The Best of Ask David: Volume 2
published November 17, 2004
by David Luciani

It was a popular suggestion last year that I annually compile some of the "best" Ask David questions from the previous year and last year was the first time I did this.  A year ago then, I drew from the complete history of the Ask David series.  This year, I'm selecting only from the 2004 set and I expect to make this compilation an annual occurrence.  To make it into this collection, a question or responses doesn't necessarily need to be one where I gave a response that proved to be correct - that would be too selective and biased in my favor as I'm picking what makes it into the set, though admittedly I do pick a couple that were fun in how they turned out but I'm trying to avoid doing that too much here.  Rather, I'm trying to select the questions and/or answers that either seem to constantly come up again (such as questions where I respond by saying "please refer back to what I said about Ron Belliard in April to see how that sort of thinking applies to this other player) or questions/answers that readers frequently write member support asking to help find because they found them informative about our methods or ideas.  For some reason, a question or response can become so popular that it eventually spins off into its own essay idea and such questions will typically also find their way into this collection.

If a player went out and surprised by hitting 40 home runs for the first time and I had predicted it, I think the reader would benefit from looking back at the reasoning.  Moreover, if I was predicting a breakout that didn't end up occurring, those can be interesting too because we can learn from our mistakes by sorting through faulty reasoning on the projection.

In some cases I've added comments here to put the questions in context or to follow up on the result.  One other change this year is that I also have my "David's Notebook" column to draw on.  Though that feature won't be returning in 2005 in the same format (to be replaced by me focusing on individual issues in essay format more often than in past years), it did offer some interesting or useful comments.  Enjoy!

From January 1, 2004

Q.  This is what has become a "Catch-22" for me in using your projections when preparing for draft day.  I rely heavily on your forecasts but I still like to do my own research.  Let's say you have Rafael Furcal forecasted to have 52 steals.  I do some of my own research and find that Furcal is feeling great and Bobby Cox wants to run him more, etc.  Do I assume my own research is already (or soon will be figured) into your forecasts or do I bump him up from, say, 52 steals to 58 steals?

A.  Firstly, I need to comment on your example in case readers get confused about our Furcal forecast.  We're only forecasting Furcal to steal 28 bases in 2004 in the latest forecast set so I assume you're just writing hypothetically.  I would just hate readers to read your question and then assume we're forecasting Furcal to do something he isn't.

I actually addressed a similar question about manager's comments in previous Ask David's, when a few years ago, Don Baylor said he was going to run Sammy Sosa more and then more recently when Joe Kerrigan, soon to be the ex-Red Sox manager at the time, was going to run Nomar Garciaparra more.  I've said it before and will again:  Do not put much emphasis on what managers say they are going to do unless it relates to the role of a player.  In other words, if Bobby Cox said he were going to move John Smoltz back to the starting rotation, I would have no choice but to consider that but if he said he wanted to see Furcal run more, I would put little weight on that with rare exceptions.  I do remember back in 1998 when Tim Johnson proved in spring training that he was going to have the entire Blue Jays team running more and so I bumped up every appropriate Blue Jay's steal forecast accordingly but it is so rare for something like that to happen.

To answer your more general question, we already do consider all available information when we make forecasts and I can only say that our forecasts are the best we can make based on the information we have available.  I don't know everything there is to know about every player - like whether he had a fight with his wife that will affect his performance - and so it's completely possible if not likely that you could theoretically have access to information that no one else has.

In general, we consider everything that is being said and make our own assessments of what to believe, what to put weight on and so on.  Whenever I get messages from readers who don't agree on a particular forecast for a player, I often remind them that there's nothing stopping them from modifying that particular forecast.  If research is truly to be information-gathering, each reader has to combine all the info they get from various sources, including ours, to form their own opinions about players.  Nothing is set in stone and by all means if you think we're wrong, I recommend modifying the forecast.  However, I do want to respond by saying we consider all such comments when making our forecasts and I already believe that the forecast we publish is at the midpoint of expectations, a phrase I often use to emphasize the type of forecasting we do.

From January 14, 2004

Q. How much does post-season performance affect your evaluation of individual ability?

A. I view post-season performance about the same way as I view regular season performance in that it gives me an opportunity to see an increased sample of a player's skills. An obvious adjustment has to be made not only for the way players are used but for the quality of competition in post-season being higher. It is unlikely to see starting rotations of disastrous quality or pathetic offenses at that stage because such teams rarely make it to the post-season. Therefore, since I tend to evaluate statistics in the context of what sort of competition a player was facing, post-season adjustments are rather difficult to make and are done on a case-by-case basis.

Though the post-season seems long, at most it's only hypothetically possible for a player to play in a total of nineteen post-season games in a given season and that's if they make it all the way to Game Seven of the World Series and if every series went its maximum length. Therefore, it's rare that a post-season offers enough data for me to upgrade or downgrade a player for the subsequent season.

If you're asking whether my Miguel Cabrera forecast is based on the post-season, the answer is that I would have projected him for 2004 almost exactly the same way if he hadn't appeared in the post-season. He's a great player and even with what he did, he's still undervalued by many.

Q. I heard that you once commented about the Pete Rose affair and explained why it can hurt the integrity of the game even if he only bet on his team to win. Can you explain this again? It would seem to me that if you're betting on your team, it can't possibly hurt your team because you want to win that much more.

A. Yes, I commented about this several years ago around the time broadcaster Jim Gray confronted Rose at the World Series and ended up having to issue a public apology to viewers for making Rose uncomfortable at the festive event. At that time, Rose was still denying that he had bet on baseball and a reader wrote me then to ask if it would be acceptable if Rose admitted betting if he had limited his bets to betting on his team to win.

My response then remains about the same. The game of baseball, particularly for a field manager, is about long runs. It's about sometimes juggling your rotation so that the best pitchers will pitch in key games relevant to your pennant success. It's about resting players from time to time so that they perform better for 162 games, even at the expense of today's win. While a manager could theoretically argue that betting on his team does not influence the outcome, the example I used then (and I've heard this was quoted by a reader on our website discussion forum this past week) is that if you have a huge bet on a specific game and you're leading 8-3 in the ninth inning, is it possible that you might bring in your closer to lock down the game, thus jeopardizing his availability for the next day's game? I'm not in any way suggesting that Rose did this and in fact, I suspect that he wanted to win for Cincinnati more than he wanted to win his wagers and so I would doubt that he did manage a different way. The problem is that it does create a reasonable perception of bias in terms of how you approach that individual game, even perhaps at the expense of your team's long run. I believe there are managers who correctly presume that though they could make adjustments in the way they use players to win a particular game today, they are more interested in the long run. I believe that when you bet on a particular game, even betting on your team to win, you jeopardize your own ability to manage a game as though you didn't have a bet on it, even if you have the best of intentions to do otherwise.

On a related note, since Pete Rose was an extremely popular topic this past week in emails to this column, for those of you who have asked whether the Dowd Report was ever published, you'll be pleased to know that the full report, including scans of the alleged betting slips and transcripts of interviews of Pete Rose, is now available online and can be found at http://www.dowdreport.com which is a site published by the former special counsel John Dowd. If you want to get a fascinating look inside this issue, I strongly encourage you to check out this website.

Q. Sometimes I've made some really dumb moves in fantasy baseball and yet other times, it seems I can do no wrong. I just wonder if even you ever make fantasy moves that look really stupid and if so, what's your worst fantasy move ever?

A. We all will make moves that don't work and if we're playing the percentages, we accept the really bad moves with the really good ones with an eye on long run success. I've made some terrible moves over the years but I like to think I've made many more winners. Most of the bad choices that I've made in fantasy baseball at least fall into a category where I can tell you the reason why I made the choice at the time and why I then expected a certain type of outcome. There is a single exception that comes to mind, back in the days when I first started playing in more competitive statistics-based leagues after making the shift from Strat-O-Matic style simulation games. The error came in a published invitational league back in 1994, before I had made the transition from being a "regular" baseball writer to one who also dabbled in fantasy baseball. That year, I had been invited to play in a variation 5X5 league with many better-known baseball writers than myself and celebrities including Bill James, John Hunt of USA Today, Robert Wuhl (the actor who plays Arliss on the HBO series of the same name), John Benson and others. As luck would have it, I had won the first pick overall in the draft and with my pick, and for the life of me I can't possibly explain why, I took pitcher John Burkett, who was coming off a 22-win season with the Giants in 1993. To this day, I don't know what I was thinking at the time as Burkett, even with the wins, had never had an ERA below 3.50 at that point of his career, had only once before topped 12 wins and ended up going 6-8 with a 3.62 ERA in the strike-shortened 1994 season. I can't even tell you what my expectations were at the time because I can't conceive of a single reason that I would ever make such a pick now, even if he had repeated or exceeded his 1993 season. I can only speculate that as the most unknown, at the time, of anyone in the league, I must have been trying to make an impression for being unconventional!

I can't think of a single example that comes close to that miscalculated (or possibly uncalculated) move in the 1994 season but so you know, we all make decisions that later we look back upon and wonder what we were thinking. I'm almost embarrassed to tell you about this one but since I often mention some of my best choices it's only fair to mention one like this.  It's worth a laugh now but I can't explain what I was possibly thinking at the time.

Q. In strong keeper league situations, is it generally better to build up prospects or just use them to trade for established major league players who may cost more budget-wise?

A. I tend to believe in the idea that top prospects are so over-hyped in this era that it is an excellent strategy to stock up on them for the purpose of trading them but it depends on the league. In leagues where you are evaluated based on an active roster and have a second roster or farm system that allows you to store top prospects, I think you should go after the big name prospects that interest you. However, in leagues that require you to keep those prospects active as soon as they make the majors, I am almost always opposed to hanging on to such prospects because there is usually a development time between the day the player arrives and the day he becomes what everyone, including you, expects him to become. By the time he does, especially in leagues that inflate salaries, you're paying for him at the rate you would have had you been forced to buy him in an auction.

There are exceptions such as a Mark McGwire or Albert Pujols but in general, a player arrives in the majors before he is ready. Consider Eric Gagne. Let's say you had known precisely what Eric Gagne was going to do when he eventually became a great pitcher. So, you stored him on your roster. That means you had to endure the disastrous 2000-2001 seasons just to keep him long enough to wait for his great 2002-2003 seasons. I don't find that worth it. Heck, let's say you could go back in time and know exactly what Roy Halladay would be destined to do. Could your roster really have endured his 2000 season? In just 67.2 innings, he allowed as many earned runs that year (80) as it would take to raise a fantasy team that has 1000 innings with an ERA of 3.50 to a fantasy team that has 1067.2 innings with an ERA of about 4.00.

I know this isn't your question but based on other similar questions I receive, I want to caution all readers that our goal should never be to impress our opponents with our vast knowledge of prospects. In other words, many competitors want to be the guy who drafted Eric Gagne when he was no good so that when he becomes good they can look intelligent but that isn't the goal. We're trying to win… we're trying to dominate our opponents. We don't care if they think we're smart or hopelessly lucky or complete idiots as long as we keep stumbling our way into first place, looking as clumsy as possible doing it. We just want to win and I believe that in general, with rare exceptions, these top prospects take time to develop and except for fantasy leagues that give us a place to store them, and I mean store them even while they're active in the majors, I'm not interested in spending too much time stocking up on such players.

So, the short answer is if you can store prospects in a place where they don't hurt your active team, then do that. If you have to put them as active until the day they become great players, I recommend drafting prospects that other people care about and trading them or drafting prospects that you think are ready from square one. I'm curious how many are still waiting for "can't miss" prospect Ruben Rivera to win them their fantasy league…

Q. Do your projected pitcher wins take into account the quality of a team?

A. Yes they do and it's not only based on a consideration of a team's offense but also the estimated quality of defense and the role the pitcher will fill on that team. That's why sometimes you'll see me forecast a mediocre pitcher to have a terrible win/loss record, such as is the case with Nate Cornejo projected to go 8-13 and in other cases, a pitcher who has subpar statistical numbers can have a good projected win/loss record, such as with Tim Wakefield and his projected 16-10 record. If one of these pitchers changed teams, their forecasted wins and losses could move dramatically and instantly.

Q. I have a question concerning your theory about maximizing value with keeper lists. I think the theory makes a lot of sense but how do you factor in draft inflation?

A. What's great about the 78% rule (i.e. you want to have a group of players that costs 78% of their projected value) is that the discount accounts for draft inflation before the draft begins. In other words, we're banking on a bunch of opponents overpaying for players and the more they overpay, the more we can swoop in late in the draft and grab all the bargains. Draft inflation correctly reflects that in the middle to late parts of an auction, there is too much money to spend and not enough value to buy. This is because everyone's keeper list, on average, will probably be getting good players at a discount. Then, in the final stages of the draft, the bargains are everywhere and only those who have managed their resources well are ready to pounce.

As I've argued, we have a specific goal in mind and that's to build a roster that costs about 78% of its real value. We don't care about overpaying because if you track things during the draft as I have outlined before (if you're not sure what I'm talking about, you can read the essay at our website called "On Paying 80% of Projected Value"), what happens late is that you hit a point where yoWanu can now afford to overpay for players. In auctions that start with $260 per team, I have found that this begins happening around the time you get down to $9 bids as the current "high end" of bids. If you've followed the draft diligently and with the methods I've described, you hit a point where now you've got all sorts of money and can outbid everyone by $1 for all the players you're really interested in. I have cautioned before that you do not want to end up in a position where you have any money left over but what's nice is that if you have an exceptionally strong keeper list, you can afford to keep a few players above cost (I explained this in the essay entitled "Perpetual Leagues: Making Your Keeper List" at the site) and if your keeper list was weak, you've probably already spent a lot of your money and are waiting for bargains anyway.

From January 24, 2004:

Q. When you have too many players on one team that played the same position, how does that affect your projections?  Do you deliberately project too many at bats on a team or do you consider where everyone's going to play?

A. When putting together the forecasts I have to consider many factors, such as whether players can move around and play other positions, who's going to play well and who's going to lose their jobs, who might inherit playing time as a result of another player's ineffectiveness and so on. We do consider that there might not be enough at bats to go around at a position and thus, some of the time, you'll see situations where we downgrade a players' projected playing time because there's no room for the player in question to play. For example, Deivi Cruz of the Devil Rays has a very low number of project at bats for 2004 compared to what he had a year ago with the Orioles. This is because, quite simply, he has no place to play as of now.

Projections can go the other way as well. Sometimes, there's no obvious candidate to play a position on a team and sometimes I project that the incumbent will lose his job and another player will step forward and take his place.  A great example of this occurred last year when I projected Desi Relaford to get as many at bats as a full-time player even though he did not start the season with the apparent regular job.  The same situation occurred to two years ago when I projected Geoff Blum to play quite a bit because I was, at the same time, projecting Morgan Ensberg to lose his job in 2002.  This is similar, in some ways, to this year's projection for So Taguchi of the Cardinals, a player whom I don't project to play well but also one whom I expect to find full-time duty out of a utility role, a projection which would be precarious at best as other transactions take place to change the nature of the Cardinals' roster.

Q. Do you think Ken Griffey Jr. can stay healthy this year?

A. It seems that I get asked this question every year in recent years and every year Griffey eventually suffers from an injury that enhances his newfound reputation as an injury prone player.  I'm not going to be the one to project him to return to the days of playing 160 games and when one considers that he hasn't even played in more than 111 games since the 2000 season, even my own forecast of 120 to 130 games could understandably seem optimistic.  That's about what I expect him to play, which would obviously represent an injury-shortened season. It seems hard to believe but the player once considered to be "the kid" turned thirty-four years old this winter. My short answer, and I don't think I'm out on a limb here, is no, I don't think he can stay healthy for the full 2004 season.

Post-2004 comment: Griffey played just 83 games in 2004, even fewer games than I was forecasting.

From February 9, 2004:

Q.  Who are ten players that you foresee having a better year in 2004 than most other prognosticators might predict?  I think Mark Teixeira and Scott Rolen are prime candidates to perform above their expected levels and I would like to know who else to add to this list.

A.  I recall answering a similar question last year and I'll answer it as I did then, not necessarily scientifically but by telling you which players come to mind so far, based on reaction to the forecasts:

1. Byung-Hyun Kim
2. Johan Santana
3. Laynce Nix
4. Brad Fullmer
5. Adam Dunn
6. Jose Contreras
7. J.J. Davis
8. Jason Bay
9. Albert Pujols (I listed him a year ago and still do - hard as it is to believe, he's STILL underrated)
10. Miguel Cabrera (unfortunately everyone knows about him because of the playoffs)

Post-2004 comment: I'm not entirely disappointed with this list.  No doubt, my #1 in Kim proved to be a complete bust but the #2 performance speaks for itself and I was happy with performances from Dunn, Jason Bay, Pujols and Cabrera.  The reader who asked the question did exceptionally well in that both of his/her picks outperformed 2003.

From February 29, 2004:

Q.  My league uses quality starts as a scoring category in place of wins but I don't know of any analysts who do projections for this statistic.  However, since you do project the individual components of a quality start (games started, innings pitched, earned runs), it would be easy for me to calculate the projected IP/GS and ER/GS for each starting pitcher.  Would there be a way to take this information to come up with a forecast for quality starts?

A.  Let me give you some good news and that is that we are considering adding quality starts for the 2005 season.  As for estimating it, here's a rough way to estimate them based on our forecasts.  This isn't perfect and I've already been working on the methodology for forecasting QS for some time but it should at least give you a better estimate that incorporates the way we expect a pitcher to pitch.  As I say, it's not perfect but try this and make sure you use it only for pitchers who have at least one start forecasted:

Estimated QS = -0.24 + (.5413 * GS) + (.1074 * W) + (.0452 * IP) + (-.1094 * ER)

The problems with the formula are fairly obvious.  Certainly, it's not possible to have a "negative" amount of quality starts but we need that modifier at the beginning to get a more reliable result.  As you can see, the formula is much simpler than we could make it.  For example, we might have started by presuming that all shutouts are quality starts to improve the formula.  Basically, the formula is giving you a base that considers games started, wins and innings and then applies a penalty based on the number of earned runs allowed.  Applying this formula to the 2003 pitching population who started at least one game, it had a standard deviation of +/- 1.75, meaning that for about 95% of the pitching population, it estimates the number of quality starts they had within about 3.5 quality starts, which I think is more than good enough for you to use in your league.  One other obvious flaw in the formula is that, theoretically, a reliever who starts the occasional game could end up with more quality starts than actual starts so I suggest you limit the result to be the lesser of games started or the result of the equation.  Undoubtedly, the formula is far too basic and needs to be refined before it could be applied to a successful long-term forecasting model but it's the best I can give you for right now until I focus my attention more on this category.

Trying it on a real life example from our 2004 forecast set (Randy Johnson), Johnson is projected to start 30 games, win 14, pitch 214 innings and allow 97 earned runs.  Inserting all this into the formula, we get:

Estimated QS = -0.24 + (.5413 * 30 Starts) + (.1074 * 14 Wins) + (.0452 * 214 Innings) + (-.1094 * 97 Earned Runs)

And So, Estimated QS = -0.24 + (16.24) + (1.50) + (9.67) + (-10.61) = 16.56 Quality Starts or 17 if we round it to the nearest integer

Post-2004 comment: As suggested then, quality starts will be added to our forecasts in 2005, based on votes taken in the annual reader survey at the end of the season.

From  March 7, 2004:

Q.  My online draft snakes 1-10 and then 10-1.  Is there any benefit to being at the beginning (picks 1-3), the middle (picks 4-7) or the end (picks 8-10)?  The guys that pick at the end always complain about their bad luck so I wondered if it made a difference.

A.  I could see mild complaining with the final pick if you had confidence that everyone drafting ahead of you knew the precise actual eventual value of the players but because we have confidence that we can forecast performance better than our competitors (or we would like to think we do), we shouldn't get too wrapped up in worrying about our draft position.  In fact, for fun, let's say that there are about 230 players to be drafted in a ten team draft and let's also pretend that somehow everyone in the league knew precisely which players would be the most valuable.  In terms of the average quality of player (with #1 being the best and #230 being the worst player drafted), check out the differences between a straight draft and a snake draft, again presuming that everyone knew exactly who would be the best player on the day of the draft:

Average Quality/Rank of Players Drafted in a Perfect 230 Player Draft

As you can see, in a straight draft, it can make a moderate difference in the average quality of player that a team is drafting but in a snake draft, the difference is almost negligible and is certainly offset if a team owner has confidence in his own drafting skills.  The only way there might be an argument about unfairness is if the gap between players is so wide that the top few players (e.g. Alex Rodriguez) make so much difference on their own that they can clinch a league.  I don't believe that's the case in just about any league.  By the way, snake drafts are not only my preference but also my strong recommendation for leagues that don't retain players from year-to-year. 

Q.  One thing I noticed about your forecasts is that, compared to other publications, you seem to put a lot of weight on the previous season, particularly for stolen bases.  For example, I'm confused as to why you have Vladimir Guerrero stealing only 13 bases and Luis Castillo stealing only 26.  Guerrero stole 40 bases just two seasons ago and Castillo stole 48 bases two years ago.

A.  You are astute to notice that for speed, I do put a bit more emphasis on recent performance than I might for other categories but in this case, the Guerrero forecast is more related to how I expect his back to impact his running game.  I have noticed what you're talking about with some other publications (though not all), who seem to use something like a three year average to project steals.  That is extremely dangerous.  As our 2003 projection wrap-up showed, stolen bases are better predicted by the previous season than they are by using a two-year or three-year average but I must clarify that our forecasts aren't just based on averages but rather my own assessment of a player's skills, which are subsequently translated to a statistical forecast.  In the case of Castillo, I quite simply don't believe he runs as quickly as he did two years ago.

Post-2004 comment:  Guerrero stole 15 bases and Castillo stole 21.

From March 21, 2004:

Q.  I checked a sports site which has many proposition bets.  Some of your projections for individual players are significantly different from their over/under marks.  If your projections are correct, wouldn't you be theoretically able to take advantage of those odds?

A.  It's difficult for me to respond to such questions because firstly, I do not encourage that members use our forecasts for gambling purposes because that's not what they're designed for.  There's a pure love of baseball and forecasting behind them.  Having said that, your question is a good enough one that it deserves answering in a public response.  Quite simply, most such bets are based on minimum standards that players have to meet.  For example, they might have an "over/under" on wins for a pitcher but require that the pitcher make at least 30 starts.  They might have an "over/under" for hits but require a batter to have played a certain amount.  Our forecasts are at the midpoint of expectations for not just performance but playing time as well.  In other words, our forecasts account for the percentage chance that a player will have a lengthy injury which is why so few of our projected players will ever get more than about 155-160 games projected.  We have pitchers like Josh Beckett, who are forecasted to get just 10 wins in the latest forecast set but are also forecasted to start only 27 games.  Many such odds would simply call it a non-bet if Beckett started only 27 games.

Anyway, even with my strong recommendation that readers don't use our forecasts for such purposes, we can't control what they do with them so the answer to your question is that anyone who uses a solid statistical foundation and who knows how to recognize overlays in a betting forum does have a betting advantage in the long run.  I just hope readers understand that this isn't horse racing and the forecasts we published are based on our estimation of not only how players will perform but whether they will even play.  Oddsmakers tend to avoid consideration of the second and simply call it a non-bet if the player gets injured or benched.  Beyond that, I really don't want to comment more because it is my sincere hope that people subscribe to our forecasts because they love them or they apply them to more friendly and less stressful forms of entertainment, such as fantasy baseball.

Post-2004 comment: I had used Beckett as the example here and he went on to start 26 games and win 9 of them (at the time we were forecasting 27 starts and 10 wins).

From April 4, 2004 (Opening Day Commentary - was in the form of narrative rather than question and answer):

"Adam Dunn:  I have to start by conceding that a few years ago, I was quite tough on Dunn when everyone else had him as a great prospect.  In fact, back of January, 2002, I wrote the following: "Everyone knows about this guy and as we finalize projections for 2002, we're looking at projecting a struggle for the future superstar.  The Reds would stick with him for at least a half a year, which could be devastating to a fantasy roster if he goes out and hits .230.  He already has the major league power but he strikes out far too often to be an immediate success.  Even if he lives up to expectations, he's not worth what many fantasy leaguers are willing to pay for him.  If you trade him now in your fantasy league, you'll get a player worth what Dunn would do even if he does well."

I quote the Dunn comment because I've done a turnabout of sorts at this stage.  I think the growing pains are over and the player who hit just .215 last year is finally going to be ready to go out and hit in the .260s with 35-40 home run power and plenty of runs scored and RBI, not to mention 15-20 steals."

Post-2004 comment: Dunn hit .266 with 46 home runs, 105 runs scored, 102 RBI but only 6 steals.

"Johan Santana:  By now, everyone knows how high I am on him and while he's coming off elbow surgery during the winter, his stellar performance in 18 starts last year was no fluke.  Even with the team he plays for, I'm expecting he's going to be a bigger winner than most expect and though I don't expect his ERA to be as low as it was, partially because I'm expecting a minor control lapse, he's going to be a huge strikeout guy and even with big expectations, still a bargain."

Post-2004 comment: Santana exceeded my expectations, obviously.

"Jeff Weaver:  His was a very tough forecast to make because you don't know which of two pitchers will show up.  He was disastrous with the Yankees last year but he seems to have a solid grip on his spot in the Los Angeles rotation and because I'm expecting the Dodgers to be good, he's a solid bet for double digit wins.  While I don't see him being an ace pitcher or anything, I could see him landing somewhere in the range he established in 2000-2001 with Detroit, which would be much better than everyone else is expecting and certainly worth having on a fantasy team for cheap.  He strikes me as the kind of pitcher who should fit in well to the National League and become a solid #3-4 type starter for quite a few years."

Post-2004 comment: Weaver went 13-13 with a 4.01 ERA in 34 starts.

"In the American League East, one flip for me is a tough one to make.  Most readers know that I grew up watching the Toronto Blue Jays and though I tend to root for them in the division, and still will in 2004, I can't let my love of the team cloud my interpretation of the player forecasts.  Where back in February, I continued to project a seventh straight year of the American League East coming out in the same order, I now believe the Orioles have moved ahead of the Blue Jays in the division and that Toronto will disappoint, compared to the expectations of many pre-season publications and prognosticators.

Believe me, I'd love to be picking Toronto to do better (and as a fan I hope I'm wrong here) but in particular, I'm forecasting disappointing seasons for Pat Hentgen and Josh Towers and while I like the new-look bullpen here, I'm concerned about the potential everyday contributions from the likes of Reed Johnson, Orlando Hudson and Chris Woodward and I don't believe that either Carlos Delgado or Vernon Wells can repeat the level they set in 2003."

Post-2004 comment:  I have looked back at all the prominent publications to see how they picked the American League East.  While no doubt I got some divisions wrong (e.g. the NL Central), I cannot find any other publication who picked Toronto to finish below third place.  As it turns out, they were even worse than I expected but my reasoning here seems to have been on track.

From April 7, 2004:

"- Thought it was interesting - A reader of my projected list called me a "moron" in reference to my top 200 prospects list.  Leaving Jesse Foppert off the list in 2003 generated similar reaction.  It occurs to me that something seemingly unique to sports, if not baseball in particular, is that it is one of the only fields in which you can be called a "moron" for your views on what will happen in the future, even before anyone knows whether you're right."

From April 11, 2004:

Q.  You have Paul Konerko forecasted to steal no bases but he already has one.  Doesn't that mean your projection for Konerko should have at least one steal in it?

A.  No.  I answer this question every year and it deserves to be answered in this space again because it's important if you're to get the most out of our forecasts.  The projection posted, after the season is already in progress, is what the player is forecasted to do for the remainder of the season.  That means that despite Konerko getting a stolen base, the forecast is saying that we don't project him to steal another one the rest of the way.  It may prove to be incorrect but that's how you should be reading the forecast anyway.  The first steal is in the books.  Our focus is on forecasting the future.

Post-2004 comment:  Konerko did not steal another base the rest of the season.

Q.  I happened to notice that Ron Belliard is leading the league in hits, so I plugged in some numbers and it looks like he's already broken through the upper 95% confidence level for his .257 forecasted average, after only 20 at bats.  Is my math right?  Too early to get excited but I'd be interested if he'd caught your eye as well.

A.  Unfortunately this isn't the case and though things have changed slightly since you wrote the question, I would like to respond.  As of games completed through this past Thursday, Belliard had 10 hits in 20 at bats, which at first glance would seem like it's outside the margin of error but there are several problems.

Firstly, even if at bats were the "denominator" of the equation (they aren't in how I look at statistics), 20 at bats would be too small a sample.  As I explained briefly in the debut of "David's Notebook" this past week, I tend to require at least 30 before I can even begin to consider a sample as meaningful and usually I prefer much larger sample sizes to make meaningful conclusions.  Remember that someone's going to be outside of the margin of error and it's always possible that a player is simply among the small percentage of players who, by luck alone, could be performing outside expectations.

But 30 at bats wouldn't be enough anyway.  In fact, here's how I break things down.  Firstly, I want to know what percentage of time a hitter is making contact.  In Belliard's case, through games completed on Thursday, he had struck out twice in twenty at bats or made contact 18 out of 20 at bats (i.e. 90% of the time).  In the latest projection update published this morning, I'm forecasting his contact rate to be about 84% which, on average over a 20 game span, would be 17 times making contact out of 20.  Therefore, even if the sample size for times making contact was large enough, and it isn't, it would mean that he's performing almost exactly where he should be in terms of making contact.

Then, within times making contact, I consider how many singles, doubles, triples and home runs a hitter gets separately.  The latest forecast set has Belliard expected to get a single about 20% of the time he makes contact, a double about 8% of the time, a triple only about 0.3% of the time and a home run about 2% of the time.

Examining what he's done so far, he has no triples and no home runs.  He has doubled about 11% of the time he's made contact (and he couldn't be closer to my expectations based on the number of times he's made contact).  He's singled about 44% of the time he's made contact, about twice what I would have expected.  The bad news is that in the history of baseball, no one has ever singled at that rate per time making contact over a full season of at least 500 at bats.  In fact, the best mark in history at making contact over a full season was back in 1898 when Wee Willie Keeler singled about 37% of the time he made contact, in an era when distinguishing between singles and errors was still unclear.  In modern baseball, the best rate actually came in the year 2000 when Luis Castillo singled about 35% of the time he made contact and went on to hit .334.  The subsequent season, he hit .263, largely because his singles rate fell back to Earth to about 25%.  In fact, there have only been 26 seasons since 1969 when a hitter has sustained a singles rate per ball in play above even 30% and most of them have been batting title winners and perennial contenders and with the exception of Rod Carew, even the best hitters only did it once or twice in their career (i.e. Wade Boggs only had a rate even above 30% once in 1985, Tony Gwynn only did this once in 1984 and so on).

I say all of this because in fact, the singles per ball in play percentage I'm forecasting for Belliard of 20% is just about at his career mark of 21%, some of which came in Colorado last year when had a 22% rate.  He's made contact 1808 times in his career before this year and I'm not prepared to set aside the weight of that against 18 times making contact this year.  It's virtually impossible that a hitter can maintain a 44% singles rate per ball in play over the long run and last year's leader, Derek Jeter, hit singles "just" 30% of the time he put a ball in play.

The short answer is that so far, most of Belliard's batting average comes from his singles and most of those will not be coming at the same rate in the long run.  I'm hoping that clears up the math of the situation.

Post-2004 comment:  Belliard hit .282 for the full season, including his hot start but he hit just .260 from May 1st on.  We were continuing to forecast a .257 hitter the rest of the way.

From April 21, 2004:

- A reader wrote to say that he does the opposite of panicking by trading for guys that are off to a cold start, which is often a pretty good strategy, but he then added something that was a bit of a fallacy or misunderstanding of the law averages.  This is worthy of either a note in the weekly notebook on in AskDL.

I'll clarify here.  What the reader had said was essentially true and that is that cold starts by players afford smart fantasy GMs to trade for slumping players but he then followed it up with "if a guy is hitting .150 at the end of April who always hits .300 every year, I know he's going to finish with a .300 average by hitting even way better than .300 from May to the end of season because of the law of averages."  It's a common misunderstanding of the law of averages.  The law of averages does not say that probabilities will balance out in the long run by something doing more than it is capable of to make up for a slow start.  It simply clarifies that the larger the sample size, the more a player or a coin flip or whatever will come closer to its real ability than in a shorter sample.  It does not mean that something is "due" to occur because it hasn't yet.  If I really believed that a player is a true .300 hitter and he starts by going 0 for April, I'm going to be projecting a .300 hitter from May on, if that's my belief, and it means that this, combined with his season-to-date miserable April, puts him on target to finish under .300 for the season, even if his real ability is to hit .300.  The best example I can make is that let's say you flip a coin ten times and you get the unlikely result of one head and nine tails.  Now, you're asking me to forecast the next fifty flips of the coin.  I'm still going to be projecting a 50/50 split the rest of the way, if I believe the coin is fair, and what that means is that the implication of my remaining forecast is that the coin will finish with more tails than heads.  You may not agree with the application of the principle to baseball forecasting but I do want to clarify that the "law of averages" does not mean that things balance out in the long run by "making up for" a strong or slumping start.  I don't think I can explain this as well as the many statistical books out there that do (and particularly under the topic of the so-called "gambler's fallacy" which is covered in many books that talk about betting schemes and so on).   Readers who want more info immediately should look up the term "gambler's fallacy" in any of the online search engines to get a great variety of explanations.

- Before the season started, many readers questioned why I was forecasting Morgan Ensberg to play "only" about 130 games.  I have to admit that it's not entirely a belief about what he'll do when healthy because I often tend to discount lesser experienced players more than veterans.  On average, it's the statistically prudent thing to do and history has shown it.  It accounts for the potential for slumps in inexperienced players and/or the potential for injury the first time a player goes through the long haul of a 162 game season.  In so doing, the margin of error on such players is significantly lowered.  But in Ensberg's case, there is another factor and that's the Jimy Williams factor.  I think Jimy is actually a better manager than some give him credit for but he does seem to rotate players through the lineup when there's no one present at a position who has several established years in the books.  Ensberg had a great year last year and he could easily exceed the number of games I've forecasted but thus far, his struggles so far may be just the sort of thing that will cause him to achieve the discounted number for inexperienced players.  Of course, he's a far superior player than the one we've seen so far and in a limited sample, he's showing much more of an ability to take walks than we're used to.  I just couldn't bring myself to forecast more games than I did because I thought to do so would be a disservice to our readers.  Job security plays a big role in the forecasts, particularly under the games projected, and despite a great season last year, I don't see Ensberg's job security as absolute here if he continues to hit to the tune of a sub-.200 average.

From April 25, 2004:

Q.  I was looking at your projections and specifically, the relative confidence ratings you publish.  Instead of doing an awesome season projected for Albert Pujols, with a relative confidence of only 25%, why not publish something lower (like a .330 season with 35 home runs) and a much higher relative confidence that is less of a gamble?

A.  I've done my best to explain relative confidence in the past but I guess I do no need to return to the topic as I've faced several of these questions since the last time I talked about the RC rating.  Firstly, I can't publish a forecast with any higher confidence.  I've used all the data that's available and based on the amount of the data, the consistency of it and other factors that influence my confidence, I assign a relative confidence rating after the forecast is created.  For example, my Opening Day forecast for Jose Contreras, like last year, has a very low confidence rating but I still made the best possible forecast I could make and then assigned a confidence rating to it after that.

The mistake many readers seem to make is that they look at great projected seasons like the one for Pujols, Contreras or Byung-Hyun Kim and they falsely believe that I could assign greater confidence to it by bringing the projected numbers closer to the typical or average player, making it (as you well put it) less of a "gamble."  This is not the case.  What I'm publishing really is what I believe the midpoint of expectations is for the player.  The confidence level does not mean that the player is less likely to achieve a high level.  Rather, it means that the margin of error is much wider.  That is to say that if I publish Pujols to hit in the .350s, it's an inconsistent .350.  In fact, on Opening Day of 2003, I published a low confidence forecast for Pujols - with an RC rating of just 26% - that had him forecasted to hit a career-best .334 with career highs in home runs (38) and walks (73).  Interestingly, I addressed a similar question then but in particular, a reader with whom I carried on a lengthy email exchange said he couldn't understand how that could be the midpoint.  He suggested, as you do, that I could create a higher confidence forecast by lowering the season more to the levels Pujols had established and I recall explaining to the reader that the margin of error was wider and that Pujols could actually exceed the expectations by a wider margin that a low-confidence player would.  Of course, Pujols actually did end up finishing on the high side of my expectations, hitting .359 with a career-high 43 home runs and 79 walks, all even higher than the career-best marks I had forecast.

I really do believe when I publish a forecast that a player is as likely to exceed the projected mark as he is to under-achieve it but to demonstrate this, let's take a look at the history of "high performance" forecasts.  I'll use the hitting side here of players forecasted to hit better than .300 as an example.  Let me show you the typical deviation.  I started publishing confidence forecasts here only in the past few years but as we were building up the system, we maintained them for two full years prior and so I'm including those in the analysis as we ended up using the same methodology as we use now.  I am not including a confidence model that I rejected after a test-run (these weren't published anyway) in the 1998 season.  Even including every possible season, the data remains quite limited here as there aren't that many plus-.300 seasons projected each season but this should make my point anyway:

Several things jump out when you look at the table.  Firstly, notice how all of the averages are right about where they should be and in fact, the lowest confidence forecasts of plus-.300 hitters, on average, actually slightly exceeded the expectations on average, though the difference is statistically insignificant.  What's more interesting, as I've been saying, is to look at the standard deviation.  I don't recommend you use this history of standard deviations just yet to make statistical conclusions about "low and high end" expectations for players but look at how neatly the standard deviation increases as the confidence value gets lower.  In the case of the relative confidence forecasts with a rating of 80 or better, the standard deviation is just +/- .007, meaning that about 95% of these high-confidence but all too rarely published forecasts, historically, have landed within about 14 points in either direction of the projected batting average.  On the bottom confidence group here, the standard deviation is extremely wide of +/- .031, meaning that 95% of the forecasts, historically, have landed within about +/- 62 points of the batting average, wide enough that you can run into real problems if you hit one of the bad ones.

I think the results here would have been even more consistent if the relative confidence rating was not assigned to the whole forecast.  In other words, I have contemplated publishing ranges or RC ratings for each category because sometimes, I'm much more confident of a player's stolen base projection than I am of his batting average projection and thus far, the RC rating hasn't captured that.

Anyway, the short response to your question is that I can't raise the confidence any higher than you see.  I've made the best possible forecast that I can for each player and while some low confidence forecasts may look extraordinarily good or bad, they really are at the midpoint of my expectations and statistically, the case has been that the players perform, on average, as the forecasts say they will.  Unfortunately, the lower confidence ones have a much wider margin of error, as you can see from the chart above.

Q.  I'm in a 5X5 keeper league and from what I remember, you are not a big Jeremy Affeldt fan.  I've been offered John Smoltz for Affeldt.  What do you think of the trade for the long-term and the short-term?

A.  For the short-term, my own forecasts give the obvious answer that I like this trade as I continue to forecast Affeldt as a pitcher who will perform below expectations this season and who will eventually lose his rotation spot.  Even for the long-term I like it because if Affeldt struggles as much as I expect him to in 2004, then he'll be easy to re-acquire in next year's draft and then you'll have Smoltz on your roster too.

Sometimes, it's tough to trade away a prospect and I know that for about one in fifteen of them, I'm going to be way off and they'll turn out to be something extraordinary that I didn't expect.  In Affeldt's case, I don't see this happening and I think that if you got Smoltz this year, you'd be a much stronger team and even if Affeldt turns out to be as great as everyone expects, you would have gained substantially more in this deal for the current season, enough to offset the unlikely scenario that has Affeldt going on to be Cy Young a few years down the road.

I might be wrong about Affeldt but I just don't see what everyone else sees.  His perceived trade value will have gained a boost as he threw seven innings of two-hit ball the other day.  Even if he goes out and has an ERA in the 3's over a full season, which I'm not projecting, I still like Smoltz better for the 5X5 league anyway which means there's almost no chance you'll be kicking yourself unless Affeldt wins 20 games or something surprising.  Heck, I'd probably trade the imagined/projected "living up to his potential" version of Affeldt for Smoltz so that should make your decision a bit easier.

Post-2004 comment:  Affeldt ended up hurt but did get a brief chance to be the closer for Kansas City.  Regardless, Smoltz obviously ended up having the superior season.

From May 2, 2004:

Q.  I was wondering if minor league performance to date influences your major league projections?

A.  Yes it does.  As you probably know, we publish weekly minor league translations for every player from Single-A on up and we are always watching for players who could get a call-up.  Also, the actual performance itself, just like major league performance, influences the direction our forecast takes.  That's why a player such as Edwin Jackson (who was downgraded to a projected 4.47 remaining ERA, almost a half run higher than in the previous issue) can see such a change in his forecast even though we haven't got any new big league data.  Basically, his minor league performance has caused us to bump up his forecasted ERA in the majors.

Q.  A trade just went down in our league that was completely one-sided.  An owner took advantage of another by trading away a couple of fast starters for a superstar that is starting slow.  I officially protested this trade.  Our league voted on it and approved it because they saw no collusion and saw it as only taking advantage of "selling high."  There is a difference between a savvy deal and one that is outright ridiculous.  Don't you think that protests should not only prevent cheating but completely lopsided deals?  That was the real point behind my protest.

A.  It's tough for me to answer this because I know you want me to agree with you here but actually, I believe that as long as cheating doesn't occur, if an owner is capable of "taking advantage" of another owner, then that's part of the game.  Moreover, and I know you say it's lopsided but consider that the owner who's being exploited does have the right to make seemingly lopsided trades against him because maybe he's sincerely convinced he made the right deal.  Part of the great joy of fantasy baseball and clearly part of the challenge is that everyone has varying opinions about players and it's the gaps between those opinions that generates trade activity and ultimately determines who wins.

Every league is different and my only recommendation to you is that if you think this owner is, to steal a term from poker, a so-called "fish" then by all means you need to go fishing yourself.  Obviously if collusion is the only grounds upon which your league will overturn a trade, then you need to get on the phone with this inferior owner and start hammering out a series of deals.  You should be able to learn how he thinks by the deal you mentioned - start offering him all our fast starters for guys whom you see as valuable and who are off to a slumping start.

I've been in this boat several times where other owners thought that I was the one being taking advantage of.  For one example and a lengthier answer to your question, I encourage you to refer to my essay called "On Fantasy Trading Rules" (republished February 14, 2004) that's available from the home page of our site.  Another example that was not mentioned in that essay but is similar was where I helped mediate a trade in another league.

One owner had tried to trade Randy Myers in the middle of the 1998 season when he had about 25 saves for Toronto at the time, if I recall correctly.  He was trying to deal Myers for Phil Nevin.  At the time, Nevin had four part-seasons in the books and again if I recall properly, was hitting around .220.  He had never shown any real power and was not getting any playing time.  The owner who protested the trade had presented his case that it was "obvious" that it was a lopsided deal and said something along the lines of "Nevin is already twenty-seven or twenty-eight years old" (I don't recall which) and that "therefore he has no prospect upside."  I recall that he also pointed out that Nevin, who should already be in his prime, was a career .230 or .240 hitter at the time and didn't even seem to have 20 home run potential.  The owner acquiring Nevin made his case to me that he saw Nevin as a future 30+ home run man and this same owner was forced to publicly state that he saw Myers as a guy on the fast decline.  After reviewing the two team cases, I allowed the trade.  I believe that Myers went on to have only three or four more saves in his entire career and within two years, Nevin became everything this astute owner had said he would be and more.  In terms of what makes a deal "ridiculous" (as you called it), where we do draw the line?  What criteria should we use to determine whether a trade meets that standard as opposed to just being "apparently lopsided but actually not?"  I won't pretend to be able to offer the answer to that question so perhaps I can't help you but I do think if you're to come to peace with this issue, you may want to pose these questions.

I know it doesn't always work out like this but deciding what's a good trade is not always black and white and fantasy GMs have the right to make "dumb" moves because many of them turn out not to be so dumb when we look back at them years later.  I understand your frustration and have shared it, on occasion, in my own leagues.  But you asked my opinion and that is that quite simply, we have the right to be foolish as long as we don't cheat.  No one has the right to tell another owner that they are better at assessing how fair a trade is because the gaps in perceptions about players or their eventual value to your fantasy team is exactly what this game is all about.

Q.  You have Ichiro projected to have a .329 average the rest of the way.  Consider that since June 16th of last year, he's hit .287 and since July 9th, he's hit just .265.  Any way you look at it, if you're using a 95% confidence model, your projection would seem to be a bit too optimistic.  What confidence level are you using?

A.  There are several aspects to your question I need to address.  We "average" out at a 95% confidence model but that doesn't mean we use a fixed statistical model only.  My own scouting observations of a player do come into play and in some cases, I use as high as a 99% confidence level and in other cases, I could use less than 95%, depending on the category and how specific I need to be.  For something like triples, a 95% confidence level works extremely well but for something like singles, I need 99% confidence or even more.  I rarely go below 95% in terms of margin of error analysis because too many players would be falsely flagged as a concern but I have done it, on occasion.

It seems that you are looking at the statistics on a "per at bat" basis, which I don't agree with doing.  First, I'm forecasting a player's ability to put the ball in play and then within that model, I'm looking at the percentage of times he gets a single, double, triple and home run per the number of times he puts a ball in play.  I refer you to my answer in this space re: Ronnie Belliard of a couple of weeks ago if you want to see the actual breakdown of the sort of way I take apart the performance.  In fact, Suzuki's projected remaining projected average is well within the margin of error when you look at it that way.

Beyond that, I don't simply limit myself to the most recent data, even though it carries greater weight and it seems that starting at June 16th of 2003 would be deliberately picking the most pessimistic starting point for Suzuki that we can find.  For example, if I start at June 1, 2003, Suzuki was a .309 hitter since then before the start of this season.  If you look at the rate he was making contact even as late as starting after the All Star break, he put the ball in play about 89% of the time after the break compared to about 91% of the time before the break, which is clearly the same player contact-wise.  On a per ball in play basis, his doubles rate was the same after the All Star break (about 5%), his triples rate was even better (about 2% compared to 1% before the break) and his home run rate was about the same (about 2% in both halves of the season).  His fall-off in average was due to the number of singles he was getting per ball in play which was an incredible drop from 42% per BIP to 21% after the break.

What's important to realize is that a 21% singles rate per ball in play is simply not indicative of the type of player we're dealing with here.  In fact, over his career including 2003, Suzuki's rate of singles per ball in play prior to 2003 is about 28% and so, a 21% performance over a half-season of just over 250 balls in play is entirely within the margin of error, though clearly on the low end of the margin.  Using a 99% margin of error (which is what I tend to use for the singles category), if he were really a 28% singles player, Suzuki could be expected to have a singles rate over that span of anywhere between about 20.8% and 35.2% and so, though he's on the extreme outer edge of the margin for singles, he isn't enough outside of it that it forces me to make an adjustment, particularly since I'm not going to confine myself to using only the second half of last year (in which he hit just .259) to make my conclusion.  I'm just saying that if I did do that, he would still be within the margin of error for the type of player I forecast that he still is.

The way I look at it, his slow start did cause me to downgrade him a few percentage points but to answer your question, we continue to use a confidence level of at least 95% on most statistical factors but I tend to use 99% on categories such as singles and walks and Suzuki's projection is within the margin of error there, even considering what he's done since the dates you picked.  Quite simply, I see the same player getting different results and so I'm sticking with my expectation that his singles rate per ball in play remains around 28%.  His contact skills haven't shown any decline at all, even in the period you mentioned.

Post-2004 comment:  Ichiro even exceeded expectations but did end up justifying my continued faith in him, even when he had gone a year without looking like a plus-.300 hitter. 

From May 12, 2004:

"I finally downgraded Barry Zito and though I've never been as high on him as many others, I thought he was a much better pitcher than this.  In particular, there are two primary areas of concern and then a few lesser ones.  The first, and most obvious, is how slow he's looked compared to the pitcher we saw in 2002.  Yesterday, I compared some tapes of a few games from 2002 to two games in 2004 and his fastball is straighter and doesn't have that same pop it used to have.  Admittedly, his curve ball has been right where it always was but what I've noticed, in these 2004 starts I've watched, is that Zito always seems to pitch as if he's tired, even at the start of the game.  More importantly, and what started my downgrade even a week ago, is that Zito is consistently leaving the ball up in the strike zone and even when he's at his best, any of his pitches can be hammered there, which they have been so far.  He's already allowed about half as many home runs as he allowed all of last year and he's pitched only about 15% of the innings.  I think more of these balls are going to start staying in the park for doubles and such, as Zito irons things out, but it's passed the point of being coincidence or bad luck.  Where I think we're going to see a shift is that he may start allowing fewer home runs but he'll continue to give up hits at a an alarming pace, which is where readers, no doubt, will recognize the shift in my latest forecasts that caused the projected ERA to balloon.  Zito, so far, isn't the same pitcher.  What I'd have to see for him to turn it around is not just a couple of good games but individual efforts that show that his pitches are hard to pull and that he can return to pitching inside effectively.  Particularly since he's been missing high and away, it's been tough to tell because even at his best, he's never been a true control pitcher and so it's difficult to determine whether he has a new approach or if his typical wildness is causing him to miss away.  I do think he's overusing the curve, even though he commented after one game that he hadn't thrown that many."

Post-2004 comment:  I list this to show the reader what I was thinking when I make a serious downgrade, as I did with Zito.  Zito didn't end up rebounding and if anything, I was a bit slow to react here but he deserved a long look.

From June 6, 2004:

Q.  I made a trade about three weeks ago in which I thought I was making out greatly by taking advantage of an owner who was down on Carlos Delgado.  Now I'm not so sure.  Should I be trying to unload Delgado or is he likely to have a strong second half?

A.  I'm not sure if in your question, you're asking about Delgado's ability or is health.  In terms of his ability, there's no way he suddenly fell into a decline and that means that the .227 average he boasts as of the time of my response here isn't a real representation of his ability.  Even his power numbers are off and in cases like his, where there's no reason to believe that he'll sustain such a low performance, you can never get fair value for a player like him because his market value drops in line with his performance.  If nothing else, if you feel forced to deal him away, at least wait for Delgado's next inevitable "hot streak" so that the receiving owner believes he's turned the corner, which the next hot streak may very well represent.  Delgado this season has played through rib cage problems and knee problems and a healthy Delgado shows a consistent track record.  I don't believe his MVP-type 2003 season was representative of his real ability as it was a bit on the high end of expectations but his 2001-02 seasons show us a consistent 30-40 home run hitter with 100+ RBI and 100+ run scoring ability and a decent average much higher than we've seen this season.  Look for him to turn things around soon as long as he can get healthy and I always discourage readers from dealing away players when they've been playing through injuries because you can never get fair value for such players in a trade.  My latest forecast has him hitting .291 the rest of the way with 27 home runs and 67 RBI.  No one's going to give you that kind of value for him if you try to move him now.

Post-2004 comment:  Delgado did have a great second half and recovered to hit .294 the rest of the way with 24 home runs and 67 RBI, almost exactly what we told the reader to expect here.

From August 25, 2004:

- On the baseball analysis side of things, I've spent a great deal of time lately revising our age charts to account for the effects of age on each pitching category.  It wasn't intentional but in previous years, I know I gave hitting more attention when it comes to discussing the effects of age and pitching has been interesting me much more lately because some call it volatile where I argue that, particularly in the case of relief pitchers, it's just that the sample size in a season is so small that we can't know whether most relievers even performed up to their abilities, even though there are some so remarkably consistent that we can't say anything but how good they are.  It does seem that the better a pitcher is, the more consistent he will be, which goes in hand with statistical theory.  An event that has a 10% chance of being successful will consistently give you failures try after try whereas an event that has a 40% chance of being successful can give you an assortment for a long run before you have any sense of certainty that you were right in the first place.  In other words, Eric Gagne and Pedro Martinez would have trouble having a bad season even over a limited number of innings but Esteban Loaiza could easily go out and do something entirely the opposite of what you expect over 50 or 60 innings.

In particular, I've focused recently on strikeouts as an age analysis category because not only are they pivotal to a projection but each time we do the analysis, it continues to support the idea of a constantly-declining ability in strikeouts as a player ages and not the more conventional peak that comes in a player's mid twenties.  It seems counter-intuitive to common sense that strikeouts would be in constant decline as most would expect that a pitcher becomes stronger in his mid-twenties and thus throws harder.

Because I've been so interested in this, I returned to update an old study that had considered data from the 1969 season through to the end of 1997 and I recently updated it to include consideration of new data through the completion of the 2003 season and also for the first time ever to incorporate translated performances for the minor leagues.  The analysis now also reaches back into the past of baseball history by adjusting for yearly strikeout totals.  All of the stats in the analysis are park-neutralized and league-neutralized and only players who have played a comparable amount in each of two seasons get considered.

What surprised me is just how straight the line is getting on the strikeouts analysis.  The declining line didn't change much from before but it's straightened quite tightly now as more data gets added to the set.  Including minor leaguers and big leaguers, there are now about 15,000 player performances considered for which there was plenty of data for two seasons and where the data remains limited now is only around the ages of 18-21.

The new analysis yielded an even more level line than we saw a few years ago and basically, and unlike any other pitching category, the pitcher's strikeout rate per batter faced (when we exclude walks from batters faced, which was something I've explained previously) was in constant decline with the exception of around his late teens.  What we saw was that a player would see a significant increase from turning eighteen to nineteen and that he would then begin a constant and steady decline which becomes much sharper as he reaches his late forties.  Here's a graph of the results which compares strikeouts per batter faced (excluding walks) of more than 15,000 neutralized performances compared to their rate of a year earlier (accounting for league, park, batters faced and level of competition):

The starting point in the graph has been selected simply so that we achieve an "average" of 1 by the time the player has reached 29-years-old but it has absolutely no bearing on the shape of the line and we could have started anywhere.  Basically, to read the chart, it's showing you relative skill in this category for an age.  That means that at the age of 18, a pitcher strikes out about 1.2 batters (that he does not walk) for every 1 he would strike out at the age of 29.

The line is so straight that it's remarkable, particularly where there are plenty of players in the analysis from the ages of 22 to 45 and this is the only such declining skill among all the pitching categories analyzed.  In the case of walks per batter faced, we see constant and steady improvement throughout a pitcher's career until his forties and we see such improvements in other categories such as the pitcher's ability to keep the ball down in the strike zone and so on.

Just for fun, I took a few well known pitchers who are perceived to have achieved their prime later in their career to see what they were doing differently.  There were some clear exceptions, such as Randy Johnson, whose strikeout ability per batter faced (when we exclude walks) has risen steadily since his mid twenties from a rate of about 30% around the time he was twenty-eight to a peak of about 40% by the time he was thirty-seven.  His is a particularly tough one to examine because of the effect of moving to the National League but unadjusted, it's clear he's an exception to the rule even though his data is included in the analysis.  What I did find, though, is that many pitchers who became successful later in their career can and do see the typical drop in strikeouts and what's moving them into the success zone is the sudden arrival of their control and in turn likely as a result, an ability to keep the ball in the park.

I'm not a complete believer in DIPS theory because I feel that pitchers can and do control what type of ball gets hit against them (e.g. Eric Gagne is excellent at making batters hit a ball weakly because they don't make square contact against him and thus don't hit a level ball) but for those who put a lot of weight on strikeouts, the history of baseball pushes us in the direction of speculating that strikeouts may be in constant decline.  This goes against both my instinct and conventional theory so I'm not going to be quick to accept the results of this new analysis though the line is so clearly straight that it makes one wonder how this could be possible.  I'm just noting it here to share with readers and am not making a conclusion at this point.