who does it best? east vs. west coast dressage

A couple of weeks ago I was catching up on The Eventing Radio Show, listening to old episodes from last year. Joe Meyer, who I love as a host and a rider, made the throwaway comment that he wouldn’t mind competing out here on the west coast because everyone does so well in dressage out here. It was meant as a joke, but it piqued my interest anyway. I have definitely complained with my west coast friends about the “easy” dressage judges out east, and I’m sure people in Texas are laughing at both coasts. But who is right?

This, it turns out, is an easy enough question to answer. So I took to Startbox and Event Entries and scraped for data on dressage scores. Then I fired up R Studio and went on a big fat fishing expedition (research slang for exploring a dataset for relationships instead of testing specific hypotheses).

The short answer? There’s no difference in average dressage score between the East and West coasts for FEI events.

Data collection & other details

There are hundreds and hundreds of events held across the US every year, and usually a few hundred people at each of those events. So to save my sanity when scraping the data, I stuck to rated events that offered the FEI levels. A few USEA-only events slipped through the cracks because I’d already opened the event and at that point, it was easier to just copy that data. I tried to get an even representation of events across the year and across the country, but of course there are more events on the East coast than anywhere else.

I ended up with a little more than 5,500 rider records before I called it quits on copying and pasting and reformatting.

For each rider record, I included the division, venue, date, state, area. I included columns that allowed me to pool similar divisions that aren’t exactly the same (CIC 2* and CCI 2*, or training and training 3-day). I also included whether the level in question was USEA or FEI rated.

I didn’t include any multi-level effects for for rider or dressage judge. Laziness was not the only reason for this — by ignoring the influence of the judge (for now), I could (kindof) see if the effects across different areas had to do with riders or judges. If the same 5 judges worked at every event across the US, then we’d expect their scores to be very reliable, and differences from coast to coast would have to do with rider differences.¬† Of course there’s many more than 5 judges and these effects wouldn’t be so obvious, but you can see what I’m getting at.

(Obviously if you have questions or quibbles, get in touch.)

Fun fact!  

Mean dressage score in the US? 34.756. Standard deviation is 5.19.

This is the density histogram of dressage scores across all levels. You can see that though the mean is 34.756, there is a peak in scores after that — right around a nice even¬†35.

So there you have it. That’s the average dressage score at (rated) events across the US.

If you’re trending below a 35 you should feel chuffed as you’re doing better than most! When you score below 30, you’re doing better than about 85% of the country. Below a 25, and you’re doing better than ~97%. Down below 20? You’re the 1%!!

(Non-eventers reading this, remember that lower dressage scores are better in our world.)

The questions

Mostly, I was interested in exploring the differences between the coasts, USEA areas, and states to see if dressage scores varied significantly from place to place. And for the most part, the differences weren’t stark or necessarily significant.

At the FEI levels, there are no significant differences in dressage scores across major geographical areas of the us — east coast and west coast, the south, and the “mid”dle. If you’re not familiar with the model outputs, the important columns here are the Estimate and the asterisks. The intercept estimate represents the average dressage score of the east coast, and the estimates below are how much the other coasts differ from the east coast dressage average. So other areas of the country do have¬†slightly higher dressage scores on average than the east coast, but not significantly so. (I’ll get to the stargazing in a second.)

You can see this reflected in the density histogram at the top of the post. There’s a¬†lot of overlap between the dressage scores of east and west coasts at the FEI levels.

But how do the USEA levels stack up when you compare things from coast to coast?

Well, things aren’t quite so tidy. Let’s do some stargazing (those asterisks are typically thought of as good things in stats land)!

For events only sanctioned by the USEA, there east coast has a significantly lower average dressage score than any other area.  Riders in the middle of the country (basically Montana) are scoring nearly 2 points more, on average, than riders out east. People in the south get about 1.4 points more, and out west we get a measly 0.7 points more.

It’s important to note here that the “points” I’m talking about are percentage points, not raw points on tests.

Let’s break this relationship down a little more by state, shall we?

In this case, the intercept state is California. And what we see here is kinda neat! West coast states seem to line up (ish) in terms of scoring, which makes sense because they would probably pull from a very similar pool of judges. The second row in the table is Canada (from Bromont’s results).

But start comparing to the east coast, and we start to see some differences! Florida and Maryland in particular appear to be preeeetttyy generous with those dressage scores! Riders in those states score 1.36 and 2.55 points better on average, respectively. On the other hand, Montana is out there hammering their riders with dressage scores an average of 1.17 points worse than in California.

Because I’m California-centric, I plotted the distribution of scores in the lowest-scoring state (Maryland) vs. California, to show what a significant difference here looks like. It’s not a HUGE split between the curves, but you can see that there are quite a few more¬†riders in Maryland in the sub-30 zone than you see in California.

So what does it all mean?

There could be lots of reasons that I found significant differences between the dressage scores of different states and areas. For one, I didn’t apply a single correction for multiple tests to this data set, and I explored tons and tons of potential relationships. Statistically speaking,¬†one of them was¬†bound to come up significant.

Could it be that riders out east are just better than riders in other states? Ummmm. I mean sure, this is one possible answer. But given the variance between states up and down the eastern seaboard, I’m not sure this hypothesis holds up.

It’s also possible that judges on the west coast and in Montana are much harsher and stricter than on the east coast. I’ve heard people say that it’s hard to “make it” in dressage in California because there’s such strong competition down here in the form of Steffen Peters and Hilda Gurney. Perhaps the presence of Olympic-level riders makes judges more strict? If so, one would expect a similar effect in Florida. And… well, the data doesn’t quite hold up to that either (but also isn’t designed to answer that question).

It seems like there might also be some hyper-local effects on the east coast, since it’s so densely populated out there. This might be due to the fact that judges out east don’t need to have a very wide travel radius in order to judge plenty of shows. If those judges tend to score a little better, then that would create pockets of shows where scores are a little more generous.

Anyway, there’s lots of potential reasons for these trends. I just enjoyed looking at them!¬†What I can pretty confidently say is that for 2018 at least, riders in California were not getting preferential treatment from dressage judges (erhrrm, Joe!).