An Overview of the Last 25 Years of Fantasy BaseballPosted: March 1, 2013
It’s fantasy baseball draft season again and here I sit…again…trying to analyze the heck out of everything. The initial trigger for what is to follow was a prospect draft and I’ll get to that at a later time. First, I want to lay the groundwork for my ranking system. I decided it would be worth my time and effort to create a post-hoc ranking system for past fantasy seasons. This is not common as fantasy baseball deals strictly with “what have you done for me lately,” not “what did you do 25 years ago.” But here we are anyway, talking about Julio Franco’s 13 home runs as a second baseman in 1989.
First, the ground rules. Players have different values based on different league rules. What statistics are counted? How many teams are there? What are the starting positions? Instead of adding adjustments for different league settings, I decided to use the Yahoo standard settings. This means 12 teams; offensive statistics of HR, R, RBI, SB and AVG; starting positions of 1 Catcher, 1 First Baseman, 1 Second Basemen, 1 Shortstop, 1 Third Baseman, 3 Outfielders, 1 Utility Player; and positional eligibility of 10 games played or 5 games started.
Now I could put a value to every single player season since 1989. I chose this year because Baseball America started ranking prospects in 1990 and I wanted to include all of their data in my database. I could go back farther, but 24 seasons is plenty good. I already had the backbone of a 5×5 fantasy system from some projection work I did a few years ago, I just needed to apply it to multiple seasons. The actual ranking system relies on z-scores to give a point total to each player’s contribution in each of the five offensive categories. Instead of using league averages and standard deviations, I took those numbers only from players with at least 320 plate appearances in a season (for the strike-shortened 1994 and 1995 seasons I used PA minimums of 225 and 284, respectively). This is roughly the top 230 to 250 players in playing time, which is a good selection of the readily available fantasy talent. The numbers may change a bit depending on this assumption, but not too much to matter. For instance in 1989, the average HR total for the top 250 fantasy players was 10.9 with a standard deviation of 8.6. This means Julio Franco gets (13-10.9)/8.6 = 0.24 HR points in 1989.
I also had to create a replacement value for every position in every season, but before I could do this, I had to find for what position each player was eligible. I used the 5 games started/10 games played requirement provided by Yahoo and matched up defensive stats for each player and gave them eligibility at the appropriate positions. Players maintained eligibility the following year for each position. Then I set the replacement value of each position as the non-adjusted fantasy value (NAFV) of the 12th best player at that position for that year (36th for outfielders). For first basemen and outfielders, I filtered the data to only show 1B, 1B/OF or OF, 1B/OF players since these positions tend to have the highest replacement value. If a player was eligible for multiple positions, his positional value was set as the most valuable of these positions for that year. This is admittedly the weakest part of the analysis, but I believe the numbers are still solid and provide an accurate representation of value. The replacement value for designated hitters was set as the least valuable of the other positions.
In summary, player value is calculated as HR points + R points + RBI points + SB points + AVG points – Position Replacement Value.
For example, we’ll look at 2012 Mike Trout. The average stat line for a fantasy-relevant player was 11 HRs, 59 Rs, 56 RBI, 11 SB and a 0.264 AVG. Trout contribituted 30 HRs, 129 Rs, 83 RBI, 49 SB and a 0.326 AVG. This gave him 1.49 HR points, 3.15 R points, 0.93 RBI points, 3.67 SB points and 2.26 AVG points for a raw value of 11.50. However, outfield was the second strongest position last year, with Ichiro Suzuki (9/77/55/29/0.283*) representing replacement level. Trout lost 2.23 points for his position and had a final PAFV of 9.27, which ranked 27th out of all fantasy offensive player seasons since 1989.
*I will be using a quadruple slash line (there are four slashes and five stats) to represent player seasons in the interest of time and virtual space. These slash lines will always be formatted as HR/R/RBI/SB/AVG.
What do these points mean? Well obviously, higher is better; however, what is great, what is good and what is horrible? Fantasy point total is not normally distributed, so I can’t (or don’t know how to) create a system similar to ERA+ where better or worse values lie above or below an average of 100. Instead, I’ll make this chart:
I included all players with at least 130 at-bats in this chart because that was the cut-off for rookie eligibility. I figured it was a good enough place to stop, but again it doesn’t change much. The top 120 players are at least average and the bottom 250 are bad or horrible. About eight players have great fantasy seasons every year.
Along with the typical baseball acronyms (HR, RBI, etc.) I will be using two brand-spanking new acronyms, PAFV and NAFV. PAFV stands for position-adjusted fantasy value. It is the amount a player is worth based on his numbers compared to the fantasy average, with a position adjustment. This is his overall value. NAFV is non-adjusted fantasy value. This is how much a player is worth, regardless of position.
Now that we have an understanding of the data, what can we learn from it? For now, I will just be looking at offensive players. I have not included pitching value in the data yet, but will eventually. However, I do have hitting data for pitchers and here’s how it looks: ugly. The best hitting fantasy season for a pitcher was Mike Hampton in 2001, but he still provided -5.58 points of value. Carlos Zambrano’s 2008 season comes in second with -6.09 points.
Of course, the ultimate question: what was the best fantasy season of all-time (post 1989)? That distinction belongs to 1997 Larry Walker who put up a ridiculous 49/143/130/33/0.366 line for a PAFV of 13.26. The second best season was an 11.18 for Matt Kemp’s 2011 with a 39/115/126/40/0.324 line. The biggest distinction between these two seasons is the positional replacement value. Here is a chart that shows the yearly replacement values by position:
(Click to enlarge)
1997 was the worst year for outfielders for whatever reason. It was very top-heavy, with Larry Walker (13.26), Ken Griffey Jr. (9.83), Barry Bonds (7.72) and Tony Gwynn (6.79) all having great seasons. After that, the quality level drops off a cliff and Rickey Henderson becomes the 36th best outfielder. Henderson was one of the greats in his prime, but by 1997 he was 38 years old and his fantasy quadruple slash was 8/84/34/45/0.248. Lots of runs and stolen bases, but not much else. That year, he was worth about the same as Al Martin and Troy O’Leary, two guys I’ve never heard of until just now.
Here is the same chart, but with 5-year moving averages instead of yearly data.
(Click to enlarge)
A few trends pop out here that were not so visible in the last graph. First, 1B and OF strength and 2B and SS strength run parallel to each other. I believe this happens because the same players are eligible at both positions. Also, OF and 2B/SS/3B strength mirror each other. When OF is strong, the non-1B infield positions are weak and vice versa. I’m not certain why this is. Maybe randomness? Maybe teams are deciding to forego offense at one position and focus on it at the other? Overall, infield offense increased through the 2000s up to a peak around 2008. It has been decreasing since then. 1B offense has been decreasing from its high in the late 90s (Mark McGwire, Andres Galarraga, Jeff Bagwell, Mo Vaughn, Frank Thomas, Larry Walker, etc…).
Who were the stars in the late 2000s when 3B became the most offensive position? Alex Rodriguez, David Wright, Adrian Beltre, Miguel Cabrera, Melvin Mora and…Chone Figgins. Yep, Figgins was worth 5.50 points in 2005 while being eligible at 2B, SS, 3B and OF, thanks largely in part to a ridiculous (for 2005) 62 stolen bases.
But Alex Rodriguez? Not many can compare to his fantasy prime. He has two of the top five fantasy seasons (2007 and 1998) and six of the top 25 (2001, 1996, 2002, 2005). Barry Bonds is the lone player to outmatch him in total points since 1989. Bonds’s career stretches outside of that range, though so the comparison is not exact. A-Rod’s worst season was his rookie year in 1995 when he struggled to hit 5/15/19/4/0.232 in 149 PAs for a total point value of -4.85. He was 19 at that time though, so give the kid a break. Since 2011, he has been worth about -1.5 points per year, due to injury and age. Bonds doesn’t have quite the peak of A-Rod because of his position. His 1993 season was similar to (and slightly better) A-Rod’s 2001, but A-Rod did it at SS while Bonds was in the OF. Bonds’s famous (infamous?) 2001 season where he hit 73 HRs was a great season (9.63 PAFV), but he gets docked on positional value again. In 2001, the 36th best OF was Jeromy Burnitz with a 34/104/100/0/0.251 line.
We are currently seeing the best offensive output for catchers in the past two decades. This is due to the likes of Joe Mauer, Buster Posey, Pablo Sandoval, Victor Martinez and Mike Napoli, all of whom have put up a season of five or more points while eligible at catcher in the last five years.Follow @stealofhome
Here is the list of the top 25 fantasy seasons since 1989
|1997||Ken Griffey Jr.||Mariners||OF||9.8|
|2011||Jacoby Ellsbury||Red Sox||OF||9.8|
And the list of the top 25 career fantasy values since 1989
|Player||PAFV after 1989|
Again, one partial season can really wreak havoc on a player’s career value. In addition to the aforemention 1995 A-Rod season, Mike Piazza, Roberto Alomar and Ivan Rodriguez all have terrible seasons that drag their overall ranking down quite a bit.
And because we’ve gone this far, I’m just going to keep posting graphs. This one compares a player’s NAFV to his wOBA that year. This chart and the next include only players with at least 130 ABs in that season. I used NAFV here because wOBA is not adjusted for position either. NAFV and wOBA have a higher correlation than PAFV and wOBA. I also tried graphing NAFV and wRC+, thinking this may be a better comparison, but the correlation is only 0.46, which is lower than the comparison with wOBA.
The outlier along the x-axis is Mark McGwire’s 2000 season. He had a wOBA of nearly 0.500, but only provided 0.26 points of fantasy value. This was largely a playing time issue. He hit 32/60/73/1/0.305 in 321 PAs, which is great for half a year, but fantasy baseball is all about accumulating counting stats, especially for hitters. The best fit line is a polynomial where lower wOBAs produce a smaller range of fantasy value. Value increases exponentially with increasing wOBA. the correlation coefficient is 0.56, so these two are linked very strongly (obviously).
If the reader is still around, they most certainly will be interested in the following chart, which compares PAFV to Fangraphs WAR. I chose to use PAFV for this comparison because WAR is adjusted for position as well.
WAR and PAFV are more closely linked than wOBA and NAFV. Also, the upper WAR values flatten out near 10 PAFV. These high WAR values may be maximizing on offensive output and have more to do with defensive WAR.
And because I’m in a giving mood, here is the trend for the yearly averages of each of the 5×5 fantasy stats since 1989.
As is fairly common knowledge, offense has been on the decline since the late 1990s. This effect is much more pronounced in runs and RBI than in home runs, though the values have trended up in 2011 and 2012. Stolen base totals are also up a bit in the past few years. Batting average has declined every year since 2006.
And so we have come to the end. The main thing I wanted to do with this post was to provide a place to familiarize everyone with the data and process I followed to create my ranking system. I will be applying this data in a following post and wanted to get all of this out of the way. It has been a brief look into the mindset I have when looking at a new data set.
What have we learned, a numbered list:
- Larry Walker’s 1997 was ridiculous.
- Mike Trout’s 2012 was ridiculous (did you know he was only 20 years old for most of that year? 20!).
- The definition of position-adjusted fantasy value (PAFV) and non-adjusted fantasy value (NAFV).
- The PAFV values for Good (6+)/Great (1.25 to 6)/Average (-1.25 to 1.25)/Bad (-1.25 to -3)/Terrible (-3 and below) fantasy seasons.
- Al Martin and Troy O’Leary are the names of two very real former Major League Baseball outfielders.
- 1B/OF and 2B/SS strength are linked to each other.
- When OF strength increases, non-1B strength decreases and vice-versa.
- Alex Rodriguez had the best fantasy prime ever.
- Barry Bonds was right there with him.
- We are in the best time for fantasy catchers in the last two decades.
- NAFV and wOBA are exponentially related and have a correlation coefficient of 0.56.
- PAFV and WAR are linearly related and have a correlation coefficient of 0.66.
- Offense decreased after the late 1990s but seems to be making a comeback.
That is all I have for now, but I will continue digging into this data much more and have some other results to share. Until next time!