'll get under that? At $4.5 mil, he's fine and the Brewers will probably need him.

To bring a free agent to Milwaukee you are going to be paying a premium either it being for an average player or an above average player. There are also some players that probably prefer to play in smaller markets, so I'm not sure how large this supposed premium the Brewers have to pay really is. I bought the premium theory more when the Brewers were pathetically bad. Prince by far is the best option for the Brewers of any future free agent where the Brewers will not have to overpay because of the city the Brewers play in. You are presuming that Prince wouldn't prefer to play for a large market team just because he was forced to start his career in Milwaukee. Also paying for free agents that are past 30 years old is not a wise decision for the Brewers injuries and decline are going to be way to risky for the Brewers to sign players after 30 years old. Teams aren't clueless. If you have two free agents with identical projections, the older will, on average, be paid less. And I'm not too excited about signing any player to a 5+ year guaranteed contract anyway.

Why? Who else are the Brewers going to be spending that money on? If the Brewers don't spend that money on Prince they are going to be spending it somewhere else. More than likely it will be on free agents who are older than Prince and are more than likely in their decline. Prince being paid up to $20m a year wouldn't be a bad investment for the Brewers. A strategy of relying on your farm system while on spending money on certain players sounds more like a great strategy than a terrible one. The Brewers really should stay away from free agency but that doesn't mean they should stay away from paying their players that they already have and need at free agent prices. I'm not even sure Prince is going to get the big bucks that some are thinking he will be getting especially if you look at the possibility of who maybe available after the 2011 season. I can't under stand how you can be confident enough to say that paying Price $20 mil/year for 5-7 years is obviously better than any other possible free agent aquisitions. If it is, other teams will be tripping over themselves to offer him more, anyway. Unless you are presuming a home town discount, the logic falls apart.

They've won four division titles with Beane at the helm, at least.With that pitching staff, they probably could have won those division titles with Dean Taylor at the helm... Well, maybe not, but you get my drift. I think GMs should only take credit for winning with bad player.

I think there's little doubt the Brewers will make an offer to Prince consistent to his worth to the Brewers. The problem is that he's worth more to a larger market team, even after the player compensation. If the Brewers want to pay Prince what he's worth to the Red Sox, I think that's a terrible strategy.

If you assume that those stolen base attempts never result in a caught stealing, it's pretty close. If you assume even a 75% steal rate, it's not.

Anytime I see the words "ESPN" and "study" together, I know I'm in for treat. Why is baseball so infatuated with phrases like "most feared" anyway? If they had simply proposed to see which batters did the best against a subjectively choosen list of good pitchers, it would have been interesting.

The whole point of pre-free agent arbitration is to give players less than what they would otherwise get on the open market. I don't believe Bush or Gerut will be an acception. If what you say was true, why would the Brewers have even offered them arbitration?

That's because you are thinking like a fan, not like a GM of of a professional baseball team.

Look for players on the market that are still good but might be somewhat undervalued. It's not like the Brewers don't have any positions that could be upgraded. And even if they aren't a bargain, you at least won't have to sign them to a 7 year deal. If Prince is willing to give a hometown discount, great. Otherwise, spend the money somewhere else.

I don't think it matters too much how your talent is distributed on offense. I also don't see how paying market value for Fielder getMs the Brewers closer to the World Series. It's easy to pay market value for wins. Allocating $120 mil to one player seems very risky for a team like the Brewers.

Aside from his two best starts, he had a 5.75 ERA.

Because FIP is fixing 2 stats and not the 3rd, it makes no sense. BABIP and LOB% aren't somehow different than HR/FB. If you are going to look at 3 years of stats I'd either look at some combination of xFIP and ERA or just xFIP by itself. FIP is a stat that makes no sense at all, it is correcting for some factors but not others. Why should I change how HR/FB works with 3 year's data but not how LOB% or BABIP works? Am I to assume after 3 years HR/FB has normalized but BABIP and LOB% still haven't, because that is what FIP is telling me isn't it? xFIP fixes some stats but not others, so by that logic, "it makes no sense" as well. The questions FIP and xFIP asks are: 1. For a given sample size, how representative is the player's stats (walk rate, K rate, etc...) representative of his true talent? 2. For a given sample size, how representative is the league average stat representative of his true talent? Over whatever sample tangotiger used (one year?), FIP determined that only for BB/IP,HP/IP,K/IP and HR/IP, was a pitcher's stats more representative of his true talent than the league average rate. The guys at Hardball Times did some research and came to the conclusion that it was best to us a pitcher's true FB% but league average HR/FB. They used year to year correlations though. What about the 3 prior years to the following year? I'm contending that over a 3 year sample, for a starting pitcher's HR/FB is more representative of his true talent than using league average. If you actually have to regress 51% towards league average, then my contention is incorrect. As for BABIP, I know you need an extremely large sample to even start picking up the pitcher's talent. I would be surprised if you regressed less than 50% over 3 years. And isn't LOB% largely a function of the pitcher's overall ability (component ERA or whatever you use)? I wish I knew the regression factors for 3 years of data. I have seen them for y-t-y correlations many times.

For a single season the HR/FB has such a huge effect on FIP that it just doesn't mean much to me, why would I correct 2 out of 3 high variance stats and completely ignore the 3rd, just makes no sense. Which is why I would prefer xFIP to FIP for ONE season's worth of data. I looked at three years and 546 innings worth of data, however. Your preference for xFIP over FIP no matter the sample is what has me utterly confused. BABIP, LOB% and HR/FB% are all skills, just skills that have a pretty narrow range and that are subject to extremely high variance year to year. Absolutely. That's why you use multiple years worth of data and a regression factor. With HR/FB, for a quick and dirty estimate, you could just use the average of FIP and xFIP over maybe three years. That would effectively regress a pitcher's HR/FB rate 50% towards league average. FIP regresses 0%. xFIP regresses 100%. Imo that is exactly the opposite of how you should look at it. With a large sample you can ignore the context, with a small sample you pretty much have to look at because the data itself is pretty useless without it. What is the context of the last two starts that justifies weighting them less (or removing them altogether)?

It just seems like the author is ignoring a lot of contextual information and just randomly calling it 'luck'. I think the author could have been more clear when he said, " I think I've established that there was practically no difference in how he pitched in his good starts compared to his bad starts." There are many other factors he missed or didn't quite capture perfectly. But I hope we can all agree that what he has captured plays a HUGE role in what makes a pitcher successful. He's captured velocity and movement perfectly. He's also at made at least a crude check on location (down the middle, on the edge strike, on the edge ball). What percent of what makes a pitcher successful do people think he has captured? Less than 50%? About "luck", I think some fans truly believe that if you did know the true odds associated with every AB, you could perfectly predict the outcome of not only that AB but the game. Not even close! For whatever reason, luck is concept that many sports fans are still very resistent to. Some even take it almost personally (see Topper's elequent response above). And even if someone is open to the idea, it's a hard thing to explain. Anything with odds has luck. IT IS ALWAYS THERE, IT CAN NEVER BE ELIMINATED. The odds of a head is 50% but you can roll 7+ heads in 10 flips pretty easily. People know this intuitively when they are rolling dice or flipping a coin. When they are dealing with something with unknown odds (like people), however, they invariably want to attribute too much of uncertainty to the odds. You are still dealing with the same sample error you are with the coin flip PLUS the unceratinty of the actual odds. If Burnett gives up an .800 BA in 10 AB, do we know it was because of only luck? Of course not. What we know is that even a "good" Burnett will still give up 8 hits in 10 AB some of the time through no fault of his own. We can even estimate how often we think he should, based on our estimateof his true talent. If, in fact, he's doing it more often than we think he should, THAT's when we should question whether we've really correctly estimated his true talent. I suggest to anyone who wants to understand the role of luck in sports, you really need to do some research. I'm not suggesting that you should be able to perform the associated calculations. I just thing you would get a better appreciation for what the hell I'm rambling on about if you went online and started reading. Specifically, learn about binomial distributions and bernouli trials. And if you spend 2 minutes googling those phrases and feel compelled to come back here to post, "These are human beings, not coin flips!!!!", you didn't spend enough time looking into it. http://forum.brewerfan.net/images/smilies/smile.gif

The Yankees can crush the Red Sox one day and lose to the Royals the next, so there's obviously a huge amount of uncertainty there. Sure personel changes and maybe a guy is more or less hurt but sometimes balls just find holes or liners get caught. There is a tremendous amount of luck involved. And It's not like we can't estimate the size of the luck compnent. If a batter is a true .300 hitter, we can estimate how often he is expected to have an O-fer game. Fans will say the batter just didn't play well but that isn't necessarily true. He might have played exactly how he always does. Of course, he might have actually not been a true .300 hitter for that day, for whatever reason (injured, pitcher strength, etc...). Observed variance = true variance + luck I agree that sometimes studies are too quick to attribute variance solely to luck. We could assume that the difference between Burnett's results is just luck but that would be wrong. There are many other factors not being considered. I don't think this author even suggests that he's captured everything, however. He's just quantified what we typically refer to as the pitchers "stuff".

What it's showing you is that a pitcher's performance might be very consistent but the results can still vary significantly, from game to game. Fans are seeing binomial variance (luck of the dice roll) and assuming performance variance. With enough starts, much of that randomness cancels itself out, which reduces the binomial variance and let's the performance shine through. Over 30 starts, good performances has a decent chance of rsulting in results. Over 90 starts, its almost guaranteed. It's the signal (talent) to noise (luck) ratio.

Agreed. Now if Escobar was 2 years older and had dominated AAA in 2008, that would have been a different story.

I agree that context matters and I also agree that it would be very helpful to know what the intended location was. But to say that knowing the speed and location of a fastball/changeup or break and location of a curveball tells us nothing is quite the overstatement, IMO. If you knew nothing of a pitcher and were given the choice of either seeing the data presented in this study (20 starts) or just the context, pitch type and location data (20 starts), which one would do you think could give you a better idea of a pitcher's future expected performance? I think I must be taking your "tells us nothing" comment too literally? Game theory is only going to take a pitcher so far. Yeh, a pitcher should throw more balls out of the zone when the batter has 2 strikes but how do you decide what is the "right" pitch in a specific situation?

That settles that.

Because it is the best tool we have for small samples of data and at this point the large sample(his career) doesn't really mean much. What you are suggesting is that best way to predict a pitcher's future HR/FB is to simply use the league average HR/FB, no matter what the sample size is. I don't believe that is correct. I find a lot of fantasy baseball websites talking about how much better xFIP is compared to FIP but it seems like they are always limiting themselves to one year of data (looking for over and undervalued players?). Furthermore, it seems that they suggest there are only two choices; either use the league average HR/FB or use the pitcher's actual HR/FB. The third choice, of course, is use a number somewhere inbetween. I believe that the correct way to project HR/FB is to use a weighted average of the last 2-3 years (most recent years weighted more), adjust for the pitcher's park and then regress some amount towards league average. Maybe you regress 75% toward the league. I don't know. Finally, add the park factor back in. There's some discussion on the topic here: http://www.insidethebook.com/ee/index.php/site/comments/pitching_components/ I think the 2007 Hardball Times Annual had a study that concluded a pitcher's HR/FB was basically irrelevant. I have 06, 08 and 09 but not 07! Anyone have it here? I didn't remove it, I said that given the fact that 2 bad starts at the start of the season is all it took to see a jump in xFIP I'm not going to assume that it is some new trend and not just variance. Trend? If you appreciate the statistical noise associated with small samples, I would think you should appreciate that looking for trends in that small sample won't get you very far. Unless I know for a fact that a player was significantly hurt, I just throw it all into the data bucket. FIP is the weakest of the 3 stats you looked at and his ERA is right where I would project it over that timeframe. You damn FIP because it uses actual HR numbers but ERA is better? It not only uses actual HR rate but also includes the dreaded BABIP!

Some players/people are going to play harder and prepare better when they know that a several million dollar, guaranteed contract is on the line. Now, once that large, guaranteed contract is signed maybe one does not feel as inclined to take extra batting pratice, work out as hard, dive for that ball, etc. Because not many of us personally know many of these players, at times, bringing up a possible "contract year" result, such as I did with Silva, should be brought up. If you were simply bringing it up as a possibility, fair enough. Anything is possible, so it's impossible to rule it out. I just find the whole "contract year" theory pretty unpersuasive in general (beyond playing time) and Silva's 2007 performance as a weak "contract year" candidate overall. If anyone really wanted to get the ins & outs of the calculations, either buy/check out the book, or shoot me a PM & I'd be happy to detail what methodology is given. Consider yourself shot. What is his data set? Is this the study that focused on "prominent free agents"? How did he define that, exactly?

I'm am convinced that if us fans saw every pitch that a pitcher threw in the zone for a year but not their outcome, we would struggle to identify the good pitches from the bad. 99.9% of the fan's observations are results based. I'd like to think the professionals would do a much better job but looking at this makes me wonder.... Of course, this is just one case study.

He hasn't had an xFIP or ERA that high any of the last 3 years so I don't see a compelling reason to expect it to jump that much. I'd put Marquis at something like a 4.70 and Suppan at like a 4.90-5.00 range, a marginal difference. Why would you even feel compelled to use xFIP for a pitcher with such a long track record? Since 2002, the league HR/FB has been 10.7%. Last year it was 10%. Suppan's HR/FB% the last 3 years as been 11.7% and he'll be 35 next year: Age IP FB HR HR/FB 2007 32 206 245 18 7.3% 2008 33 178 191 30 15.7% 2009 34 162 186 25 13.4% 546 622 73 11.7% I would not be expecting a league average HR/FB at this point in his career. While I agree with you to be skeptical of one year's worth of data (ERA, for instance), we are talking about 546 IP over thee seasons. Let's look at the rest of the data: Age IP ERA FIP xFIP 2007 32 206 4.62 4.42 4.90 2008 33 178 4.96 5.51 4.79 2009 34 162 5.29 5.70 4.65 546 4.96 5.21 4.78 Looking at just that, 4.95 ERA projection for Suppan in 2010 looks pretty optimistic. By any measure, he's had to bad years in a row and he is also another year older. Over the past 4 years Suppan has had an xFIP trend of 4.66, 4.90, 4.79, 5.26 and that 5.26 came mostly due to a bad 2 starts to start the year, the rest of the season it was right in the expected 4.90ish range so I'm not ready to assume it is some new level and not just variance. As I'm sure know by now, I'm am not fond of selectively removing data for no good reason. Does it really matter when he had those two bad starts? If one had been in April and the other in August, would it not be OK to remove them, then? It's not like Suppan was pitching through a publicly known injury (at least not one I'm aware of). In the end, your projection of Suppan is your project, so I'm not going to argue that you are wrong. I just thing the statistical evidence you are using to support your opinion is not very compelling.