The STALLVALUE SYSTEM 2004 is really three systems in one. It is first a
summation of a player's actual performances for the prior three years. Second,
it is a projection of his statistics for the upcoming year. Lastly, it is a
valuation of his statistics for fantasy league purposes.
For more information on the interrelation of these 3 "systems" please see the essay entitled Performance, Playing Time, Price .
This essay deals primarily with the aspect of prognostication which we call "performance". By this we mean the methods used to determine what a player’s propensity to perform is per unit of playing time (i.e. how many home runs per at-bat).
Prior performances
The first step in creating projected performances for each player is to create a data base on which to base these projections. One method would be to take his prior year's performance and project him to duplicate it. This might be acceptable if a player never got better (or worse), never was injured, never had good or bad years. For obvious reasons, we do not use this method.
Another method would be to take a player, look at his lifetime statistics, calculate how much you expect him to play and estimate his expected performance based on your beliefs about his current capabilities. Thus, if you read that Jason Tyner had just beaten Barry Bonds in an arm wrestling contest you might figure Tyner to hit 60 home runs and project him accordingly. While this method is better than the first it is founded too much on hunches (and therefore overvaluing favorite players) for our taste.
Having eliminated the two methods above, we should then consider those methods which apply prior performance to expected playing time to mathematically compute a player's performance. Once it is decided to use this type of method it is necessary to determine the relevant time frame to use as our yardstick, as well as the weight to be given to each year.
One such method would be to use a player's lifetime ratios and apply them to this year's expected playing time. This system does not work well with older players since how a player performs will change over time. For example, Jim Thome is a better and more powerful hitter today than when he broke in in 1991 and age and injuries dictate that it is highly unlikely that Ken Griffey Jr. will ever again reach the 56 homer and 20 steal plateau as he did in 1998. Thus there comes a point at which looking to the past starts to distort the picture of the statistics that a player will produce today.
Thus, if we reject a player's entire history we come to the point of looking at anywhere from 1-5 years. As stated above, a 1 year period creates too much of a possibility for distortion based upon injuries, career best or worst years, etc. Additionally, while a two year period would smooth out the volatility, it would still be giving too much weight to aberrational years. We also reject a 5 year period simply because it looks too far back into a player's career (i.e. if the character of a player's statistics have started to change, a 5 year period would be too slow in reacting to these changes).
Thus we need a time period which will mitigate the effect of aberrational years while being short enough to react to true changes in a player's performance to allow us to take advantage of this information in a yearly draft.
Using this reasoning we come to choosing between 3 and 4 year periods. We choose a 3 year period because it is sufficient to iron out glitches in performance and since it will react to true changes faster, it is the preferable time period.
Another issue that we had to resolve was whether we should weight our ratios to give greater weight to certain years or whether to weight all years in the base equally. There are 3 major problems with weighing current years more heavily.
The first major problem is that our statistics have shown us that weighing the immediately preceding year more heavily than the second preceding year is STATISTICALLY WRONG!!
To determine this we studied 3 year histories of a large group of players over multiple years. We used 2,118 hitters and 1,579 pitchers. The last year we added to this database was 2003. We wanted to determine whether year 3 was closer to year one or year two.
For our control groups we only use players who have played predominantly or entirely in the majors for the 3 preceding seasons with at least 100 ABs or 50 IP for each of those years, including the current season.
For 2003 there were only 236 major league hitters who met that criteria. Using only this group each year ensures we weed out those players with only a few ABs or IP who could throw off the results with one “fluke” hit, etc. It also ensures that we are not worrying about the accuracy of our minor league conversions (yet another essay topic covered in Major League Equivalents)
Following are the results broken down by category. We did not review wins or
saves since they are too dependant on factors beyond the player's control.
|
PRIOR YEAR COMPARISONS BY CATEGORY |
||
|
Category in Year
3 |
Closer to Year 1 |
Closer to Year 2 |
|
Batting Average |
49.10% |
50.90% |
|
Home Runs |
49.45% |
50.55% |
|
RBI |
48.59% |
51.41% |
|
Steals |
47.78% |
52.21% |
|
WHIP |
49.84% |
50.16% |
|
ERA |
49.21% |
50.79% |
As you can see, the study favors the more recent year but only ever so slightly. Certainly nothing to indicate that the most recent year should have a significantly higher value placed on it that the second year back. Also, when you factor in a player's age the differences virtually disappear. For example look at Craig Biggio heading into 2003. He not only had a career average of .288 but had hit below .288 2 of the last 3 seasons. In 2002 he hit .253 while in 2001 he hit .292. Thus, given his age you would expect that in 2003 he would be closer to his 2002 average but probably above it. Of course he dropped to .264 in 2003 making him closer to year 2 (2002) than year 1 (2001) in our study. In our study the group that stayed closest to the immediately preceding year was either the younger or older players. Exactly what you would expect. Thus, when you factor in ages there is simply no statistical basis for weighing the most recent year more heavily other than it sounds good to the uninformed.
This is why you will see that our projections are more accurate overall than other draft guides that are out there. Virtually every other guide gives a significantly greater weight to the immediately preceding year because intuitively this sounds good. However, in doing so their projections are actually farther off. This study may also help explain why free agents so often appear to be overpaid. The fact is that they will not repeat that great year they had just before they signed.
While were are on the subject of looking at prior year performances we would like to touch on the subject of constant increases or decreases in a player’s performance. While this is dealt with more fully in the essay on Peak Years , for the purposes of this essay, we want to look at the probability of a player continuing to show improvement or regression in the various categories. Thus, in studying the players we also tracked the instances when the player’s stats headed in the same direction for two consecutive years. i.e. if a player hit .250 in year 1 and .260 in year 2 what was the probability of hitting more than .260 in year 3. The results may amaze you.
We tracked the percentage of players who moved up or down for two
consecutive years versus those players who moved in one direction one year and
the opposite direction the next.
|
Statistical Movement |
||
|
Category |
2 Up or 2 Down |
1 Up 1 Down; 1
Down 1 Up |
|
Batting Average |
34.80% |
65.20% |
|
Home Runs |
37.44% |
62.56% |
|
RBI |
36.93% |
64.07% |
|
Steals |
40.18% |
59.82% |
|
WHIP |
36.48% |
63.52% |
|
ERA |
38.06% |
61.94% |
As you can see the odds are significantly against a player improving 2 consecutive years. Of course we also tracked the ages of the players in these groups and the ones improving 2 straight years were the youngest, those declining were the oldest and those flipping were in the middle. There was an important exception to this pattern but this is just a tease to make sure that you read the essay on Peak Years .
A second reason to use an equal weighting for the years is aberrational years occur more frequently than do true changes in a player’s performance. Giving greater weight to the immediately preceding year will tend to distort the projections.
While occasionally a player's performance will change permanently (as Luis Gonzalez’ did in 1998 when he went from averaging 12 HR/year the prior seven seasons to averaging 32/year over the next six), it is much more common for a player to have an aberrational year. Other than Gonzalez, how many other players have dramatically altered their performance and then proved that change not to be a fluke? Not coming up with too many names are you? Neither did we. However, we came up with numerous instances where players had one aberrational year and immediately returned to "pre-aberration" form. Some examples are Gary Gaetti’s home run totals from 1980 to 1988. (starting in 1980: 22, 32, 25, 21, 5, 20, 34, 31, 28) In that period, his lowest total was 5 in 1986 and in that year he had the second most at-bats (588) in the 9 year period. Some other examples are: Wade Boggs 24 HRs 1987 (made double digits for only the second time in his career in 1994); Cal Ripken 14 homers in 1992 (has never failed to hit 20 in any other season except strike shortened years until 1997 and 1998); Kevin Mitchell 9 home runs 1992 and 47 home runs 1989 (he has not come within 10 of either extreme in any year with at least 350 at-bats). Jeromy Burnitz hit 27 or more homers per year 1997-2001, hit 19 in 2002 and then 31 in 2003. Barry Bonds hit 73 HR in 2001 but has never hit 50+ in any other season.
From recent years we see two more glaring examples. Two
players who hit the home run leader boards for the first time were Richard
Hidalgo in 2000 and Rich Aurelia in 2001. Prior to 2000
We could go on with this list for some time, the point being that every player will (by definition) have a career best and a career worst year. By weighing the immediately preceding year too heavily we would be stating that we believe that a player's performance in that year is a greater indication of that player's capabilities than any other year and as noted above we cannot find any statistical data to support that contention.
The third major problem with weighing the most recent year more heavily is the "sheep dilemma". Now it has probably occurred to you that even if you weigh the years differently, at the end of the "base period" each year will have received equal weight. thus, using a 3 year period if the weight was 50%, 30% and 20% as opposed to 33.3%, 33.3% and 33.3%, then, at the end of three years any year analyzed would have contributed the same overall weight to the statistics (i.e. 100% over 3 years). However, the sheep dilemma dictates against overweighting the most current year. The sheep dilemma is created by the fact that many fantasy league players are sheep. They look at what a player did last year and run with the crowd thereby overvaluing or undervaluing a player based on last year which was most likely an aberrational year.
Seven years ago in this space we warned you about Brady Anderson and how he was highly unlikely to come anywhere near his 1996 50 homer total. Five years ago we warned you about a few players who saw their power zoom up in 1997 and who were unlikely to repeat that in 1998. They were Jose Cruz Jr. (26 HR '97 to 11 HR '98), Tino Martinez (44 to 28), David Segui (21 to 19) and Jeff Kent (29 to 31). OK, 3 out of 4 went down, a pretty good percentage. The one player we predicted as likely to rebound into the mid 40 homer range was Albert Belle (30 to 49). The bottom line is that is doesn't get much better than that. OK, OK it does get better. In 1999 we listed 6 players likely to go down in homers with about the same ABs as the prior year. Two were Moises Alou and Andres Galarraga who had significant injuries that year so we won't count them. Of the other 4 all went down including Albert Belle's plummet from 49 to 37. We picked 6 hitters to increase their production in homers with about the same number of ABs. All 6 increased including Butch Huskey (13 to 22), Gary Sheffield (22 to 34) and Larry Walker (23 to 37)
For the 2000 list we warned about drops for Dante Bichette, Chipper Jones, Jay Bell, Steve Finley, Rafael Palmeiro, Fred McGriff, Shawn Green and Ivan Rodriguez. That group went from 307 homers in 1999 to 229 in 2000. Only Steve Finley went up (34 to 35) We also projected increased production from Ray Lankford, Frank Thomas, Mo Vaughn, Jim Thome, Raul Mondesi and Adrian Beltre. That group went from 144 homers to 186 with only Raul Mondesi showing a decrease.
Ready for the 2004 list? Here are players we think will have a
significant decrease from their 2003 numbers: Javy
Lopez, Bal, Jim Edmonds, StL,
David Ortiz, Bos, Jason Varitek,
Bos, Craig Monroe, Det, Dmitri Young, Dcet, Shea Hillenbrand, Ari, Hank
Blalock
Here are the players we think will have significant power increases over 2003: Lance Berkman, Hou, Roberto Alomar, ChW, David Bell, Phi, Danny Bautista, Ari, Marlon Byrd, Phi, Hideki Matsui, NYY, Eric Hinske Tor and Shawn Green, LA.
Also, as noted in the instructions you must also pay attention to "positional statistics" which may alter a player's value based on the position he will play. For example, if a player who was historically a closer in the minors and will be moved by his team to a set-up role, you should expect that his saves will decrease over his historical ratios. Again, this is explained more fully in the instructions.
We also factor in a player's age, so younger players project to greater than the average of the prior 3 years, while older players will project to less.
The years that we have players peaking and starting to decline vary depending on the category as you will see when you read the essay on Peak Years . However, when we factor this in, our projections are even better since (as noted above) we know that younger players generally improve in the categories. Thus, there will be some younger players whom we do project to show a second or a third straight year of improvement.
In conclusion, based on the combination of factors and the detail we pay to research you will see that our projections are unmatched by any other system on the market. That should go a long way towards giving you the edge you need to succeed.
One final note here is that you need to be mindful of the impact of expansion if it happens again. There are more hitters who are nearly ready for the major leagues than there are pitchers. Thus, when you have expansion you will generally see an increase in offensive output. Why? Because the expansion hitters are essentially major league hitters while the expansion pitchers are essentially AAA.
For example, look at the 1993 rosters of the Marlins and the
When we looked at a sample of 1992 to 1994 for pitchers we found (not surprisingly) that the percentages rose dramatically for pitchers being closer to the immediately preceding year (1993). For WHIP it was 54.29% and for ERA it was 55.35%. Thus we saw swings of 6 to 8% in those two categories. Would it surprise you to know that the pool of pitchers expanded by 7.7%? The same held true in 1998. In ERA pitchers were closer to the immediately preceeding year to by 51.48%.
In other words, pitchers in 1994 we closer to 1993, but only because of expansion. The same holds for 1998 but to a lesser degree because the dilution was spread between two leagues. As time passes, those percentages will return closer to historical norms.
The logical follow-up to this is "Shouldn’t we have also seen the same type of 7-8% improvement in hitters stats?" The answer here is no because so much of the "explosion" of offensive stats were concentrated in the hands of a few. While 1994 saw a dramatic increase in the number of home runs hit for each team, the increase could be almost entirely attributed to 1 or 2 players. For example, in 1993 the White Sox hit 121 home runs not including Frank Thomas’ 41. In 1994 the White Sox were on pace to hit 119 home runs (excluding Thomas). However, Thomas was on pace to hit 54. Thus the pro-rated team total for 1994 was 173 versus 162 in 1993, a 6.8% increase. However, the increase was due almost entirely to Thomas. The same holds true in 1998.
An example in the NL is the Giants. They hit 168 home runs in 1993 and 130 not counting Matt Williams. With essentially the same team in 1994 (Will Clark had 14 in 1993 and Todd Benzinger was on a 13 homer pace in 1994.) they were only on pace to hit 113 home runs (not counting Williams). Williams was on pace to hit 61 bringing the projected team total to 174. Again the entire team increase can be traced to one player.
Therefore, the conclusion that is reached is that the decrease in the
quality of pitching did not have an impact on all hitters, but rather had a
dramatic impact on the very best hitters. We saw this again in 1998. In 1997
the
Return to Strategic Essay List
Return to PPI Home Page
Email questions to PPI or email
support@stallvalue.com
Copyright (c) 1990-2004, Player Projections, Inc. All rights reserved.