Chad Peltier
Guest
Ohio State’s 2019 recruiting class is really good, but still step down from recent years
Chad Peltier via our friends at Land-Grant Holy Land
Visit their fantastic blog and read the full article (and so much more) here
Box plots help show how the 2019 class is skewed more than the total composite ratings or mean player average suggest.
The narrative for the Ohio State Buckeyes’ newly-signed 2019 recruiting class depends on how you look at the recruiting data.
If you just casually look at the 247 Composite rankings, the Buckeyes’ 14th ranking isn’t very impressive.
But, as Morgan and Richard have written, that ranking probably isn’t a big deal, and looking at the class using a different metric — average player rating — shows the Buckeyes with a top-3 class.
That’s because the 247 Composite rankings uses both quality (a player’s rating) and quantity (number of commits) to determine the rankings. Ohio State’s small class of just 17 players was far smaller than anyone else in the top-25. In fact, the smallest class in the top-10 had 24 players, or 41% more players than the Buckeyes’ class. If you look only at recruit quality — using the average player ranking — then Ohio State’s class ranks third with an average rating of 91.87, behind Alabama’s 94.38 and Georgia’s 93.32.
But even ranking classes using the mean player rating can still hide the fact that classes may be skewed in one direction or another.
For example, a 20-player class with a mean player rating of .9000 could mean 10 players at 1.000 and 10 players at .8000, or it could mean all 20 players are ranked at .9000. You don’t know whether a class is skewed by a few awesome players (and/or by a few lower-ranked recruits) without looking at the full distribution of player ratings.
Box plots can help with that. I took Ohio State’s last five classes, removed any special teams signees (who skew the class’s player rating because they’re all ranked relatively low), and created box plots for each class.
The lines off each box show the upper and lower range of each class – the highest and lowest ranked recruits. The box’s top and bottom represent the upper and lower quartiles of player rankings – the top and bottom 25% — so the box itself represents the interquartile range (IQR). The line in the middle(ish) of the box shows the median player ranking – the middle player rating in the full distribution of player rankings. Finally, the “x” shows the class’s mean, or average, rating.
Box plots can help us identify whether recruiting classes are skewed by a few super-highly ranked or super-lowly ranked recruits. A skewed class can be seen by the length of the lines coming out of the box and whether the median line is higher or lower inside the box.
Also, the full range of player ratings (i.e., from .99 to .85) doesn’t tell us a whole lot here – every OSU class has some players rated as five-stars and some rated around .85 (three stars). The range doesn’t say a lot about where most players in the class rank. But the size and position of the interquartile range – the box – does tell us a good bit about how the class is distributed. A bigger box means a larger IQR, and therefore a wider distribution of player ratings for the middle 50% of the class.
OK — here are the box plots:
/cdn.vox-cdn.com/uploads/chorus_asset/file/13754686/OSU_crootin.jpg)
Here are my immediate takeaways:
/cdn.vox-cdn.com/uploads/chorus_asset/file/13754747/osu_crootin_meanmedian.jpg)
TL;DR: The 2019 class is good, but its rating is still skewed by a few highly rated recruits. Most of the class is lower ranked than in the previous few classes.
As multiple articles have pointed out, Ohio State’s 14th ranking in the 2019 247 Composite doesn’t do justice to the level of talent in the class. If you instead look at the average player rating, Ohio State jumps up to third in the country, behind only Alabama and Georgia.
But the mean player rating doesn’t indicate how skewed Ohio State’s 2019 class is, with most recruits rated lower-than-usual, but with three five-stars. Those three five-stars — Zach Harrison, Garrett Wilson, and Harry Miller — were more than all but four other schools, and of those four, only Georgia (5) had more five-stars than Ohio State.
The three five-stars raised the player average enough to mask the fact that most of the class was a little lower rated than normal. The median player rating (and box plots even more) shows this better than the mean:
/cdn.vox-cdn.com/uploads/chorus_asset/file/13754891/OSU_crootin_median.jpg)
If you remove Ohio State’s three five-stars, the average recruit is rated .9034 — a lower-rated four-star. If you remove the three highest rated players from the 2017 class, the average player is still rated .9439.
This definitely doesn’t mean that this is a bad class. The 2019 class is still phenomenal, and could absolutely win a championship under Ryan Day. The three five-stars look like no-brainer sure-things, and I’m really excited to see what other guys like Marcus Crowley, Cormontae Hamilton, and Jaden McKenzie do in Columbus.
But it’s also worth noting that the class as a whole is still a step down from the recent standard for Ohio State recruiting under Urban Meyer. It’s only a slight step down, and one class usually doesn’t mean much in the grand scheme, but it’s still worth noting.
Continue reading...
Chad Peltier via our friends at Land-Grant Holy Land
Visit their fantastic blog and read the full article (and so much more) here
Box plots help show how the 2019 class is skewed more than the total composite ratings or mean player average suggest.
The narrative for the Ohio State Buckeyes’ newly-signed 2019 recruiting class depends on how you look at the recruiting data.
If you just casually look at the 247 Composite rankings, the Buckeyes’ 14th ranking isn’t very impressive.
But, as Morgan and Richard have written, that ranking probably isn’t a big deal, and looking at the class using a different metric — average player rating — shows the Buckeyes with a top-3 class.
That’s because the 247 Composite rankings uses both quality (a player’s rating) and quantity (number of commits) to determine the rankings. Ohio State’s small class of just 17 players was far smaller than anyone else in the top-25. In fact, the smallest class in the top-10 had 24 players, or 41% more players than the Buckeyes’ class. If you look only at recruit quality — using the average player ranking — then Ohio State’s class ranks third with an average rating of 91.87, behind Alabama’s 94.38 and Georgia’s 93.32.
But even ranking classes using the mean player rating can still hide the fact that classes may be skewed in one direction or another.
For example, a 20-player class with a mean player rating of .9000 could mean 10 players at 1.000 and 10 players at .8000, or it could mean all 20 players are ranked at .9000. You don’t know whether a class is skewed by a few awesome players (and/or by a few lower-ranked recruits) without looking at the full distribution of player ratings.
Box plots can help with that. I took Ohio State’s last five classes, removed any special teams signees (who skew the class’s player rating because they’re all ranked relatively low), and created box plots for each class.
The lines off each box show the upper and lower range of each class – the highest and lowest ranked recruits. The box’s top and bottom represent the upper and lower quartiles of player rankings – the top and bottom 25% — so the box itself represents the interquartile range (IQR). The line in the middle(ish) of the box shows the median player ranking – the middle player rating in the full distribution of player rankings. Finally, the “x” shows the class’s mean, or average, rating.
Box plots can help us identify whether recruiting classes are skewed by a few super-highly ranked or super-lowly ranked recruits. A skewed class can be seen by the length of the lines coming out of the box and whether the median line is higher or lower inside the box.
Also, the full range of player ratings (i.e., from .99 to .85) doesn’t tell us a whole lot here – every OSU class has some players rated as five-stars and some rated around .85 (three stars). The range doesn’t say a lot about where most players in the class rank. But the size and position of the interquartile range – the box – does tell us a good bit about how the class is distributed. A bigger box means a larger IQR, and therefore a wider distribution of player ratings for the middle 50% of the class.
OK — here are the box plots:
Here are my immediate takeaways:
- Without knowing anything about box plots, you could look at the chart above and see that the 2019 class is a solid step down from the 2017 and 2018 classes, which were absurdly good.
- It’s also easy to see that the 2019 class most resembles the 2015 class in overall quality and the distribution of player ratings. For example, the 2015 third quartile is .9536 and 1st quartile is .8694 for an interquartile range (IQR) of .0842. 2019’s 3rd quartile is .9554 and 1st quartile is .8755 for an IQR of .0799. Those are the two largest IQRs of the last five years. They’re also the two lowest IQRs — 2016 is pretty close, but has a significantly higher floor than 2015’s or 2019’s.
- It’s pretty easy to tell which classes are skewed. The 2017 class has a crazy skew towards the top, with the median near the top of the box and a short top line and long lower line. The 2019 class is also skewed, but in the opposite direction, and less extreme. The median is only .0276 points from the 1st quartile, but .0523 points from the 3rd quartile.
- The 1st quartile (the bottom of the box) is a good way to look at the class’s floor — better than just looking at the low end of the range (all but 2018 are around .85). That’s because, without a box plot, we have no way of telling whether there’s just one recruit ranked at or near .85, or if most of the class is actually ranked near there. So, one of the main things that jumps out to me is just how (relatively) low the 2015 and 2019 1st quartiles (.87 and .88) are compared to the absurd 2017 and 2018 classes. In those latter two classes, 75% of the recruits were rated .92 or better.
- Similarly, in 2019, half of the recruiting class is rated .9031 or worse — that is, a low-four star or below. In comparison, in 2017, half of the class is rated .9712 or better.
- In fact, just looking at the class medians says a lot. From 2015 to 2019: .9003, .9294, .9695, .9517, and .9031. Medians are really helpful because they’re not affected by outliers or skewed data.
- The larger the difference between the mean and the median, the more the class is skewed. You can see the difference between the mean and median here, which again shows how skewed the 2017 and 2019 classes are, but in opposite directions.
TL;DR: The 2019 class is good, but its rating is still skewed by a few highly rated recruits. Most of the class is lower ranked than in the previous few classes.
As multiple articles have pointed out, Ohio State’s 14th ranking in the 2019 247 Composite doesn’t do justice to the level of talent in the class. If you instead look at the average player rating, Ohio State jumps up to third in the country, behind only Alabama and Georgia.
But the mean player rating doesn’t indicate how skewed Ohio State’s 2019 class is, with most recruits rated lower-than-usual, but with three five-stars. Those three five-stars — Zach Harrison, Garrett Wilson, and Harry Miller — were more than all but four other schools, and of those four, only Georgia (5) had more five-stars than Ohio State.
The three five-stars raised the player average enough to mask the fact that most of the class was a little lower rated than normal. The median player rating (and box plots even more) shows this better than the mean:
If you remove Ohio State’s three five-stars, the average recruit is rated .9034 — a lower-rated four-star. If you remove the three highest rated players from the 2017 class, the average player is still rated .9439.
This definitely doesn’t mean that this is a bad class. The 2019 class is still phenomenal, and could absolutely win a championship under Ryan Day. The three five-stars look like no-brainer sure-things, and I’m really excited to see what other guys like Marcus Crowley, Cormontae Hamilton, and Jaden McKenzie do in Columbus.
But it’s also worth noting that the class as a whole is still a step down from the recent standard for Ohio State recruiting under Urban Meyer. It’s only a slight step down, and one class usually doesn’t mean much in the grand scheme, but it’s still worth noting.
Continue reading...