Applying the Pareto Principle (80-20 Rule) to Baseball

Pareto works best with counts or other data that must be positive so maybe WAR isn’t the best stat to use. Maybe runs created would be better.

How about doing a split comparing pitchers and hitters?

What percentage of players does it take to reach 80% of run production?

And in this case, isn’t WAR almost “misguided”? Negative WAR doesn’t mean negative production, it means less than replacement production.

As a result, wouldn’t it make more sense to look at it from a zero baseline?
For example, Jose Guillen has been responsible for say, 20 created runs while a replacement player would have created 25 runs at this point — that results in him being a -.5 WAR player; however, he’s still responsible for producing 20 runs, regardless of whether that’s better or worse than a AAAA-scrub.

I reran your study but limited it to position players. I did that so I could use traditional stats rather than the esoteric WAR. And since pitchers aren’t in MLB for their batting prowess I removed their offense as well.

According to fangraphs, 949 players had at least one PA in 2009. 344 of those were pitchers. Take them out and you’re left with 605. 20% of 605 is 121. So I compared the totals for the Top 121 batters (in terms of PA’s) to the MLB non-pitcher totals.

PA’s: The Top 121 had 76,502 PA’s out of 181,051. So 20% of the players got 42.3% of the opportunities to produce offensively. How well did they do?

Hits: They got 19,225 hits in 67,521 AB’s, a .285 BA. All non-pitchers went 42,836 for 160,769, .266. So the Top 20% were indeed more productive than the rest, as you’d expect. But nowhere near 80%. In fact in percentage terms the difference looks insignificant: they had 42% of the AB’s and 44.9% of the Hits. How about extra-base hits?

XBH: The Top 20% had 6,787 XBH out of 14,590 for all non-pitchers, which is 46.5%. So yes, they did better in this category than they did as singles hitters (44.0%). But once again the difference is surprisingly small. How about HR’s?

HR: 2,493/5,015 = 49.7%. So we can say 20% of the batters produced 50% of the HR’s if we round off. (Of course I could get a higher number if I select the Top 20% in terms of HR’s hit instead of PA’s, since some of them were not included in my Top 20%. But I think that’s pushing it.) That’s the highest percentage I found for any of the traditional stats, with one exception: Intentional Walks.

IBB: 616/1,179 = 52.2%.

Almost every stat I looked at fell in the range between 42% and 50%. The only exceptions beside IBB’s were:

Sacrifice Bunts: 27.7%. (You don’t ask you’re best hitters to bunt very often.)
Strikeouts: 39.1%.
Triples: 41.3%.

What about the bottom line, Runs and RBI? The Top 20% produced 46.1% of R, 46.7% of RBI.

So there you have it. Yes the Top 20% of batters produced more than their share of the offense. But it’s mostly because they got more opportunities. They come to the plate 42% of the time and produce between 42 and 50% of the results. Nowhere near 80%.

For professional purposes, I use Pareto charts all of the time so I thought I’d share a few observations that are consistent with fjm235’s findings…

- This is most effective for smaller numbers of items. Looking across all players is almost certainly too large of a sample.

- When very large samples are used, the percentage contribution from 20% of the items looked at will rarely approach 80% and is more likely in the ranges seens above (40-50%).

- Let’s not pretend we’re debunking scientific law – the Pareto principle is simply a phenomenon that may have good practical application (in the quality world it is used to identify which defects occur most frequently so those can be addressed first) but has little theoretical basis other than the known fact that few things are distributed uniformly.

- We will most likely never get good results for baseball, because the sample we are looking at excludes players that do not perform well enough to play in the majors. So there’s a whole population of people that would be batting .050, for example, that would greatly inflate population size without increasing the total number of hits, BB, etc.

- When we say the “top 20%” it always has to be in terms of the metric of interest. Looking at the top 20% of PA and then counting HR is not good application of the pareto principle anyway.

One area we would most likely see the principle be effective is salaries…20% of the players are probably making close to 80% of the money.

I like the discussion thus far!