THE BOOK--Playing The Percentages In Baseball

Where is this research from?

I don’t believe the conclusions (that peak off. age is now almost 30) for a second!

For one thing, this guy does not explain enough of his methodology or give enough details on the data to critique the study, so I would have to duplicate it. 

There are so many potential problems (especially selective sampling) and traps with these aging pattern studies, that I think he may have fallen into one of them.

I also don’t have any idea what smaller parks would have to do with againg patterns, and I am not sure that an emphasis on power rather than speed should affect aging patterns for overall offense either.  IOW, I think that fast players with little power will show the same peak age as slow, power players.

The only possible thing that might make his conclusion valid (and the data will speak for itself) is steroids.  There may be enough older players who took steroids in the late 90’s and early 00’s to make it “look” as if peak offensive age has shifted.

I don’t think using mean age is the best for judging these trends.  One should at least look at the median and mode as well to get a true picture of what is going on.  I think it is very interesting that the curve for peak age is not a steady upward progression.  I would think that it would be a steady upward progression if improved health care and conditioning were the main causal factor.  Since his numbers seem to show that players who start their careers later peak later, wouldn’t that imply that major league experience plays a significant role along with innate physical skill in reaching peak performance?  And since he is only measuring offensive value, does it really take 6 or 7 years of professional experience to learn to be the best hitter you can be?  Is it players being drafted out of college that is the main cause of later starting age or is it players with long term contracts blocking younger players from moving up?  Certainly a lot of areas for future research.

I think MGL hit the nail on the head when he mentioned steroids as the probable cause.  Also, don’t forget amphetamines, which were the performance enhancers of choice for most players in the late 1980’s and early 1990’s.

This is just something I did in under 10 minutes, so take it FWIW.

I only looked at OBP, since I did work on this a few weeks ago, when I looked at z-Scores, and the spread in hitting talent.  Here’s the process:
1 - Take each hitter’s z-score in OBP, minimum 200 PA
2 - Find the year with his highest z-score
3 - Figure his age as year minus birth year
4 - Repeat for all players
5 - For each birth year, find the average age of each player’s highest z-score

I looked at each birth year from 1854 to 1966.  The average age is 28, and it pretty much oscillates between age 26 and 30.

As recently as birth year 1946, the average peak age was 26.6.  Birth year 1955 has an average peak age of 26.8.

On the flip side, the highest peak age was 30.1, and this happened in birth year.... 1966.  Which is the last year in my study.  Birth years 1964-1966 had an average peak age of 29.2.

If I extend the study to birth year 1973 (meaning the players were as young as 32), the peak age hovered at almost 29 years old (average of 28.7 from 1967 to 1973).  I have about 50 players in each birth year.

So, I’d say the peak age has probably increased by about 1 year. 

Remember, I’m only looking at OBP.

***

Now, what if I only looked at players’ career between the ages of 24 and 34?  That is, if some guy peaked at age 23 or 35, then I discard that year, and look for his next best peak.

Once again, the average is 28.  Once again, birth year 1966 has a peak average of 30.  And, once again, birth years 1967-1973 averages 28.7.

***

Finally, what if I only looked at players between ages of 24 and 30?

Expectedly, the average peak age drops (to 27.2).  This is expected, since I threw away any late peaks.  The oldest a player could peak, under these guidelines was 30.

However, once again, the birth year 1966 is near the top of all years at age 28.0.  (A couple of other years are at 28.2).  Birth Year 1973 is also at 28.0.  The 1967-1973 birth years averaged 27.6, which is higher than the overall average of 27.2.

***

When I do 7-year rolling averages, I get a peak of 29.0 with the 1910 birth year (meaning 1907-1913).  From there, until birth year 1944, it was a steady decline to a peak age of 27.2.  From that point onwards, until today, peak age has grown steadily back up until birth year 1969 (i.e, 1966-1972), with a peak age of 29.0.  This is also the last year in the study.

This growth is not without precedent.  But, if it continues to grow, it will be without precedent.

Interesting.  I’m a little concerned about using a 200 PA minimum, though.  Someone (like Barry Bonds) could miss well over half the season due to injury and still turn in a very high OBP.  For me, peak performance requires durability, not just occasional brilliance.  Could you run it with a 500 PA minimum?  How about OPS (or 1.8*OBP+SLG)?

Don’t forget, I said I used z-scores.  So, if we assume a league average OBP of .3333, then the following all have 3 SD from the mean:

200 PA: .433 OBP
400 PA: .404 OBP
600 PA: .391 OBP