The "Rebound" Theory

[da_big_kid_94]

I've been kicking this idea around for a while, but haven't been able to word it exactly how I would like - so I have to fall back on a poor technique - demonstration via anecdote. To wit - what criteria do you use to categorize players when it comes to consecutive diametrically opposed offensive seasons? Are they guys coming off career years who should never be expected to put up numbers like that again or guys who just had an off year and should be reasonably expected to come at least close to their performance two years prior?

I present this blind stat line as an example:

G AB R H 2B 3B HR RBI TB BB SO OBP SLG AVG
2006 87 237 32 67 13 2 9 33 111 17 58 5 8 .328 .468 .283
2007 140 505 86 149 33 9 24 81 272 36 99 .353 .539 .295
2008 157 612 76 164 45 6 20 91 281 27 109 .300 .459 .268

This is a young player who probably went very high in drafts or was a very expensive player at auction. I realize that you probably feel you can't make the call without knowing who he is, or even who he plays for - which is maybe the point. I'm willing to bet that the projections for this player for the 2008 season were significantly higher than what he actually produced (I know - it happens). His HR and RBI numbers are very similar and there's not a real big difference in BBs and Ks between 2007 and 2008. Yet his average drops 30 points and his slugging and OBA percentges drop through the basement.

Now ...you look at these numbers in the raw - what is your first impression of this player? You may change your mind after finding out who he is - but, for now, you don't know. Is he a candidate for a rebound or did he have his career year early in his career? And most importantly - why did you categorize him as you did?

[ShawnC]

I certainly don't like the lack of walks, especially the downturn they took last year - 27 walks in over 600 AB's is putrid. There's some other things I don't particularly like about this player, but for the sake of keeping his name a mystery, I'll leave it at that for now.

My guess for 2009 would be a line of .270/.325/.490 25 HR 90 RBI 110K 40BB

I can see him having his career year within about 2-3 years, something like .290/.350/.570 w/ 35 HR 110 RBI, but I see him as somewhere in between his '07 and '08 seasons for most of his career.

[Todd Zola]

There is a significant difference between walks from 2007 and 2008, which is a pretty big aspect to the analysis. Truth be told, the majority of his walks in 2007 came in the first half, so his discipline took a downturn for the past year and a half.

Last season, his BABIP dropped, which along with fewer walks leads to the big drop in BA and huge drop in OBP.

From a pure fantasy sense, this player is going to be drafted "high" for his SB so looking at his draft position may not be the best indicator to address the question of what will he HR and BA be this season?

From a strategical sense, he's the type you will not have to pay full value for anymore, or probably not anyway as his stock has dropped in terms of the perception of his value. And he does have the ability to bounce back in a big way.

We've got him for .275-21HR. Probably not good enough to wear sunglasses at night. But add in 21 SB and he's a top-60 or so player, 4th-5th round in 15-team, mixed, 5th-6th round in 12 team mixed.

[Guest]

You're asking a more theoretical question, so I'll answer that and ignore one of baseball's most likely candidates to star in a remake of License to Drive, Blown Away or Lost Boys.

Anyways - we use three year averages, weighted, with aging adjustments and including minor league translations with more weight placed on MLB stats.

I actually think this player has a pretty stable stat set by our methodology, has three year translated BABIP of .337, .332, and .299 but has some fluctuations in K, HR and BB rates to go with it, you're looking at probably a guy who you'll hedge to the 3 year weighted mean and go from there, with a pretty safe bet that if he stays healthy you recoup the investment.

Im not sure where else you'd go but kinda hedge the middle here - really, if you include his minor league and major league history (which we do) he's not really all over the place.

But general rule would be to hedge towards the mean in an outlier year situation unless there was a concrete reason why not to take it into account. For example, a hitter who had a busted knee in 2006, didn't steal bases that year, came back healthy and for 2 years stole 50 bases each year, it would be inappropriate to somehow consider the 0 SB in that year, given two years of proven health. But thats an extreme example, in most cases you have to hedge or else you're out on a limb somewhat. (which is why we provide profiles as well)

[Mojo Jojo]

see i am not so sure that there is a rebound here and using a 3-yr weighted average may not give you an accurate projection.

first off, i think todd's point about downturn in walks being a big part of this analysis is huge.

the walk-rate difference between the first and second half would have been a red flag going into the third year and after seeing the results of the third year, i would be more inclined to believe this year's projection should be quite similar to last year's stat line. obviously, there are tweaks that should be made based on other analysis and his stat line could change if he becomes more selective/takes more walks. but, he would have to prove that he can take more walks before i believe it to be the case. one piece of the puzzle (for me) would be his historical walk rate.

but...on the positive side

looking at the overall power numbers, true the hrs dropped but doubles increased. often times it is just a fraction of an inch that keeps a ball in the park or not. another key to whether those doubles revert back to hrs is the hitter's fb/gb rate and fb/hr rate (again historical averages can provide trends). also, the 30 point drop in batting average is not outside the normal range of statistical variation. so yes, while our hitter suffered a batting average drop there is nothing that statistically says that can't change. his walk rate obviously plays a role here, but statistically speaking a batting average rebound could occur without a change in walk rate.

why are all my post centered?...i would rather have them left justified

[da_big_kid_94]

I see some of the questions posted, and they seem to beget more of the same;

• Is a - 9 walk total and a + 10 K total really all that significant here? I understand it was done over an additional 110+ ABs, but it's not as if he were very selective in the prior year. He had an additional 107 ABs in 2008, but failed to put the ball in play only in 19 of those ABs - yet his OBA dropped fifty points. He wasn't a very selective hitter in 2007 either but his value for drafts/auction really went through the roof for 2008. It would logically follow that his BABIP had to take a nose dive in this situation - but was a matter of hit luck in 2007 or had he regressed to his mean in 2008? Again, anecdotal - but when one first looks, did you categorize him as a one year wonder or someone due to bounce back?
• To be honest, I deliberately omitted his SB totals - because I felt that they would be misleading in this context - if people had seen the SB numbers, they may have been unduly influenced towards his roto categorization rather than his real world categorization. As an aside, to me, it may be an anomaly that he stole as many bases as he did last year considering how often he failed to reach base, relative to the 2007 season.
  
• Edit: Knew there was something I forgot - only 4 less HR than the prior year, and that's offset by 12 more doubles than in 2007 - yet an 80 point drop in slugging? very odd looking

[ShawnC]

• Is a - 9 walk total and a + 10 K total really all that significant here?

Yeah, I think it is. Given his walk rate the prior year and his minor league track record, we should have seen 45-50 walks in 612 AB's during 2008, but he only got 27 - that's more like a difference closer to 20 walks, not 9, especially given the fact he should be improving (or at least maintaining) at this stage of his career, not taking steps backwards. Todd's point about the second half magnifies that concern. Young players go through their ups and downs, but a tailing off of plate discipline is something that raises some significant concerns.

[Todd Zola]

Once our projections, profiles and corresponding description of the methodology comes out, I am sure we will have a plethora of interesting discussions with respect to the merit and shortcomings of the way we do things and the way others do things, and quite frankly, that is the major reason was I decided to leave Fantasy Planet and team up with Gary on a private site.

We have a developed a highly rigorous, objective means of looking at a player's anticipated performance. That said, it allows for some subjective seasoning, but the beauty of the method is in order to alter what our translations say 'should' happen, we need more than some anecdotal evidence or gut feel. Some may argue that and they will have their forum, trust me. Gary mentioned discounting injury years, that's one means of adjusting as how much to discount is subjective. Looking at how "lucky or unlucky" a player was is another, and can be accounted for by regressing BABIP or HR/FB more or less than our translations say they should -- again, how much is subjective. Deciding if a player owns a new skill or has lost a skill more or less than would be attributed to normal aging is another way to alter the projection -- you guessed it, how much is subjective. A great example of how I failed at this is last season, I "predicted" that Ian Snell's improved walk rate was a new skill and he would not only keep it up, but further refine it. I also felt that Casey Kotchman's hitting more flies would carry over and an increase in HR would ensue

[cwk1963]

I have him pegged for a bit of a bounceback or rebound. He's still young enough to learn to make adjustments and I think he will. I'm looking at him as somewhat of a bargain potential - not a few dollars but less than what he was projected for last year.

[JP Kastner]

For a player like this you have to look at context and not just statistics. Was he hurt? Did a situation change?

I know this player well. At the All-Star break he was hitting .289/15 in 357 AB. Stretch that over his total 612 AB's and he would have hit 26 HR. Digging deeper he was doing just fine until a brutal September in a playoff run. That month he hit .173/0 in 98 AB.

What I see is a quality young player who in the heat of a playoff run was pressing. His whole team was pressing. He probably needed a day off or two, but as the team was in the fight of their playoff lives, they kept putting him in the lineup.

I remember looking at his face with runners on second and third and two outs and it was clear he didn't want to be there. He swung hard at the first two pitches but missed. After that he tried to protect the plate as he was behind in the count and popped up to the shortstop. The crowd, tired of years of loosing seasons and frustrated with their team fading at the end booed him mercilessly.

As he is young, I'll assume he learned a lesson last year and pencil him in for his first half, .290/25, but I'll make a note to watch him carefully. He doesn't do well in pressure situations, so I may pull him in September if he is in the same situation.

[da_big_kid_94]

First, let me thank you all for the replies - there has been some very enlightening information posted. It seems that, from some of the responses, some of you had to find out it was Corey Hart before making the call... and I guess that was my point. What I was really asking for was to take his numbers in the context presented - raw and unvarnished without any mitigating circumstances - and let me know what your very first impression of this player was - career year or rebound candidate - and why you categorized him as such.

I do the same thing - I have to know who he is before justifying my decision - and in our game of roto, I understand the importance of that. But I seemingly come upon an interesting juxtaposition - in our game, it's the stats that matter - but I have to know who put them up in order to make a decision on him? Why? Because if I know it's Corey Hart that put up those numbers, am I more likely to say that 2008 was the anomaly rather than 2007 because of his youth and the Brewers in a pennant race? But if I take Ryan Dempster's numbers and put them in the same meat grinder, I'd say that 2008 was the career year and he can't hope to put those numbers up again.... without knowing it's Dempster. Now, I'd be a fool not to notice a guy who had that many big league seasons behind him and still figuring him as a young pup - thereby decreasing his likelihood of putting up similar numbers once again ... but I think you get the point.

Do we have to know who it is before we're willing to make a call as to career year or rebound candidate? And if we do, how does that knowledge affect our decision?

[viper]

People too often make selections based on name value. I attended Shandler's Arizona First Pitch forum. Although I didn't participate, I attended his annual 3-man draft contest breakout session. You have $50 in an auction league to buy three players. After each team has 3 players, Ron will run a standings - available several days later. About 30 participated. The standard mix is two hitters and one pitcher. The stats used are the just completed year's numbers. In other words, you know exactly what you are drafting. What I observed is that players with name appeal but who actually provided less bang for the buck were still taken and bid up while lesser known productive players went for significantly less, if at all. Even in a perfect projection environment, people draft names.

[Todd Zola]

Do we have to know who it is before we're willing to make a call as to career year or rebound candidate? And if we do, how does that knowledge affect our decision?

Here's my take on this.

Gary and I have developed a projections system that is similar to others in that it uses a 3-year weighted average as a foundation, incorporating our own regression, growth, aging and MLE translations. It has been well established that this process is a viable means to predict future performance. You can apply this process to skills and convert skills to numbers as well.

Obviously, it is not 100% perfect, or else everyone would know the magic formula. Part of the issue is projecting playing time but not all. After objectively computing a player's 3-year base, you have to ask yourself "is there some luck involved that has not been accounted for in the regression" and "are the skills accurately depicted or have they grown or declined more than should normally be expected?"

This is where you need to know MORE ABOUT THE PLAYER.

Was there an injury in the past 3 seasons? A position change? A change in the batting order? A coaching change leading to a change in approach?

The 3-year base is objective. These changes are subjective. Anytime there is subjectivity there will be bias. Keeping this subjectivity in check and having a level of consistency within this subjectivity is important.

I am going to post this in the 'Josh Hamilton' thread, but this is a PERFECT example of how subjectivity can be even "more biased".

Because of the nature of the story that is Josh Hamilton, most of us WANT him to succeed, it is human nature. So even though for the past 2 seasons, his skill set has been virtually identical, many are suggesting he will improve upon that skill set more than the normal 28 going on 29 year old player would, because he had been out of the game and really has a HIGHER skill set that will finally manifest this season. I am personally of the mind we will maintain this present skill set as opposed to improving any more.

The second part of the bias is with playing time. Last year was the first one EVER that he played basically a full season. If this were Geoff Jenkins back in the day, people would be saying "no way he makes it through another year playing CF every day in that Texas heat" and would discount his playing time. We have Hamilton for 552 AB, as compared to 625 he got last season. This is going to drop him to a very controversial ranking amongst outfielders, but unless I am convinced otherwise, I feel some level of injury risk needs to be considered.

[da_big_kid_94]

And if anyone hasn't realized it already, by reading these posts .... THIS is what I missed about last year!

[Mojo Jojo]

corey hart

i honestly did not know that was who we were talking about when i gave my response. would i have answered differently had i known who i was talking about?....to quote mrs palin "you betcha!"

so...the question is...is our own personal bias/subjectivity of a player a good or bad thing?

personally, i think it is a good thing. we are given (or come up with our own) projections we use to determine value/tiers/etc for draft purposes. for this example, let's say josh hamilton, vlad guerrero, ryan braun and corey hart all have identical numbers and identical values.
i am given the following projection for all 4 players: 32 hrs, 110 rbis, 18 steals

we look at those numbers, but we need to have a "feel" for what they really mean. are we talking about the numbers for josh hamilton who as todd points out may not survive the texas summer heat and get the quality/quanity of at bats he likely needs to get to his projected hr/rbi totals? or are talking about vlad guerrero who we should probably classify as a declining hitter in a productive lineup who because of skill decline may not meet our projections? or are we talking about ryan braun a young improving slugger who with a few more fly balls and less ground balls could hit 40+ hrs? or corey hart, a young speedy/slugger who because of (pick your reason) declined down the stretch?

our own way of looking at the numbers (or what numbers we choose to look at) and our known 'history' of a player (is he declining, returning from injury, improving, what team he plays for (sorry aaron rowand), did he display certain skills at the minor league level but those skills haven't yet transitioned to the majors) all provide data points for our decison/subjectivity and all come into play in our decision process of who we draft given the above choice. because i think we consciously or subconsciously look for "better than the projection" or upside and given the choice above i would rather have ryan braun than vlad because he will more likely beat his projection than vlad will. so i guess this might be the subjectivity todd touched on in his post...i bump braun up some hrs and rbis and i drop vlad down some hrs and maybe steals thus changing values. whether i do it in my projections or whether i do it in my head at the draft table using someone elses projections i think "how WE feel" about a player plays a role in our decision process

i guess my point is that despite all the given numbers there is some other process we use and ultimate refinement of that process is key to our success in this endeavor. not sure anyone will ever come up with a fool proof formula for projection (personally pretty sure they don't) but until then i think our intuition based on the numbers we use goes a long way in improving our odds of doing well.

for me, looking at ryan braun and saying...hmmm one of the better power options in the outfield with a slight uptick in flyballs will net me maybe the best power option in the outfield is so much more important than whether he is projected to hit 35 hrs to soriano's 32.

and oh by the way....yes i think corey hart rebounds. hart is still young and giving him the benefit of the doubt because i value the skills he displayed in the minors and believe those will have more of an affect on his numbers this year, than his decline in the 2nd half of last year.

[Bodhizefa]

big kid, this was an excellent read. I know that I personally hold bias on many players and allow it to aid me into bad decisions more often than I should. Perhaps the greatest skill of an owner is finding that perfect mix of objectivity along with the sprinkling of subjectivity to really take the final step towards awesomeness. Or maybe it's just drafting Corey Hart in his career year.

[da_big_kid_94]

I agree with you, mojo ...it is our own personal peccadillos and values that we place on certain skill sets that make our game. But has the production of a given player become secondary to who he is and what we think about him? I think Jason was the one who tried to show everyone they should get off the Esteban Loaiza bandwagon several years ago ... and to me, that was a case of people looking at the stats rather than who put those stats up! In other words, the converse of the discussion here! But, IMHO, it was because many people thought he had finally found "it" ...whatever it was ...those guys are like rookie closers ...few and far between.

Victor Martinez's season was most probably not his fault last year - and most people will likely have him penciled in as a rebound candidate - can't see much fault in that. But when he had hit zero homers in April, hardly anybody blinked. It was his leg, supposedly. Now of course, it turned out it wasn't his leg .... but Wedge knew about the elbow since spring training ... but they kept throwing him out there, bad wing and all. No DL .. try to play through it , Vic. And the same guys who made that decision are still in charge - and Vic isn't getting any younger. To me, Vic's a big risk ...but because he is VMart, will people ignore this and value him too highly based on past accomplishments that may not be a valid picture anymore? That is a question for each of us to answer to his or her own satisfaction - and there are many others.

This discussion has been an enlightening one - so much so that I may have formulated the question I was trying to ask and couldn't word;"At what point in your analysis of a player's projections does the identity of the player weigh equally to past accomplishments or tilt the scale towars future rather than past?"

Many folks have spoke of the three year weighted average - which is one of the more intelligent and well known methodologies being used today. Yet, in many cases, that methodology, if one chooses to look at it in a simplistic matter, is nothing more than a 2/3rds majority. This is not an attempt to put the methodology down - nor is it an attempt to show it's flaws - hell, all of our methodologies have flaws. I'm just trying to frame it like this - if 2 out of the three years in the weighted average were good - are we more likely to throw out the bad year depending on who is involved? Conversely, if 2 out of those three years were bad - are we more likely to brand the 2 bad ones the norm based on the player's identity.

Which leads me to one final thought and food for discussion - a guy who had three consecutive successful seasons automatically goes into our "no brainer" file unless there are extreme mitigating circumstances - not much the three year weighted average is going to tell us here we didn't already know. It is how we judge the "2/3rds" player that will most likely be the largest contributing factor to our success.

[Guest]

Many folks have spoke of the three year weighted average - which is one of the more intelligent and well known methodologies being used today. Yet, in many cases, that methodology, if one chooses to look at it in a simplistic matter, is nothing more than a 2/3rds majority. This is not an attempt to put the methodology down - nor is it an attempt to show it's flaws - hell, all of our methodologies have flaws. I'm just trying to frame it like this - if 2 out of the three years in the weighted average were good - are we more likely to throw out the bad year depending on who is involved? Conversely, if 2 out of those three years were bad - are we more likely to brand the 2 bad ones the norm based on the player's identity.

Which leads me to one final thought and food for discussion - a guy who had three consecutive successful seasons automatically goes into our "no brainer" file unless there are extreme mitigating circumstances - not much the three year weighted average is going to tell us here we didn't already know. It is how we judge the "2/3rds" player that will most likely be the largest contributing factor to our success.

You can attribute this specific issue to much of why Todd and I have staggering cell phone bills.

And once everything rolls out, you'd better believe we intend to point precisely this out as one of the flaws of our own methodology (yes, you read right, we intend to spend as much time questioning ourselves as anything else - probably the biggest reason behind this whole endeavor).

I owe a bit bigger post to this issue later today but we got some snow and well, snowblowing is about to begin.

[cwk1963]

This is exactly what I missed on the 'other' boards and why my money will be here. Thanks, kid, for an excellent read. And thanks to everyone else - Todd, JP, mojoe, viper, Gary, etc etc etc who's participated and are making this the best forum on the web.

[JP Kastner]

There are two parts to projection, the Machine Element and the Human Element. The art is putting the two together. The Machine Element will get you 30% of way there. The Human Element will get you 30%. That leaves 40% that cannot be predicted. At least one player picked in the first round of forum's mock draft will get hurt and miss significant time. That means that some other player will get an unexpected significant increase in playing time. Someone will figure things out. Someone will collapse under pressure.

If you focus too much on the Machine, you will miss honest to goodness breakouts. Focus too much on the Human Element and you corrupt the projections with false expectations. I'm curious to see what Todd and Gary's projections will look like.


JP

[Todd Zola]

The more rigorous the objective machine part, the less subjective one needs to be. For instance, incorporating aging into the machine reduces the amount of subjectivity when doing the human part.

But what about Justin Upton, who started in the bigs at an early age? Is he subject to the same aging trends as players similar in years? How about Kevin Youkilis and his late start? Not to mention Josh Hamilton who lost several years dealing with his personal demons.

For a little more insight into the human/subjective side of projections, check out the Josh Hamilton thread as a couple of different ways to subjectively analyze the same data is there.

Perhaps Gary may have a different opinion, but in general, if you are expecting to look at our projections alone and notate several "breakout" or "bust" candidates, you will likely be disappointed. However, if you read our profiles in association with the projections, we will identify candidates most likely to improve or decline and we will explain the reason, so you can then make your own decision with respect to how much you want to pay or where to draft a player.

One more thing, playing time will be incredibly important. If you feel a player will get more at bats than we project, his counting stats will obviously go up. As opposed to other sites, you will be able to manually make that adjustment in our automated value calculator and compute the player's value using your playing time estimates.

[Guest]

To be honest I think this focus on "The Machine" is somewhat overstated, and without criticizing those who are trying to make a good point that human intervention simply has to be part of the decision consideration, lets not lose sight of that the machine is doing:

We get fed a lot of information on a player over their career or three year path or whatever in-between. They play in different parks, leagues, levels, they're different ages, etc. We have to do something with all that information to try and normalize it so that we handle the statistical output of all players consistently.

Again, all we're trying to do is get the process started with a consistent methodology that is as reasonable and robust as possible (by robust, meaning factor in as many things as possible). I don't see where anyone can question this as an approach. (Note, I dont think anyone is).

This thread has focused on two points that need addressing while a third has been touched on and I want to hit them now...

1) If you want to look at the 2 out of 3 issue, and I agree its a substantial issue, by nature you have to also accept that a player's stat line is not really 3 seasons, but the 1000-2000 PA he likely has had over those three years, with accompanying results in all of those PA. Obviously we weight the ones most recent more because the most recent data is in our opinion most likely to reflect current status but we're otherwise not using the season standard for anything more than trying to determine playing time and for weighting the information. Everything we do is evaluated on a per PA basis.

2) The idea of rebound is a great discussion point but ultimately the forecaster has to ignore it as much as possible. I'll go to the coin toss example - if 2 out of 3 were heads, we know enough about the coin to know that the fourth throw is still 50/50. But we don't know enough about many of these players! Hell, I'd argue depending on weather, home vs road, day vs night, whether the guy got laid the night before, etc, he's probably a different skilled player every day of the season. Again, the forecaster doesn't know it's a 50/50 coin, they know the information in front of them, therefore they take all of it and try and get to a best approximation of what the "coin toss" percentages are. And that for the most part reflects the projection.

3) This is the piece alluded to but not directly said (I think, its been a long thread), which is who the player's we're really talking about are. And really it's three types - the minor leaguers, the guys with a real uptick, the guys with a real downturn. The minor leaguers you probably will find us more conservative on than many others - but guess what, a lot of these minor leaguers fail, especially in their first run in the majors. If anything we're keeping you out of trouble devoting resources at a bunch of minor leaguers who may or may not handle the transition well. Yes, you might miss on Ryan Braun, but in a single-season league, how much trouble do you get in going for a more consistent guy who's 25-100 and missing out on Braun's 35-110, but you've saved all the risk inherent in Braun (as a rookie, not talking today). For every Longoria, there's a Gordon (and I'd argue Gordon has been a decent player, just not the stud his AA numbers in 2006 might have influenced some he'd be right away). Clay Buchholz didnt set the world on fire last year. Felix Hernandez has needed a few years to take the next step. We remember the misses much more than we remember where conservatism would have been the best approach.

And then there's the uptrenders and downtrenders. Todd has done this piece before - guys who go up come down, guys who go down come back up, its basic regression to the mean at its finest. So the real question you're asking is where a player's skill has actually improved or degraded to the point that their earlier output is no longer relevant to the conversation. And I'll concede that in some cases this simply has to be addressed. Dustin Pedroia is one - he had almost no power in 2006 in AAA and then somehow in the majors, it showed up. You may have to ignore 2006 for the most part if you think the 2008 power is real, for example. And a guy like Andruw Jones who didn't deserve any hedging on his 2007 downturn. I'd argue, however, that the LA Dodgers who have a lot more resources to evaluate talent, plus scouts, gave him $36 million assuming, I'd imagine, he'd tick back up. It's hard for me to to think a fantasy evaluation system should be held to a higher standard than the GM of the LA Dodgers insofar as addressing the likelihood he'd move towards career norms at the age of 30, but to each their own. Yeah, averaging might miss the guys surging or fading to some degree, but my real question is whether you'd prefer a fairly well tested system that's being constantly updated for new information as it's available to consider the information or a gut hunch. In many cases we'd just as soon keep as much bias out of it as we can. Not all, however.

I'll conclude with a couple more points:

1) When we go out on a limb with a minor leaguer, its a sign their performance really backs it up. Matt Weiters will be an example of this. Clear evidence he's worth the risk.

2) The standard we'll set for ourselves is reasonability more so than pure accuracy. Does the projection make sense? Is it reasonable? We use profiles as well as other written information to reflect more qualitative concerns. I'm not sure how responsible it is to put a projection out reflecting a guy as a $22 player who's stat line reflects he's never put up more than $14. Yes, you miss some, but that's going to happen anyways. No projection process whether it's picking numbers out of a hat all the way up to pure computer or somewhere in between was projecting that for Josh Hamilton.

3) And don't forget, these are single season values. Dynasty/keeper values don't work exactly the same way (different, and likely important other thread). In football this seems a concept more easily considered, but we produce single-season values which have to take the risk a guy flames out into account for the purpose of allocating resources. Different animal if you're going after a guy to keep him for a few years.

Hope some of this is readable. My fingers hurt. "And now your back's gonna hurt, because you just pulled landscaping duty". (Some of you may get the movie reference).

[JP Kastner]

I'd like to continue this thread by looking at a different player, also an outfielder. This player remained on the same team and in the same park.

Code: Select all

Age AB   R 2B 3B HR RBI SB BB  AVE
23 557  63 12  5  0  31 14 16 .296
24 691  80 29 13  4  74 21 41 .288
25 680 119 37  6 31  96 20 34 .328

I hope I have the ages right. How would (as I call it) the Machine predict age 25? Hitting .290 seems a safe bet. There is a noticeable increase in power. One could assume that some of those doubles would turn into HR as he continues to grow. Is 10 HR a good guess? How about SB. Clearly he has speed. We all know that a SB is more than purely speed, it is also knowing when to steal, and that requires good coaching and experience. Would one say with confidence that 25 is a safe bet?

If the Machine predicts .290-10-25 and he hits .328-31-20, then the Machine is wrong in this case. Can the Machine be assembled in some way to predict Age-25? If not, how can the Human element predict Age-25?

I would argue that the Machine can be assembled to predict Age-25, but I'll save that until I see how the minds tackle this problem. By the way, this is a real player.

[da_big_kid_94]

J.P. - a question - the age 25 year you have here is a projection for what he will do in the 2009 season or has this year already taken place and these are the numbers this player put up in 2008? I ask because the HR number looks very much out of proportion.

[msugray]

J.P. - a question - the age 25 year you have here is a projection for what he will do in the 2009 season or has this year already taken place and these are the numbers this player put up in 2008?

Big_Kid,

These are actual statistics of a player that is no longer active.

Edit: Yes, Kid, the player did have an interesting career arc with respect to HR. A leap from 4 HR to 31 HR at age 26, a fall from 24 HR to 9 HR at age 29, followed by an increase in HR rate each of his next 5 seasons. The question is.........can this be foreseen?

[Todd Zola]

The era this player played in is important to know what to load into the machine.

That said, I have trouble believing a machine that did not have a flux capacitor in it would be able to predict that power surge and a human that wasn't in possession of such a machine with knowledge of an impending lightning strike could not either.

Color me curious.

[JP Kastner]

The player is Kirby Puckett. He fits a type of player I call the "Hidden Power" player. Other hitters like this would be Jim Edmonds and Alex Rios. No minor league indication from a statistical perspective that he has power. From a scouting perspective, they will say that the power is developing.

[da_big_kid_94]

Ok ... first of all, a jump for 4 to 31 HR over a single season is Brady Anderson-esque. I don't think any methodology could come close to predicting such a leap. Having said that, we take a closer look at the numbers.

A +27 HR differential that only translated into a +12 run differential. In addition, it also produced, in essence, a -5 differential in RBI. Speed is obvious - one does not hit 13 triples in a season without some pretty good wheels. The ABs indicate a guy who is usually quite healthy and hits near the top of the lineup. The fact that the number of hits was omitted is probably significant. Without seeing his K total, can't tell if his pitch selectivity really improved. I would suspect that his team had a down year in this player's 25th year with the decreased RBI totals (drove in more of him than he did his teammates over the previous season - they probably weren't getting on base as much - could have been a lot of solo shots) and the slight increase in run totals despite a 40 point BA increase.

The three year weighted average here (or the Machine) would be an interesting decision on this guy. With what I see here, I would guess career year and his future career regressed.

As for the machine - as I said before, who's got a methodology that would allow for a +27 single season HR differential? ESPECIALLY when the prior year he didn't crack double digits in a similar number of at bats. None that I know of. And there is the difference between predicting it and it actually happening. Were I actually still buying into "lineup protection" theory, I would venture a guess that a new big bat found its way into the lineup just in time for his age 25 year. I would also assume that a huge non statistically quantifiable factor changed for him as well ...to the positive, of course.

Given the age 24 and 25 season stats, every methodology would be wrong - who would predict such a huge jump in power numbers, and yet not see a proportionate increase in runs scored or RBI? I think machine, human, Miss Cleo or ouija board all come up short on this one.

Edit - BTW J.P. - that was a nice piece of work you came up with here

[Todd Zola]

Interesting, especially since the 31 came in 1986, a year before 1987, otherwise known as the year that Wade Boggs hit 24 HR with the lively ball.

Could an increase in power and speed have been predicted? Sure, look at the triples as well for some speed potential.

But 31?

MY EDIT -- Kid beat me to it, funny we both thought of Boggs. I suspect JP is very familiar with this season from his Strat background

[JP Kastner]

I picked someone prior to 1995 to hopefully take 'roids out of the equation. Yes, I was playing Strat when I watched this happen. Puckett was a top pick because he was good defensively in a tough position. When he made that jump, he rocketed to a first round pick.

This has happened before and will happen again to players like this. If we can predict it, we have a gem.

JP

[Guest]

If we can predict it

JP (and others) - Ive made the argument to Todd most notably, though Jason and Rob in the past, that reasonability is the most important thing that a projection set should provide. Once projections start to get unreasonable, in my opinion they become somewhat unwieldy (By the way, Im not suggesting Todd, Rob or Jason agreed in entirety).

Kirby Puckett's history in the minors, majors to that point said basically he didn't have a ton of power. I don't know how quantitatively, a fantasy site can recommend to its customers they pay for stats that the player has never demonstrated the capability to put up.

In fantasy football you always have a guy's college record, and while the translation isn't that smooth you see a RB ran for 1300 yards and 10 TD in college, so when they've sat on the bench for 2 years and get a chance in year 3, there's some sort of record of this ability. In baseball, you have the minors, and while it's not a perfect translation, it's the best you can do.

I'm the first to admit our model isn't going to suggest massive breakouts consistently. We'll refine it to do as best as it can. But the record of guys not breaking out has to be taken into account to. As I believe Ron Shandler mentioned that Peter Kreutzer had said (I may be off on this) - you can't let outliers shape you away from doing a great job with the majority of guys. (paraphrasing too much). Puckett's an outlier. They happen.

BTW - he's also an excellent example of the limitations we have in projecting players.

[Todd Zola]

Quick follow-up from the fingertips a co-proprietor of a fantasy baseball website that is attempting to carve out a niche in a horrible economy where disposal income is limited if not completely dried up---

In 1986, our "job" would not have been to predict Puckett to hit 31 HR. Our job would have been to alert you that he had the potential for a power increase. Our job would have been to make you want to pay $1 more than the next highest bid or draft him 1 round before everyone else.

[Guest]

The other co-proprietor would like to point out that it is indeed the understanding of the limitations involved in doing projections and in overall projection theory and the type of strategic things one can do to counteract these issues that makes this site one we believe is worth your time and money.

[msugray]

JP,

The suggestion of hidden power can be found in batted ball data (Ground Ball (GB), Line Drive (LD), Fly Ball (FB)) , but I haven't yet found a definitive projection formula.

I do know the likelihood of a player hitting a home run is improved if they hit fewer ground balls and hit more fly balls. That statement is obvious but I feel often overlooked. Ground balls have little to no opportunity to become a home run. Therefore, I would not be looking for a player to find his "hidden power" until I observed a change in their hit trajectory.

For instance, Casey Kotchman and Delmon Young have batted ball data more closely resembling Jacoby Ellsbury and Michael Bourn than power hitter profile.