We rely on statistical measures heavily in our minor league analysis. Over the years, we've made a lot of assumptions about what numbers can tell us. Are we on target? Let's start from square one.
Which team stats have the strongest correlation with runs scored?
| Correlations with Team Runs Scored | ||||||||
| Rk | Stat | Median | 2008 | 2007 | 2006 | 2005 | 2004 | 2003 |
| 1 | OPS | 0.9391 | 0.9437 | 0.9505 | 0.9344 | 0.8795 | 0.9712 | 0.8761 |
| 2 | wRC | 0.9319 | 0.9381 | 0.9485 | 0.9257 | 0.9107 | 0.9598 | 0.8786 |
| 3 | wOBA | 0.9316 | 0.9415 | 0.9552 | 0.9216 | 0.9046 | 0.9725 | 0.8872 |
| 4 | wRAA | 0.9310 | 0.9386 | 0.9562 | 0.9233 | 0.9060 | 0.9722 | 0.8873 |
| 5 | SLG | 0.8691 | 0.9033 | 0.8847 | 0.8534 | 0.7899 | 0.9285 | 0.8432 |
| 6 | OBP | 0.8434 | 0.8339 | 0.8750 | 0.8002 | 0.7829 | 0.8754 | 0.8529 |
| 7 | AVG | 0.7310 | 0.6797 | 0.7619 | 0.6726 | 0.7001 | 0.8032 | 0.7699 |
| 8 | ISO | 0.6898 | 0.6977 | 0.5892 | 0.6356 | 0.6819 | 0.7807 | 0.7622 |
| 9 | BABIP | 0.4934 | 0.5045 | 0.7257 | 0.5329 | 0.4032 | 0.4740 | 0.4823 |
| 10 | BB% | 0.4849 | 0.4900 | 0.3317 | 0.3836 | 0.4798 | 0.5542 | 0.6122 |
| 11 | BB/K | 0.4625 | 0.5086 | 0.2268 | 0.3474 | 0.4341 | 0.4909 | 0.7823 |
| 12 | Spd | 0.1429 | 0.0270 | 0.4769 | 0.0345 | 0.2513 | -0.0201 | 0.3821 |
| 13 | K% | -0.0694 | -0.0761 | 0.0120 | -0.0626 | 0.0342 | -0.1867 | -0.4353 |
Methods
We took every team's regular season offensive stats from 2003 to 2008 and correlated them, on a year-to-year basis, with runs scored. Note that while our data came from fangraphs.com, we recalculated walk and strikeout percentage (BB/PA and K/PA). View this link for more information on wRC and wRAA.
Analysis
I was hoping that we could piggyback the results of this study as a guide for weighting metrics to expose top prospects. That was an unrealistic plan. Assuming that MLB team stats can be used on individual minor leaguers is quite a leap in logic. But we still noticed some interesting results.
MLB strikeouts aren't that big of a deal
Mark Reynolds, Carlos Pena, Russell Branyan, Jim Thome, Adam Dunn, Brandon Inge, Justin Upton and David Wright are currently showing us that well-above-average hitters can be a full deviation above the league average in strikeout percentage, too. So it shouldn't be a big surprise that strikeout percentage only had a -0.0694 median correlation with runs scored from 2003 to 2008.
Does it hurt to watch your team's hitter strike out with the bases loaded? Absolutely.
Will teams with above-average strikeout rates typically be below-average in runs scored? No.
I wouldn't take this result and attempt to apply it to individual minor leaguers. There are plenty of examples that show how too many strikeouts in the minors can lead to careers that never take off. Besides, I tried to last night with the 2006 Texas League population that I used for a previous study and it didn't go well :).
OBP and SLG have very similar correlations
Ever heard the argument that OPS would be better if you gave on-base percentage more weight than slugging? I don't think so.
Note that slugging percentage has had a stronger correlation with team runs scored than on-base percentage in five of the last six years. Over the entire span of our study, the two stats came out with pretty similar correlations -- .8691 median slugging; .8434 OBP. Perhaps on-base percentage would have a stronger correlation with individual success than slugging percentage, but even if it did, doubt it would create much of a gap between the two stats.
Batting average
It's clear that batting average is a relatively weak tool for measuring a team's ability to score runs. But it actually has a solid correlation with runs scored.
Still, there are plenty of offensive statistics that have stronger correlatations to runs scored than batting average. If you look at batting average before any other offensive statistic, you're doing yourself a disservice.
OPS vs. wOBA
I've become a fan of weighted on-base average over the last few years because it makes more sense to me than on-base plus slugging (OPS). But the two stats have had very similar correlations to team runs scored -- OPS .9391 median from 2003 to 2008; wOBA .9316.
If someone prefers OPS because it's easier for them to calculate and think about than wOBA, I wouldn't hold it against them.
Potential Problems
The data is what it is, but analyzing it is a different story:
1. Team runs scored may not be the best way to measure offensive success.
2. We stretched the time span of our study from three years to six years because we saw some correlations varying more than we expected from year to year -- when I say we, I mean my research team leader Chris Vercammen and me. Our data became more precise as we took more years into account. Adding additional years could further sharpen our data, though spanning too many could complicate the data -- ex. how do you account for the effect of steroids or impact of altering the height of the pitcher's mound.
3. Assuming that team correlations reflect individual correlations may be a bit of a reach.
Questions about this study can be directed toward Adam Foster at adamf@projectprospect.com.