Intro
A new draft season brings a couple new draft studies. I think these are much better than the last round more focused, more thorough, a better stat. Either way, heres the first one. In the prior iteration I used Win Shares because it was published during the initial research and one of the stated uses was scoring drafts. I think WS does a pretty reasonable job of scoring and ranking hitters. It does a good job of scoring and ranking pitchers. It does a poor job of mixing hitters and pitchers.
One of the things I wanted to do was to record the number of solid and better seasons that each draft class produced. The nice round number of 15 WS worked well enough for hitters to establish the bottom of the solid range for hitters, but it excluded far too many pitchers who seemed to have had equally good years.
This problem of comparing pitchers to hitters has also been noted by studes, the person who runs www.baseballgraphs.com. At least amongst people who comment publicly he is probably the leading expert working with WS. Over the winter he did a study establishing baseline WS levels for each position. His WSAA (win share above average) addressed the problem of setting replacement level for pitching too low and he even found a passage from James book that suggested the need for this type of evolution for WS. Thats all well and good, but it undercuts the easy use of WS for any kind of study that compares position players and pitchers.
If youre interested that work can be found here:
www.baseballgraphs.com/bl...39_0_1_0_C
Enter WARP3. Wins Above RePlacement is the stat developed by Clay Davenport and is the basis of Baseball Prospectus historical DT cards. Like WS its a single number that represents a players contribution in all facets of the game. In this case the 3 means its translated across historical eras. Thats also an advantage because it already translates the strike shortened 1994 and 1995 seasons into full season stats for easy comparisons across draft classes.
Aside from the integration of hitters and pitchers, the two systems are in good agreement. A rank order of just hitters or just pitchers would probably show very little difference. The big disagreements would all come from mixing the two types and at least to my eye the mixed rankings of WARP3 make more sense.
This study only covers 1987-1992. The draft rules were changed in 1987 to eliminate the January and June Phases. Since these are the rules that continue to govern the draft, that makes 1987 an obvious starting point for a draft study. If you assume a HS player in 1987 was drafted at 18 and a college player was drafted at 21, then those players were 34 and 37 in 2003. Their careers arent over so you cant put a final number on the career value, but theyre all well past their peaks.
I used 1992 as the end of the range to be studied because I wanted all the players in the study to have at least played through their twenties peaks. An 18 year old HS pick in 1992 would have played 2003 at 29. A 21 year old college pick would have played 2003 at 32.
So the college players in this study are age 32-37 and the HS players are age 29-34. Im going to save most of the college vs HS comparisons for a second post, but the age difference is something to keep in mind. All of the college players are well past the traditional peak and are pretty close to their final WARP career total. Many of the HS players are still just passing their peaks and have several productive years ahead of them. This issue is exacerbated by the fact that the 1987-1989 drafts were very college heavy and the 1990-1992 drafts leaned more towards the HS side. Just about all of the best C players in the study are at least 35 while just about all of the best HS players are 29-31.
Part 1. How many and what type of players come from draft classes
The left hand column is the WARP3 total. The number in that column, eg 70 refers to the entire 10 WARP group 70-79.9. In the next section Ill put some names to WARP3 numbers to give you a better idea of what type of player fits in those ranges. In general, Ive used 20 WARP as the cutoff for a useful career. A good career could maybe start at 35, but 40 keeps it simple and doesnt change things too much. To keep it as a pattern of 20 WARP segments a very good career could be considered anything over the 60 WARP level and a great career is anything over 80 WARP. The key demarcations are really 20 (useful), 40 (good) and 80 (great) WARP.
Lets start with the line Total thats about ¾ of the way down the table. I probably missed a few bad players here and there, but each of the drafts produced a lot of players who at least made the majors with a range of 140-169 players and an average of 155.
The line above Total "20 less" shows that the vast majority of those players do not have useful careers. For this reason studies that focus on the number of players who make the majors should be taken with a large grain of salt.
Of the players with useful career about half also exceeded the 40 WARP threshold and had good careers. A handful of those players also exceeded the 80 WARP threshold for a great career.
The last 4 lines of the table express the data as percentages. On average over 80% of the players who make the majors do not have useful careers. Less than 10% have good careers and less than 2% have great careers. Remember, those are percentages of players who made the majors, not total picks or total players signed. The total number of picks is probably 10-fold higher and the total number of players signed is probably 4-fold higher.
The number of players with more than 20 WARP careers is pretty tight between 28-35 from 1978-1991. The 1992 total is much less and that draft was quite weak overall. The weakness of the 1992 draft is also something to keep in mind throughout the study.
The 40 over group is pretty tight between 14-18 for the years 1897-1989 followed by a big drop off. A large part of that decline is due to the younger current ages of successful HS picks from 1990-92. There are a handful of good HS picks that are still in the 20-40 range that will certainly finish above 40. Those players will raise the 40 over numbers at least for the 1990 and 1991 drafts.
That same issue is true of the 80 over group. The college heavy 1987-1989 drafts are already maxed out at ~5 players while the HS oriented 1990-92 drafts still have players working towards that level.
Part 2. So what does the 5th best player from a draft class look like anyway?
This section is to put names to the different classes of WARP. Its also a shorthand way to look at the depth in a draft class. These tables are just the 5th, 10th, 15th, etc best player, as ranked by WARP3 through 2003, for each draft class.
These three mostly college oriented drafts are generally made up of players whove reached their ultimate career WARP level. Of the players listed at 5-20 only Edmonds, DeShields, Navarro and Avery were not college products.
The first thing that struck me is how tight the three classes are. The 5th best player is right around the 80 WARP mark. The 10th best player is somewhere in the 40-60 WARP good career range. The 15th best player is just at the 40 WARP good mark. By the 20th best player were down to low 30s WARP careers that are useful, but aside from a couple great years from Avery not especially good. By the 30th best player were down to the low 20s and the bottom of the useful range. By the 40th best player were at the top of the useless range though Groom has had some good years as a LOOGY and Deion Sanders was a great football player.
In the past Ive made the claim that most drafts produce somewhere between 15-25 good players depending on how stringently you want to define good. I think this section is a pretty good indicator that any kind of stringent definition of good would lead you to use the low end of that scale. There arent a lot of good players coming out of each draft.
These drafts with much lower WARP totals are much more heavily weighted towards HS players who are still in the middle of their careers. Of the players in the 5th, 10th and 15th spots only Lieber, Wickman and Grudzielanek are college products.
Is this group of 5th best players going to reach the 80 WARP level? Pettitte will and Cameron might. Damon, from the weak 1992 draft, is a tougher call. The high end from these drafts probably wont be quite as high or deep as the 1987-1989 drafts.
However, at least for the good draft years of 1990 and 1991 I think these groups of 10th, 15th, etc best players will end up comparable to the players in the 1987-1989 years.
The following table is just the WARP averages for each slot for these three periods.
Again, from my perspective the 10th best player is pretty clearly good and the 20th best player is more some shade of useful and the 15th best is right on the borderline.
Im not sure how many of the specific names in the 5th best slot would be considered truly great by most fans, but that at least gives you a floor to the number of great players from a draft class. Its probably something less than 5.
Part 3. What is the average return per pick for every round?
Well, every round up to 30 and then just the ones that produced at least 30 WARP for the six years combined.
First column is simply the round.
Second column is the total WARP through 2003 for every pick in that round during the 1987-1992 period.
Third column is just that WARP total divided by 6 to give you an idea of what each round produced per draft year.
Fourth column is the total WARP divided by the number of picks. From 1987-1991 there were 26 MLB teams drafting and in 1992 there were 28 teams. The numerator is 158 for all of the full rounds. The first round supplemental totaled 56 picks during this period. One year MLB let a couple independent teams make picks in rds 4-10, but those picks arent considered official MLB picks even though one of those players turned into decent MLB player Mike Lansing. Ive included Lansing into the study, but otherwise ignored those picks. I meant to count up the number of players who actually signed, but havent had the time, but it wouldnt matter very much. If you figure 10-20 per round dont sign which is high for the early rounds that people care about then you drop the numerator to 138-148 and that doesnt change the final numbers much at all.
The last two columns are the low and high yearly totals for each rd up to 20. I added that because the average per pick can be somewhat misleading since its really driven by a few very good players trying to offset many useless players. It also shows that every round except the first will have completely non-productive draft years. Basically the difference between the low and high years is the presence or absence of one or two good to great players.
The WARP/pick column is the meat of the table so lets focus on that. Every pick on average is useless. The end.
Using an average to study something with such a small success rate is really problematic. The overwhelming number of failures easily swamp the rare successes and reduce every pick to an average of useless. Its better to look at the probability of finding those few successes. Ill do that in the next section.
The big thing is that the draft is not primarily a crapshoot. If it was youd expect the overall value to be randomly distributed. Thats not the case at all. Despite some fluke late round spikes the draft crapshoot is heavily weighted towards the top.
The first round clearly stands out from the rest of the draft and deserves to be thought of as something different from the rounds that follow. From this dataset, you can group rounds 2-6, rounds 7-8 and rounds 9-11 together. With a different set of years youd certainly find separate break points. The last time I did this it was more like rds 2-4, rds 5-8 and rds 9-11 or something like that.
The fact that rd 3 is higher than rd 2 is largely due to John Olerud who fell to the 3rd rd due to health concerns that were expected to lead to his return to college for his senior year. The Blue Jays took a flyer on a player who was a candidate for the #2 overall pick prior to his seizure and eventually signed him for a bonus that was similar to what the #1 overall pick, Ben McDonald, received. Players who fall for signability reasons and then sign bonuses well out of the norm for their round will always slightly skew these types of studies. I think Olerud is the only significant example in this dataset.
Thanks to Google you can find a 15 yr old story on the Olerud negotiations written by some guy named Jim Callis by searching "john olerud draft bonus".
In general it probably makes sense to think of the first round only as high picks, then rounds 2-6 as mid-rd picks, then rounds 7-13 as late round picks and everything else as complete flyers.
Part 4. What percentage of picks become useful, good and great?
This is a table of just players who have crossed the 20 WARP career threshold (with maybe a small handful below that who will make it, been a while since I put it together and I forgot. Sorry).
First column is just the round and it has every round that produced a 20 WARP player.
The next three columns are the Total Players in each group: 20-39.9, 40-79.9 and over 80.
The next two columns focus just on the the 40 WARP and over players which Ive dubbed Good to Great or G2G. There are some players in the 20 WARP category that will cross this threshold, but I didnt project any of them in. %Tot G2G is the percentage of Good to Great players in each round. For example, there are 25 of the 77 G2G players in the first round for %Tot of 32%.
The last three clumsily named columns are the percentage of picks that became 20 or 40 or 80 WARP players. Again, Im using the 158 picks as the denominator, but changing that to say 140 signed players wouldnt change very much.
Ill start at the bottom with the totals. On average each draft class produced 15.8 players with 20-39.9 WARP careers, 10.2 with 40-79.9 WARP careers and 2.7 players with 80 and more WARP careers. The number of Good to Great players on average was 12.8. If you assume 5-10 of the 20 WARP players will cross over into the G2G range you end up with about 15 useful players and 15 Good to Great players per draft class. Again, somewhere around 15 players seems to be the divide between good and useful.
Nearly one third of the G2G players came from the first round. Between the supplemental round and round 5 there is another 27%. Rounds 6-8 produced 11% for a total of 70% of the G2G players in the first 8 rounds. The other 30% is comprised of some fairly big spikes in rds a 11, 13, 17, 18 and 30 that produced 2 or 3 players and the rest randomly distributed down to Mike Piazza in rd 62.
Note that the first round is the only one to average more than one G2G player per draft class. By the 6th round most of the production spikes amount to one G2G player every 2 or 3 years.
In the %pick columns you see that 33% of 1st rd picks became at least useful though most of them are in the 20-39.9 WARP range. Only 11% of first rd picks became what I labeled good and just 4% became great. Its hard to find even modest success in any of the other rounds. In the %pick20 column only rounds 2 and 6 show a percentage that became useful over 3%. In the %pick40 column only rounds 3, 5 and 8 showed more than 1% who became good. In the %pick80 column rds 2-4 each have 1% chance to become a great player. There are only three other 1% chance rounds scattered throughout the draft.
Successful picks are exceedingly rare. When something exceedingly rare occurs my first inclination is to think that luck played a huge role. Lets look at the Sox 4th rd pick from 1989 Jeff Bagwell. On average 4th rd picks (including him) produced 3.1 WARP in their careers. Bagwell produced 120.8 through 2003. If we look at the preceding table we see that during this time period 1% (actually 0.6%) of 4th rd picks became great. If we widen the scope of comparable players to the 2nd to 6th rd we get the same 0.6%.
So, were the Sox lucky to draft a player worth roughly 40-fold more than the average 4th rder? If yes (if?), how much of the 117.7 WARP beyond the 3.1 average should be attributed to luck?
Purely off the top of my head, Id attribute at least 80% of that value to luck. That would leave the Sox scouting department with credit for an extra ~24 WARP, which is a still significant 8-fold over average, but it would leave most of that value - ~94 WARP - to pure luck. Considering the Sox gave away Bagwell for a couple months of Larry freakin Anderson, its probably pretty likely that most people would agree that Bagwells future value represented luck more than skill or competence.
Thats all well and good until you hypothetically - wonder over to Baseball Primer (or any sabremetric community) and toss out the idea that 6th rd pick Tim Hudson is mostly luck. That can become quite a show hypothetically speaking of course. We can do the same type of calculations that we did with Bagwell and use these tables to show that a sixth rd pick like Hudson might be expected to produce ~3 WARP of value and have a fraction of one percent chance to be great or good. And while some luck in the Hudson selection will eventually be acknowledged nobody really wants to say that the value of the Hudson pick was mostly luck, much less 80% luck. Of course, the As kept Hudson and have a rational and somewhat validated philosophy so isnt it easier to give the majority of credit to that rather than to luck?
Perhaps in this case it would be more accurate to say that 80% of the future value is skill and competence and only 20% is luck. To go back to Bagwells numbers that would lead us to give the Sox scouting staff credit for 94 WARP and leave luck with just the residual 24 WARP. Thats a huge difference on the single player level and it only gets magnified when we look at the team level in the next section.
In order to really evaluate draft philosophies you have to first have a handle on how divide successes between luck and skill. After all this research, Im personally comfortable calling 80% of draft success luck across the board. Im open to being convinced down from that baseline for specific teams over reasonable periods of time, but I do think thats the best way to initially look at rare draft successes.
Part 5. Team Draft Rank: 1987-1992
Finally a simple table. This is just the sum total WARP for every player drafted by the 26 teams that participated in all six drafts between 1987-1992. MLB added Florida and Colorado for 1992, but I left them out of the table.
One of the things I want to do, but havent had a chance is to try to see if the rank order of these 1987-1992 drafts correlates at all with team winning percentage in say, 1995-2001. I would guess that any correlation would be weak, but it is at least interesting to see the Indians dominate this ranking a they would the AL in the mid to late 90s.
From the Ave column you can see group breaks from #2-7, #8-12, #13-22 and #23-26.
The team average is 388 WARP (or just 65 WARP per draft) and that falls nicely in the big gap between #12 KC and #13 Texas. If you use that as an MLB average you can see that some teams wildly exceeded or missed that league average. The Indians were nearly 300 WARP over league average. Thats spread out over a lot of years, but its a significant advantage. One the opposite end teams like the Cubs and Phillies were 200 WARP below league average. Thats a lot of ground to make up just to get back to average. I know the divisions were different back then, but how would you like to be a fan of Detroit (ranked 19 with 281 WARP) trying to compete with Cle (ranked 1 with 680 WARP) and the White Sox (ranked 2 with 589 WARP)? Actually, that mismatch played itself out in the mid to late 90s and beyond pretty well.
On an organization level are these huge differences luck? If you think its 80% luck, then the portion of the Indians huge 300 WARP advantage attributable to skill is only 60 WARP. If you think its 80% skill, then the portion attributable to skill is 240 WARP.
As a first pass attempt to shed some light on that question I looked at each draft and put it in one of three bins Top (rank 1-8 ) , Middle (rank 9-18 ) , Bottom (rank 19-26). How consistent are teams from year to year? As an aside, MLB as an industry is not very consistent at all. Rank order correlation from year to year is just over 0.1.
The first thing I looked for is the number of times a team was in the same bin at least 4 times. That was true for 8 teams (just under 1/3). Toronto, Seattle, Houston and Montreal all were in the Top four times. Atl was in the Middle all six drafts. SF was in the Middle four times. LA was in the Bottom four times and the Cubs were in the Bottom five times.
There are at least some teams that show some consistency, but there are also teams like Hou and Minn who were either in the Top or Bottom. Theres also a team like LA that was in the Bottom four times and finished in the Top once thanks to their 62nd rd pick.
Is this a reasonable way to moderate the rather extreme 80/20 luck/skill issue that I presented? Perhaps individual picks that appear wildly lucky from teams that over a period of several years are consistently good drafters should be viewed as 40/60 luck/skill while teams that appear consistently bad should have their Piazza successes viewed as 80/20 luck/skill. It adds a layer of complexity to the analysis, but some kind of several year context is probably necessary.
Part 6. Timelines
The last part I promise and actually the one new thing that I really wanted to do. After every draft the newest and greatest crop of prospects immediately gets tagged with an eta and they all seem to be pretty aggressive and in most cases irrelevant. Were not really interested in when a prospect arrives (usually for a meaningless Sept callup); we want to know when hell be a good player capable of contributing to a winning team.
For each of the 900+ players from these drafts I recorded their season by season WARP in addition to their career WARPs that I used in the previous tables. I was then able to determine the total WARP produced by each class in each year after the draft. I set the draft year as year zero and then just went up to year 16 for the 1987 draft and year 11 for the 1992 draft. In an effort to save space Im going to first present the years 0-11 that are covered for all six drafts.
The top line on the table is Draft Year 0 to 11.
The next two lines give the ideal age of HS and C players drafted at 18 and 21 in year 0. That is supposed to give a little perspective on the timing of those two different classes of players.
The rest of the table is the summed WARP values rounded to the nearest whole number to again save some space.
Ill start at the bottom with the six year Total. Theres very little production in years 0-2. Year 3 shows a jump in production, but the overall total is still pretty modest. What production there is very ealry comes from very, very good players like Jef Bagwell, Frank Thomas, John Olerud (all college) and Ken Griffey.
Theres a large 300 WARP jump between year 3 and year 4 and that should be considered the first year that a draft class starts to be productive. In absolute terms it is by far the biggest year to year jump. Note in year 4 college players are starting their physical peaks at 25 while HS players, at 22, are still a few years away. Year 5 is a modest jump and serves as a transition year between the year 4 breakthrough and the year 6 start of a broad peak. The absolute peak for this six year period is year 8 (college at 29 and HS at 26), but there is a broad plateau peak between years 6-10. Those five years seem to be a draft classes sweet spot.
Year 11 shows a perhaps surprisingly small decline and note that its still higher than year 5. Draft classes seem to wind up slower and longer than you might think, but last longer. Even the college heavy 1987 and 1989 drafts dont show much decline between year 10 (age 31) and year 11 (age 32).
The next table is the incomplete years 12-16 for these drafts. To make the totals somewhat comparable to years 0-11 I filled in the missing years with the average values of the years that do have data. Thats reasonably good for years 12 and 13, but probably pretty rough for the last few years.
The interesting thing to me is that years 12 and 13 are still a relatively slow decline even though the players are 30/33 and 31/34. The total WARP isnt much different that year 5 when players are 23/26.
Im not sure if this makes the data easier to understand, but this is the same data expressed as a percent of peak. I set each draft class top WARP as 100% and compared every year to that.
All of the peak years are between years 7-9. Year 4 is the first year over 50% of peak value and that value again steps up to a solid peak between years 6-10.
This is years 12-16 in the same way as above.
Again you see years 12 and 13 holding up reasonably well and comparable to year 5.
Since Ive made the point that 1987-89 were college heavy drafts and 1990-92 were more HS oriented, I thought it might be interesting to see how those different drafts progressed. These are average WARP totals for the two different three year periods.
Clearly the more college oriented drafts ramp up more quickly with a substantial advantage in years 4 and 5 when college picks are 25/26 and HS picks are 22/23. Interestingly, the gap is fairly stable at ~20 WARP in years 6-8 and ~10 WARP in years 9-11. Considering how weak the 1992 draft is its also a bit surprising that the gap is that small. It looks like the peak years 6-11 are pretty comparable.
This is definitely going to be the last series of tables. The previous set dealt with cumulative WARP for every player who made the MLB. This set deals with the number of solid, very good and great seasons that occur in each year after the draft. I set a 4-6.9 WARP season as a solid, productive season. You can argue that players a bit below that are making decent contributions too, but I wanted the low end of the range to be a bit better than just a decent contribution. The next group is 7-9.9 WARP. At these levels players are at least fringe All-Stars. A 10 WARP or better season is clearly a superstar season.
This table is the three year average for the 1987-1989 draft classes. The top three lines are the normal years after draft and ideal ages of HS and college picks.
The next three lines are the number of seasons that fell into those three WARP ranges. I stopped at year 14 because thats the last year that includes all three draft classes.
Ov4 is the total number of seasons over 4 WARP and ov7 is just the number of seasons over 7 WARP.
For ov4 the first breakthrough is at year 4 and then after a plateau year there is second jump in year 6 to a broad peak that lasts thru year 11. This should be sounding pretty familiar by now.
Theres a year 8 peak of 22 seasons with more than 4 WARP, but the general peak is in the high teens. That fits with roughly 15 good players per class and a handful of other players who have a good year or two to mix in.
Interestingly, the very good ov7 group stays pretty consistent from year 4 thru year 11. These are generally excellent players like Griffey, Bagwell and Thomas who were very good players almost from the day of their MLB debuts.
This table is the same as above showing the three year average of the 1990-1992 draft classes.
The interesting thing to note is that this mostly younger group only seems to be delayed by about one year. The breakthrough is in year 5 instead of year 4 and the peak is in year 9 instead of year 8. The broader peak is the same years 6-11 though.
This table is the total number of seasons produced by all six draft classes.
Obviously the pattern is basically the same. Again, I think its interesting that the combined draft classes are as productive in year 11 (ages 29/32) as year 6 (ages 24/27).
This is just the average for all six draft classes.
It looks like the 2004 draft should start to produce double digit good seasons in 2009 and peak in 2012 at just under 20 good seasons. Its a long weight for a small number of truly meaningful successes.
Conclusions
The most amazing one is that its actually possible to do a draft project without obsessing over the C vs HS debate or using the word Moneyball. Fortunately or unfortuantely, I am going to get to that in another post that I can hopefully finish by the end of next week.
The biggest issue to come from this is the relatively stable and very low draft output over a everal year period. What happened in 1987-1992 doesnt guarantee that similar results have happened recently or will happen in 2004, but it does give a really good baseline for expectations. About 15 good players will come from a draft. Most of them will come from the first few rounds though all individual picks have very, very low probabilities of success. Aside from truly great players draft classes will start to produce value in year 4 and hit their peaks in year 6.
Those things are all true about the 1987-1992 drafts and are pretty likely to be true about the 2004 draft.
Any questions and issues or improvements in the study, Ill take a shot at answering.
References
My original draft lists were taken from Baseball America Almanacs 1988-1993.
In the last couple months The Baseball Cube put complete draft lists on the web with any players that made MLB highlighted. Those lists probably added 3 dozen fringe players to each draft class. Its a tremndous resource.
I did catch some mistakes in those lists so I cross referenced everything with the most recent BA Draft Almanac.
All WARP values come directly from Baseball Prospectus.
A new draft season brings a couple new draft studies. I think these are much better than the last round more focused, more thorough, a better stat. Either way, heres the first one. In the prior iteration I used Win Shares because it was published during the initial research and one of the stated uses was scoring drafts. I think WS does a pretty reasonable job of scoring and ranking hitters. It does a good job of scoring and ranking pitchers. It does a poor job of mixing hitters and pitchers.
One of the things I wanted to do was to record the number of solid and better seasons that each draft class produced. The nice round number of 15 WS worked well enough for hitters to establish the bottom of the solid range for hitters, but it excluded far too many pitchers who seemed to have had equally good years.
This problem of comparing pitchers to hitters has also been noted by studes, the person who runs www.baseballgraphs.com. At least amongst people who comment publicly he is probably the leading expert working with WS. Over the winter he did a study establishing baseline WS levels for each position. His WSAA (win share above average) addressed the problem of setting replacement level for pitching too low and he even found a passage from James book that suggested the need for this type of evolution for WS. Thats all well and good, but it undercuts the easy use of WS for any kind of study that compares position players and pitchers.
If youre interested that work can be found here:
www.baseballgraphs.com/bl...39_0_1_0_C
Enter WARP3. Wins Above RePlacement is the stat developed by Clay Davenport and is the basis of Baseball Prospectus historical DT cards. Like WS its a single number that represents a players contribution in all facets of the game. In this case the 3 means its translated across historical eras. Thats also an advantage because it already translates the strike shortened 1994 and 1995 seasons into full season stats for easy comparisons across draft classes.
Aside from the integration of hitters and pitchers, the two systems are in good agreement. A rank order of just hitters or just pitchers would probably show very little difference. The big disagreements would all come from mixing the two types and at least to my eye the mixed rankings of WARP3 make more sense.
This study only covers 1987-1992. The draft rules were changed in 1987 to eliminate the January and June Phases. Since these are the rules that continue to govern the draft, that makes 1987 an obvious starting point for a draft study. If you assume a HS player in 1987 was drafted at 18 and a college player was drafted at 21, then those players were 34 and 37 in 2003. Their careers arent over so you cant put a final number on the career value, but theyre all well past their peaks.
I used 1992 as the end of the range to be studied because I wanted all the players in the study to have at least played through their twenties peaks. An 18 year old HS pick in 1992 would have played 2003 at 29. A 21 year old college pick would have played 2003 at 32.
So the college players in this study are age 32-37 and the HS players are age 29-34. Im going to save most of the college vs HS comparisons for a second post, but the age difference is something to keep in mind. All of the college players are well past the traditional peak and are pretty close to their final WARP career total. Many of the HS players are still just passing their peaks and have several productive years ahead of them. This issue is exacerbated by the fact that the 1987-1989 drafts were very college heavy and the 1990-1992 drafts leaned more towards the HS side. Just about all of the best C players in the study are at least 35 while just about all of the best HS players are 29-31.
Part 1. How many and what type of players come from draft classes
The left hand column is the WARP3 total. The number in that column, eg 70 refers to the entire 10 WARP group 70-79.9. In the next section Ill put some names to WARP3 numbers to give you a better idea of what type of player fits in those ranges. In general, Ive used 20 WARP as the cutoff for a useful career. A good career could maybe start at 35, but 40 keeps it simple and doesnt change things too much. To keep it as a pattern of 20 WARP segments a very good career could be considered anything over the 60 WARP level and a great career is anything over 80 WARP. The key demarcations are really 20 (useful), 40 (good) and 80 (great) WARP.
WARP3 1987 1988 1989 1990 1991 1992 ave120 1 0 1 0 0 0 0.3110 1 0 1 0 0 0 0.3100 0 0 1 1 0 0 0.3 90 0 0 0 1 0 0 0.2 80 3 3 2 0 1 0 1.5 70 0 2 2 0 0 1 0.8 60 2 2 2 2 1 0 1.5 50 2 4 2 4 2 2 2.7 40 6 6 3 5 5 3 4.7 30 7 4 8 7 6 5 6.2 20 13 11 8 9 13 7 10.2 10 18 15 13 19 27 14 17.7 0 89 73 76 95 82 97 85.3 neg 21 23 21 27 22 26 23.380 over 5 4 5 1 1 0 2.740 over 15 18 14 12 9 6 12.320 over 35 33 30 28 28 18 28.720 less 128 111 110 141 131 137 126.3Total 163 144 140 169 159 155 155.0% 80 3 3 4 1 1 0 1.8% 40 9 13 10 7 6 4 8.1% 20 21 23 21 17 18 12 18.6useless 79 77 79 83 82 88 81.4
Lets start with the line Total thats about ¾ of the way down the table. I probably missed a few bad players here and there, but each of the drafts produced a lot of players who at least made the majors with a range of 140-169 players and an average of 155.
The line above Total "20 less" shows that the vast majority of those players do not have useful careers. For this reason studies that focus on the number of players who make the majors should be taken with a large grain of salt.
Of the players with useful career about half also exceeded the 40 WARP threshold and had good careers. A handful of those players also exceeded the 80 WARP threshold for a great career.
The last 4 lines of the table express the data as percentages. On average over 80% of the players who make the majors do not have useful careers. Less than 10% have good careers and less than 2% have great careers. Remember, those are percentages of players who made the majors, not total picks or total players signed. The total number of picks is probably 10-fold higher and the total number of players signed is probably 4-fold higher.
The number of players with more than 20 WARP careers is pretty tight between 28-35 from 1978-1991. The 1992 total is much less and that draft was quite weak overall. The weakness of the 1992 draft is also something to keep in mind throughout the study.
The 40 over group is pretty tight between 14-18 for the years 1897-1989 followed by a big drop off. A large part of that decline is due to the younger current ages of successful HS picks from 1990-92. There are a handful of good HS picks that are still in the 20-40 range that will certainly finish above 40. Those players will raise the 40 over numbers at least for the 1990 and 1991 drafts.
That same issue is true of the 80 over group. The college heavy 1987-1989 drafts are already maxed out at ~5 players while the HS oriented 1990-92 drafts still have players working towards that level.
Part 2. So what does the 5th best player from a draft class look like anyway?
This section is to put names to the different classes of WARP. Its also a shorthand way to look at the depth in a draft class. These tables are just the 5th, 10th, 15th, etc best player, as ranked by WARP3 through 2003, for each draft class.
1987 1988 1989 5 Steve Finley 83.1 Jim Edmonds 75.4 Tim Salmon 81.8 10 D DeShields 48.8 Charles Nagy 53.7 S Erickson 58.515 D Segui 40.8 W Williams 42.6 Ben McDonald 39.320 J Navarro 33.9 Steve Avery 31.8 Cal Eldred 32.830 Dave Hollins 25.3 Mark Wohlers 21.3 Brian Hunter 21.240 Buddy Groom 16.9 D Sanders 16.4 Joe Grahe 11.6
These three mostly college oriented drafts are generally made up of players whove reached their ultimate career WARP level. Of the players listed at 5-20 only Edmonds, DeShields, Navarro and Avery were not college products.
The first thing that struck me is how tight the three classes are. The 5th best player is right around the 80 WARP mark. The 10th best player is somewhere in the 40-60 WARP good career range. The 15th best player is just at the 40 WARP good mark. By the 20th best player were down to low 30s WARP careers that are useful, but aside from a couple great years from Avery not especially good. By the 30th best player were down to the low 20s and the bottom of the useful range. By the 40th best player were at the top of the useless range though Groom has had some good years as a LOOGY and Deion Sanders was a great football player.
In the past Ive made the claim that most drafts produce somewhere between 15-25 good players depending on how stringently you want to define good. I think this section is a pretty good indicator that any kind of stringent definition of good would lead you to use the low end of that scale. There arent a lot of good players coming out of each draft.
1990 1991 1992 5 A Pettitte 59.3 Mike Cameron 48.7 J Damon 45.9 10 R White 41.6 C Floyd 36.9 J Lieber 31.015 B Wickman 39.1 Grudzielanek 30.2 M Tucker 24.320 M Lansing 29.3 J Hamilton 23.6 C Counsell 18.430 J Bere 18.3 Hollandsworth 19.2 C Holt 10.740 T van Poppel 13.2 A Ochoa 15.0 Chad Fox 7.7
These drafts with much lower WARP totals are much more heavily weighted towards HS players who are still in the middle of their careers. Of the players in the 5th, 10th and 15th spots only Lieber, Wickman and Grudzielanek are college products.
Is this group of 5th best players going to reach the 80 WARP level? Pettitte will and Cameron might. Damon, from the weak 1992 draft, is a tougher call. The high end from these drafts probably wont be quite as high or deep as the 1987-1989 drafts.
However, at least for the good draft years of 1990 and 1991 I think these groups of 10th, 15th, etc best players will end up comparable to the players in the 1987-1989 years.
The following table is just the WARP averages for each slot for these three periods.
1987-1989 1990-1992 1987-1992 5 80.1 51.3 65.710 53.7 36.5 45.115 40.9 31.2 36.120 32.8 23.8 28.330 22.6 16.1 19.340 15.0 12.0 13.5
Again, from my perspective the 10th best player is pretty clearly good and the 20th best player is more some shade of useful and the 15th best is right on the borderline.
Im not sure how many of the specific names in the 5th best slot would be considered truly great by most fans, but that at least gives you a floor to the number of great players from a draft class. Its probably something less than 5.
Part 3. What is the average return per pick for every round?
Well, every round up to 30 and then just the ones that produced at least 30 WARP for the six years combined.
First column is simply the round.
Second column is the total WARP through 2003 for every pick in that round during the 1987-1992 period.
Third column is just that WARP total divided by 6 to give you an idea of what each round produced per draft year.
Fourth column is the total WARP divided by the number of picks. From 1987-1991 there were 26 MLB teams drafting and in 1992 there were 28 teams. The numerator is 158 for all of the full rounds. The first round supplemental totaled 56 picks during this period. One year MLB let a couple independent teams make picks in rds 4-10, but those picks arent considered official MLB picks even though one of those players turned into decent MLB player Mike Lansing. Ive included Lansing into the study, but otherwise ignored those picks. I meant to count up the number of players who actually signed, but havent had the time, but it wouldnt matter very much. If you figure 10-20 per round dont sign which is high for the early rounds that people care about then you drop the numerator to 138-148 and that doesnt change the final numbers much at all.
The last two columns are the low and high yearly totals for each rd up to 20. I added that because the average per pick can be somewhat misleading since its really driven by a few very good players trying to offset many useless players. It also shows that every round except the first will have completely non-productive draft years. Basically the difference between the low and high years is the presence or absence of one or two good to great players.
Rd WARP3 WARP/rd WARP/pick Rd Low Rd High1 2791.0 465.2 17.7 375.5 573.91s 390.3 65.1 7.0 33.4 120.72 617.8 103.0 3.9 41.0 242.13 787.4 131.2 5.0 37.4 293.44 493.4 82.2 3.1 14.2 197.45 519.0 86.5 3.3 19.6 145.56 495.4 82.6 3.1 17.4 132.47 374.1 62.4 2.4 0.7 113.38 366.1 61.0 2.3 -0.1 190.3 9 182.2 30.4 1.2 1.1 50.310 187.6 31.3 1.2 0.1 70.511 280.2 46.7 1.8 3.0 145.112 127.8 21.3 0.8 5.2 61.813 350.9 58.5 2.2 1.2 158.014 121.4 20.2 0.8 -0.1 50.215 91.2 15.2 0.6 0.5 36.816 48.1 8.0 0.3 20.117 225.2 37.5 1.4 2.7 90.318 203.8 34.0 1.3 0.0 87.519 35.6 5.9 0.2 15.920 223.3 37.2 1.4 118.821 54.8 9.1 0.3 22 81.0 13.5 0.5 23 55.8 9.3 0.424 124.6 20.8 0.825 79.0 13.2 0.526 11.8 2.0 0.127 14.5 2.4 0.128 73.9 12.3 0.529 30.4 5.1 0.230 124.8 20.8 0.832 55.1 9.2 0.336 32.1 5.4 0.243 55.8 9.3 0.444 39.6 6.6 0.345 41.8 7.0 0.358 47.9 8.0 0.362 84.2 14.0 0.5Total 10,109.0 64.0
The WARP/pick column is the meat of the table so lets focus on that. Every pick on average is useless. The end.
Using an average to study something with such a small success rate is really problematic. The overwhelming number of failures easily swamp the rare successes and reduce every pick to an average of useless. Its better to look at the probability of finding those few successes. Ill do that in the next section.
The big thing is that the draft is not primarily a crapshoot. If it was youd expect the overall value to be randomly distributed. Thats not the case at all. Despite some fluke late round spikes the draft crapshoot is heavily weighted towards the top.
The first round clearly stands out from the rest of the draft and deserves to be thought of as something different from the rounds that follow. From this dataset, you can group rounds 2-6, rounds 7-8 and rounds 9-11 together. With a different set of years youd certainly find separate break points. The last time I did this it was more like rds 2-4, rds 5-8 and rds 9-11 or something like that.
The fact that rd 3 is higher than rd 2 is largely due to John Olerud who fell to the 3rd rd due to health concerns that were expected to lead to his return to college for his senior year. The Blue Jays took a flyer on a player who was a candidate for the #2 overall pick prior to his seizure and eventually signed him for a bonus that was similar to what the #1 overall pick, Ben McDonald, received. Players who fall for signability reasons and then sign bonuses well out of the norm for their round will always slightly skew these types of studies. I think Olerud is the only significant example in this dataset.
Thanks to Google you can find a 15 yr old story on the Olerud negotiations written by some guy named Jim Callis by searching "john olerud draft bonus".
In general it probably makes sense to think of the first round only as high picks, then rounds 2-6 as mid-rd picks, then rounds 7-13 as late round picks and everything else as complete flyers.
Part 4. What percentage of picks become useful, good and great?
This is a table of just players who have crossed the 20 WARP career threshold (with maybe a small handful below that who will make it, been a while since I put it together and I forgot. Sorry).
First column is just the round and it has every round that produced a 20 WARP player.
The next three columns are the Total Players in each group: 20-39.9, 40-79.9 and over 80.
The next two columns focus just on the the 40 WARP and over players which Ive dubbed Good to Great or G2G. There are some players in the 20 WARP category that will cross this threshold, but I didnt project any of them in. %Tot G2G is the percentage of Good to Great players in each round. For example, there are 25 of the 77 G2G players in the first round for %Tot of 32%.
The last three clumsily named columns are the percentage of picks that became 20 or 40 or 80 WARP players. Again, Im using the 158 picks as the denominator, but changing that to say 140 signed players wouldnt change very much.
Rd Total Players 20 40 80 G2G %Tot G2G %pick20 %pick40 %pick80 1 29 18 7 25 32 18 11 41s 5 3 0 3 4 3 2 02 8 2 1 3 4 5 1 13 5 4 2 6 8 3 3 14 2 2 2 4 5 1 1 15 4 5 0 5 6 3 3 06 10 2 0 2 3 6 1 07 4 2 0 2 3 3 1 08 5 4 0 4 5 3 3 0 9 3 1 0 1 1 2 1 010 3 0 0 0 0 2 0 011 2 2 0 2 3 1 1 012 0 1 0 1 1 0 1 013 3 1 2 3 4 2 1 114 2 0 0 0 0 2 0 015 1 0 0 0 0 1 0 017 0 1 1 2 3 0 1 118 2 2 0 2 3 1 1 020 3 1 0 1 1 2 1 0 21 1 0 0 0 0 1 0 022 0 1 0 1 1 0 1 024 1 1 0 1 1 1 1 028 0 1 0 1 1 0 1 030 1 2 0 2 3 1 1 032 0 1 0 1 1 0 1 043 0 1 0 1 1 0 1 044 1 0 0 0 0 1 0 045 0 1 0 1 1 0 1 048 0 1 0 1 1 0 1 058 0 1 0 1 1 0 1 062 0 0 1 1 1 0 0 1Total 95 61 16 77
Ill start at the bottom with the totals. On average each draft class produced 15.8 players with 20-39.9 WARP careers, 10.2 with 40-79.9 WARP careers and 2.7 players with 80 and more WARP careers. The number of Good to Great players on average was 12.8. If you assume 5-10 of the 20 WARP players will cross over into the G2G range you end up with about 15 useful players and 15 Good to Great players per draft class. Again, somewhere around 15 players seems to be the divide between good and useful.
Nearly one third of the G2G players came from the first round. Between the supplemental round and round 5 there is another 27%. Rounds 6-8 produced 11% for a total of 70% of the G2G players in the first 8 rounds. The other 30% is comprised of some fairly big spikes in rds a 11, 13, 17, 18 and 30 that produced 2 or 3 players and the rest randomly distributed down to Mike Piazza in rd 62.
Note that the first round is the only one to average more than one G2G player per draft class. By the 6th round most of the production spikes amount to one G2G player every 2 or 3 years.
In the %pick columns you see that 33% of 1st rd picks became at least useful though most of them are in the 20-39.9 WARP range. Only 11% of first rd picks became what I labeled good and just 4% became great. Its hard to find even modest success in any of the other rounds. In the %pick20 column only rounds 2 and 6 show a percentage that became useful over 3%. In the %pick40 column only rounds 3, 5 and 8 showed more than 1% who became good. In the %pick80 column rds 2-4 each have 1% chance to become a great player. There are only three other 1% chance rounds scattered throughout the draft.
Successful picks are exceedingly rare. When something exceedingly rare occurs my first inclination is to think that luck played a huge role. Lets look at the Sox 4th rd pick from 1989 Jeff Bagwell. On average 4th rd picks (including him) produced 3.1 WARP in their careers. Bagwell produced 120.8 through 2003. If we look at the preceding table we see that during this time period 1% (actually 0.6%) of 4th rd picks became great. If we widen the scope of comparable players to the 2nd to 6th rd we get the same 0.6%.
So, were the Sox lucky to draft a player worth roughly 40-fold more than the average 4th rder? If yes (if?), how much of the 117.7 WARP beyond the 3.1 average should be attributed to luck?
Purely off the top of my head, Id attribute at least 80% of that value to luck. That would leave the Sox scouting department with credit for an extra ~24 WARP, which is a still significant 8-fold over average, but it would leave most of that value - ~94 WARP - to pure luck. Considering the Sox gave away Bagwell for a couple months of Larry freakin Anderson, its probably pretty likely that most people would agree that Bagwells future value represented luck more than skill or competence.
Thats all well and good until you hypothetically - wonder over to Baseball Primer (or any sabremetric community) and toss out the idea that 6th rd pick Tim Hudson is mostly luck. That can become quite a show hypothetically speaking of course. We can do the same type of calculations that we did with Bagwell and use these tables to show that a sixth rd pick like Hudson might be expected to produce ~3 WARP of value and have a fraction of one percent chance to be great or good. And while some luck in the Hudson selection will eventually be acknowledged nobody really wants to say that the value of the Hudson pick was mostly luck, much less 80% luck. Of course, the As kept Hudson and have a rational and somewhat validated philosophy so isnt it easier to give the majority of credit to that rather than to luck?
Perhaps in this case it would be more accurate to say that 80% of the future value is skill and competence and only 20% is luck. To go back to Bagwells numbers that would lead us to give the Sox scouting staff credit for 94 WARP and leave luck with just the residual 24 WARP. Thats a huge difference on the single player level and it only gets magnified when we look at the team level in the next section.
In order to really evaluate draft philosophies you have to first have a handle on how divide successes between luck and skill. After all this research, Im personally comfortable calling 80% of draft success luck across the board. Im open to being convinced down from that baseline for specific teams over reasonable periods of time, but I do think thats the best way to initially look at rare draft successes.
Part 5. Team Draft Rank: 1987-1992
Finally a simple table. This is just the sum total WARP for every player drafted by the 26 teams that participated in all six drafts between 1987-1992. MLB added Florida and Colorado for 1992, but I left them out of the table.
Total Ave1 CLE 680.0 113.32 CWS 589.2 98.23 TOR 574.8 95.84 SEA 560.2 93.45 MINN 547.6 91.36 BAL 536.6 89.47 HOU 527.1 87.98 CAL 477.2 79.59 MON 450.3 75.110 NYY 449.9 75.011 BOS 435.5 72.612 KC 417.6 69.613 TEX 331.5 55.314 ATL 327.9 54.715 LA 317.4 52.916 NYM 306.9 51.217 STL 303.1 50.518 OAK 290.9 48.519 DET 281.2 46.920 PITT 276.5 46.121 MIL 268.8 44.822 CINN 257.6 42.923 SF 222.9 37.224 SD 199.0 33.225 PHL 188.9 31.526 CUB 155.7 26.0TOTAL 10079.3 11679.9
One of the things I want to do, but havent had a chance is to try to see if the rank order of these 1987-1992 drafts correlates at all with team winning percentage in say, 1995-2001. I would guess that any correlation would be weak, but it is at least interesting to see the Indians dominate this ranking a they would the AL in the mid to late 90s.
From the Ave column you can see group breaks from #2-7, #8-12, #13-22 and #23-26.
The team average is 388 WARP (or just 65 WARP per draft) and that falls nicely in the big gap between #12 KC and #13 Texas. If you use that as an MLB average you can see that some teams wildly exceeded or missed that league average. The Indians were nearly 300 WARP over league average. Thats spread out over a lot of years, but its a significant advantage. One the opposite end teams like the Cubs and Phillies were 200 WARP below league average. Thats a lot of ground to make up just to get back to average. I know the divisions were different back then, but how would you like to be a fan of Detroit (ranked 19 with 281 WARP) trying to compete with Cle (ranked 1 with 680 WARP) and the White Sox (ranked 2 with 589 WARP)? Actually, that mismatch played itself out in the mid to late 90s and beyond pretty well.
On an organization level are these huge differences luck? If you think its 80% luck, then the portion of the Indians huge 300 WARP advantage attributable to skill is only 60 WARP. If you think its 80% skill, then the portion attributable to skill is 240 WARP.
As a first pass attempt to shed some light on that question I looked at each draft and put it in one of three bins Top (rank 1-8 ) , Middle (rank 9-18 ) , Bottom (rank 19-26). How consistent are teams from year to year? As an aside, MLB as an industry is not very consistent at all. Rank order correlation from year to year is just over 0.1.
TOP MIDDLE BOTTOMCLE 4 2 0CWS 3 2 1TOR 2 4 0SEA 4 1 1MINN 3 0 3BAL 3 3 0HOU 4 0 2CAL 3 1 2 26 13 9 MON 4 2 0NYY 3 2 1BOS 1 2 3KC 3 1 2TEX 2 2 2ATL 0 6 0LA 1 1 4NYM 1 3 2STL 2 1 3OAK 1 3 2 18 23 19 DET 1 3 2PITT 1 3 2MIL 1 3 2CINN 0 3 3SF 0 4 2SD 0 3 3PHL 0 3 3CUB 1 0 5 4 22 22
The first thing I looked for is the number of times a team was in the same bin at least 4 times. That was true for 8 teams (just under 1/3). Toronto, Seattle, Houston and Montreal all were in the Top four times. Atl was in the Middle all six drafts. SF was in the Middle four times. LA was in the Bottom four times and the Cubs were in the Bottom five times.
There are at least some teams that show some consistency, but there are also teams like Hou and Minn who were either in the Top or Bottom. Theres also a team like LA that was in the Bottom four times and finished in the Top once thanks to their 62nd rd pick.
Is this a reasonable way to moderate the rather extreme 80/20 luck/skill issue that I presented? Perhaps individual picks that appear wildly lucky from teams that over a period of several years are consistently good drafters should be viewed as 40/60 luck/skill while teams that appear consistently bad should have their Piazza successes viewed as 80/20 luck/skill. It adds a layer of complexity to the analysis, but some kind of several year context is probably necessary.
Part 6. Timelines
The last part I promise and actually the one new thing that I really wanted to do. After every draft the newest and greatest crop of prospects immediately gets tagged with an eta and they all seem to be pretty aggressive and in most cases irrelevant. Were not really interested in when a prospect arrives (usually for a meaningless Sept callup); we want to know when hell be a good player capable of contributing to a winning team.
For each of the 900+ players from these drafts I recorded their season by season WARP in addition to their career WARPs that I used in the previous tables. I was then able to determine the total WARP produced by each class in each year after the draft. I set the draft year as year zero and then just went up to year 16 for the 1987 draft and year 11 for the 1992 draft. In an effort to save space Im going to first present the years 0-11 that are covered for all six drafts.
The top line on the table is Draft Year 0 to 11.
The next two lines give the ideal age of HS and C players drafted at 18 and 21 in year 0. That is supposed to give a little perspective on the timing of those two different classes of players.
The rest of the table is the summed WARP values rounded to the nearest whole number to again save some space.
0 1 2 3 4 5 6 7 8 9 10 11 HS 18 19 20 21 22 23 24 25 26 27 28 29 C 21 22 23 24 25 26 27 28 29 30 31 32 87 2 5 30 72 130 165 198 157 215 190 191 18988 1 12 31 89 150 166 163 206 185 182 175 131 89 0 16 57 83 139 120 188 205 212 191 173 16690 2 9 32 61 67 122 171 176 166 188 180 16491 0 1 22 39 97 139 193 197 186 198 191 18092 0 2 10 42 86 112 125 133 173 149 130 113Total 4 45 181 387 668 824 1039 1073 1136 1098 1040 943
Ill start at the bottom with the six year Total. Theres very little production in years 0-2. Year 3 shows a jump in production, but the overall total is still pretty modest. What production there is very ealry comes from very, very good players like Jef Bagwell, Frank Thomas, John Olerud (all college) and Ken Griffey.
Theres a large 300 WARP jump between year 3 and year 4 and that should be considered the first year that a draft class starts to be productive. In absolute terms it is by far the biggest year to year jump. Note in year 4 college players are starting their physical peaks at 25 while HS players, at 22, are still a few years away. Year 5 is a modest jump and serves as a transition year between the year 4 breakthrough and the year 6 start of a broad peak. The absolute peak for this six year period is year 8 (college at 29 and HS at 26), but there is a broad plateau peak between years 6-10. Those five years seem to be a draft classes sweet spot.
Year 11 shows a perhaps surprisingly small decline and note that its still higher than year 5. Draft classes seem to wind up slower and longer than you might think, but last longer. Even the college heavy 1987 and 1989 drafts dont show much decline between year 10 (age 31) and year 11 (age 32).
The next table is the incomplete years 12-16 for these drafts. To make the totals somewhat comparable to years 0-11 I filled in the missing years with the average values of the years that do have data. Thats reasonably good for years 12 and 13, but probably pretty rough for the last few years.
12 13 14 15 16HS 30 31 32 33 34C 33 34 35 36 37 87 166 163 103 76 4688 106 101 94 7989 124 121 8890 142 12391 14092Total 812 763 570 464 274
The interesting thing to me is that years 12 and 13 are still a relatively slow decline even though the players are 30/33 and 31/34. The total WARP isnt much different that year 5 when players are 23/26.
Im not sure if this makes the data easier to understand, but this is the same data expressed as a percent of peak. I set each draft class top WARP as 100% and compared every year to that.
0 1 2 3 4 5 6 7 8 9 10 11 HS 18 19 20 21 22 23 24 25 26 27 28 29 C 21 22 23 24 25 26 27 28 29 30 31 32 87 1 2 14 33 60 77 92 73 100 89 89 8888 0 6 15 43 73 81 79 100 90 89 85 64 89 0 8 27 39 66 57 89 97 100 90 82 7890 1 5 17 33 36 65 91 94 88 100 96 8791 0 1 11 20 49 71 98 100 94 100 97 9192 0 1 6 24 50 65 72 77 100 86 75 65Total 0 4 15 32 56 69 87 90 95 92 88 79
All of the peak years are between years 7-9. Year 4 is the first year over 50% of peak value and that value again steps up to a solid peak between years 6-10.
This is years 12-16 in the same way as above.
12 13 14 15 16HS 30 31 32 33 34C 33 34 35 36 37 87 77 76 48 35 2188 52 49 45 3889 58 57 4290 76 6691 7192Total 67 62 45 37 21
Again you see years 12 and 13 holding up reasonably well and comparable to year 5.
Since Ive made the point that 1987-89 were college heavy drafts and 1990-92 were more HS oriented, I thought it might be interesting to see how those different drafts progressed. These are average WARP totals for the two different three year periods.
0 1 2 3 4 5 6 7 8 9 10 11 HS 18 19 20 21 22 23 24 25 26 27 28 29 C 21 22 23 24 25 26 27 28 29 30 31 32 87-89 1 11 39 82 139 150 183 189 204 188 180 16290-92 1 4 21 47 83 124 163 169 175 178 167 152
Clearly the more college oriented drafts ramp up more quickly with a substantial advantage in years 4 and 5 when college picks are 25/26 and HS picks are 22/23. Interestingly, the gap is fairly stable at ~20 WARP in years 6-8 and ~10 WARP in years 9-11. Considering how weak the 1992 draft is its also a bit surprising that the gap is that small. It looks like the peak years 6-11 are pretty comparable.
This is definitely going to be the last series of tables. The previous set dealt with cumulative WARP for every player who made the MLB. This set deals with the number of solid, very good and great seasons that occur in each year after the draft. I set a 4-6.9 WARP season as a solid, productive season. You can argue that players a bit below that are making decent contributions too, but I wanted the low end of the range to be a bit better than just a decent contribution. The next group is 7-9.9 WARP. At these levels players are at least fringe All-Stars. A 10 WARP or better season is clearly a superstar season.
This table is the three year average for the 1987-1989 draft classes. The top three lines are the normal years after draft and ideal ages of HS and college picks.
The next three lines are the number of seasons that fell into those three WARP ranges. I stopped at year 14 because thats the last year that includes all three draft classes.
Ov4 is the total number of seasons over 4 WARP and ov7 is just the number of seasons over 7 WARP.
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14HS 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 C 21 22 23 24 25 26 27 28 29 30 31 32 33 34 354 0 1 2 5 7 7 12 10 16 11 14 10 8 11 97 0 0 1 2 4 5 3 4 5 5 4 5 3 1 110 0 0 0 1 1 1 2 2 1 2 1 2 1 1 0ov4 0 1 3 8 12 13 17 17 22 18 19 17 11 13 11ov7 0 0 1 3 5 6 5 6 6 7 5 7 4 3 2
For ov4 the first breakthrough is at year 4 and then after a plateau year there is second jump in year 6 to a broad peak that lasts thru year 11. This should be sounding pretty familiar by now.
Theres a year 8 peak of 22 seasons with more than 4 WARP, but the general peak is in the high teens. That fits with roughly 15 good players per class and a handful of other players who have a good year or two to mix in.
Interestingly, the very good ov7 group stays pretty consistent from year 4 thru year 11. These are generally excellent players like Griffey, Bagwell and Thomas who were very good players almost from the day of their MLB debuts.
This table is the same as above showing the three year average of the 1990-1992 draft classes.
0 1 2 3 4 5 6 7 8 9 10 11 HS 18 19 20 21 22 23 24 25 26 27 28 29 C 21 22 23 24 25 26 27 28 29 30 31 32 4 0 0 0 4 5 8 9 10 11 14 9 107 0 0 1 0 1 2 4 4 5 4 4 310 0 0 0 0 0 0 0 1 1 1 1 1ov4 0 0 1 4 5 10 14 14 17 19 14 14ov7 0 0 1 0 1 2 5 4 6 5 5 4
The interesting thing to note is that this mostly younger group only seems to be delayed by about one year. The breakthrough is in year 5 instead of year 4 and the peak is in year 9 instead of year 8. The broader peak is the same years 6-11 though.
This table is the total number of seasons produced by all six draft classes.
0 1 2 3 4 5 6 7 8 9 10 11 HS 18 19 20 21 22 23 24 25 26 27 28 29 C 21 22 23 24 25 26 27 28 29 30 31 32 4 0 2 6 25 36 46 62 60 79 76 70 597 0 0 5 8 13 20 22 24 32 28 25 2410 0 0 2 2 3 4 7 8 4 8 4 8ov4 0 2 13 35 52 70 91 92 115 112 99 91ov7 0 0 7 10 16 24 29 32 36 36 29 32
Obviously the pattern is basically the same. Again, I think its interesting that the combined draft classes are as productive in year 11 (ages 29/32) as year 6 (ages 24/27).
This is just the average for all six draft classes.
0 1 2 3 4 5 6 7 8 9 10 11 HS 18 19 20 21 22 23 24 25 26 27 28 29 C 21 22 23 24 25 26 27 28 29 30 31 32 4 0 0 1 4 6 8 10 10 13 13 12 107 0 0 1 1 2 3 4 4 5 5 4 410 0 0 0 0 1 1 1 1 1 1 1 1ov4 0 0 2 6 9 12 15 15 19 19 17 15ov7 0 0 1 2 3 4 5 5 6 6 5 5
It looks like the 2004 draft should start to produce double digit good seasons in 2009 and peak in 2012 at just under 20 good seasons. Its a long weight for a small number of truly meaningful successes.
Conclusions
The most amazing one is that its actually possible to do a draft project without obsessing over the C vs HS debate or using the word Moneyball. Fortunately or unfortuantely, I am going to get to that in another post that I can hopefully finish by the end of next week.
The biggest issue to come from this is the relatively stable and very low draft output over a everal year period. What happened in 1987-1992 doesnt guarantee that similar results have happened recently or will happen in 2004, but it does give a really good baseline for expectations. About 15 good players will come from a draft. Most of them will come from the first few rounds though all individual picks have very, very low probabilities of success. Aside from truly great players draft classes will start to produce value in year 4 and hit their peaks in year 6.
Those things are all true about the 1987-1992 drafts and are pretty likely to be true about the 2004 draft.
Any questions and issues or improvements in the study, Ill take a shot at answering.
References
My original draft lists were taken from Baseball America Almanacs 1988-1993.
In the last couple months The Baseball Cube put complete draft lists on the web with any players that made MLB highlighted. Those lists probably added 3 dozen fringe players to each draft class. Its a tremndous resource.
I did catch some mistakes in those lists so I cross referenced everything with the most recent BA Draft Almanac.
All WARP values come directly from Baseball Prospectus.
