CDIW Amsterdam 2015

The dressage world was once again dazzled by the performance of Valegro at the CDIW last weekend in Amsterdam. In the Grand Prix they would have taken a new Word record if the final Piaffe had not gone wrong. Plugging in her “usual” score of 9.6 for the Piaffe and 9.5 for the Transitions would have had them reach a final score of at least 87.86 and edge above their Olympia record of 87.74. Oh well, that’s for another day….

The first place was never seriously in doubt, and the second place both in the Grand Prix (78.32) and in the Freestyle was safely secured by Danielle Heijkoop and Siro. Rather than compare them with the other riders in Amsterdam we instead compare them with Edward Gal and Glock’s Undercover at Olympia (80.12) a month earlier. In the spider-web plot we show the scores of each for each figure type:

Comparing the two we see that the canter, passage, piaffe work is essentially identically scored between the two. Siro had a bad Rein-back in Amsterdam and Undercover really picked up points in the Trot Half-Passes while doing slightly better in the Collected Walk and Halts, both combinations show excellent Piaffe scores of 9. Two strong Dutch combinations that we will hopefully see in Aachen in the summer.

Another interesting feature of the Grand Prix in Amsterdam was the very close scores of the riders in 3^rd and 4^th places. 74.88 for  Jessica von Bredow-Werndl and Unee BB of Germany and 74.86 for  Diederik van Silfhout with Arlando of the Netherands. (A combined score difference summing all judges of just 0.5 points!) But again, the spider-plot shows us they have different strengths:

Unee BB picks up points in the Halt, Rein-Back and Walk paces, while Arlando has the stronger Canter movements. Both combinations have very similar Trot and Passage notes though possibly Unee BB had a stronger Piaffe.

Now a little observation about judging differences.  The full results of the Grand prix are visible on the Sports Computer Graphics site at http://results.scgvisual.com/2015/amsterdam/r10.html and the event was as well judged as any other event at this level. But as everyone is aware, sometimes there are large judging differences that are often poorly understood. First the important news, there were no big changes of rank due to any single judge. Swapping in and out the judges would result in the 3^rd, 4^th and 5^th places flip-flopping but that’s because they were very very closely scored as noted above. One ride (Danielle Heijkoop) had a score difference of 5.9%, that’s the difference between one judge (in this case Isabelle Judet at M) and the average of her 4 colleagues. But note, this is “pretty normal”, with 105 judgements it is very likely to have one of them more than 5% different. When these big differences occur it is always because one judge is consistently higher or lower than their colleagues across the whole test, it is impossible for a ”mistake” –such as a missed error in a change line – to introduce such big differences.

Big scoring differences are seldom about position of the judge either. But it only takes being low or high by 0.5 points in each figure to end up 5% different at the end of the day. That’s one of the reasons the half-points were introduced, and I think one of the reasons we will eventually move to decimal scoring. (0.1 point precision for the scoring per figure)

These big differences always occur because for essentially the entire ride a judge is either consistently higher or lower than their colleagues. In this case the judge at M averages 0.5 points below the judge at B and more like around 2 points less for every Piaffe and Transition. It is not about a different view of the test – over the entire test then such effects average out, it is about a different overall appreciation of the ride.

How important is it? Well in this case, if the judge at M’s notes had not been taken into account the final score for Danielle would have been 79.50 instead of 78.32. If the proposal currently being considered by the FEI to correct the very low or very high score to the next nearest judges score were applied, then she would have ended up with a 79.24. If instead a more balanced approach of removing the highest and lowest score for every figure had been adopted, then her final score would have been 78.433, a very small change compared to the awarded score, but perhaps one that has more justice to it, in this case the majority of the jury (for each figure) would determine the final score and a single high or low judge would neither strongly influence the final score nor risk having their scores entirely dropped.

All this just to help you think about the possibilities of improving the system that can be considered.

Freestyle

First Analysis plot from yesterdays Freestyle. Comparision of Valegro and Damon Hill scows for each figure. What is plotted is Valegro Score minus Damon Hill score taking into account the various coefficients. So when this is positive (above the axis) the difference is in Valegro’s favor, when it is below it is in Damon Hill’s favor. Valegro is ahead in almost all Technical notes, and more so in the Artistic notes.

Spinning the Numbers for the 2014 World Equestrian Games

With just a week before the WEG gets underway I’ve been checking out the numbers in the GDA database to see how things might pan out in the Team and Individual competitions. The goal is not to make the actual competition uninteresting but to make it even more exciting by pointing out the possibilities and perhaps where the surprises might come. To do this I analyze the history of all the combinations and for the top combinations their strengths and weaknesses and their progression over the year in the principle movements. Of course, luckily, it is sport, there are two athletes out there in each combination that can have good or bad days, but on the whole combinations don’t do much better than normal in any given competition, though things can of course go wrong in big ways.

The Team event will be decided by the sum of the best 3 rides for a country in the Grand Prix. This time around (unlike in London) the 3 member team does not need to be declared before the competition, it’s the standard CDIO format where the best 3 of 4 counts. Also unlike London, the Special is for the Individual medal and does not contribute to the team score.

The whole dressage world knows that it is almost inconceivable for GER not to win the Gold medal this year! The depth of talent in the DOKR riders and horses is formidable. They will be fielding 4 combinations all of whom have gone above 80% in the Grand Prix. If they lose completely even their top combination they can still end up 5-6 points clear of their nearest competitors. The only plausible way for them not to be on the top of the podium is to have two of their riders eliminated, nobody on the planet could want that, so they seem to be assured of reaping the benefits of their immensely strong Dressage program.

This year round NED looks to be safely in Silver position with a 5-point lead over GBR, they also have two combinations that can perform over 80%. But, if they lose one of those riders, then they go dead-level with GBR for a Silver/Bronze fight. GBR however has one really strong combination, they should be safely in Bronze position, but the weight of the team success rests on Charlotte’s shoulders. Any problem there and they would almost certainly drop out of medal contention.

And waiting to fight it out in 4-6 position are three very close teams, SWE, USA and DEN. If Valegro goes according to form, it is not likely that these three will be in the medals, but SWE and USA are almost level and ready to step in if GBR has any slip-up. If Don Aurelio and Scandic come out at the top of their form and Santana or Darcia hit their top too, then they can get within a couple of points of a plausible GBR score. They have everything to go for. By the same measure Legolas and Verdedas are on a steep curve of improvement and right on the threshold of the podium too.

With Digby out of retirement for this competition, DEN has also put its best team forward and they can also be very close to SWE and USA. The USA had  a set-back with Steffen’s illness before Aachen, but he came out well in Verden and  Laura Graves’ rise has been meteoric. I think the motivation for these three teams is immense and the competition will be exciting to watch, they have to be at their best, because if GBR has any problem, then one of them gets to the podium and books their tickets to Rio 2018.

Table showing predicted scores for the top 10 teams. The other columns show what could happen if the team hits their best form and what could happen if the predicted top combination on the team had a major problem.

So, what about the Individual Competition? This is decided just on the Grand Prix Special and will be between the top 30 riders from the Grand prix, there are no limitations on the number of riders per nation that can do the Special. I think it is going to take around 71.something, maybe close to 72, to qualify for the Special, so it is going to be a very select club in this competition. The Special has even more weight on the Passage-Piaffe-Transitions (PPT) than the Grand Prix so this is going to tip things slightly in the favor of those horses with excellent PPT. The top 10 riders coming out of the Grand Prix are likely to be dominated by GER again, there is a good chance that all 4 Germans are in that group, 3 Dutch and one each from GBR, SWE and AUT. But the podium is much more difficult to predict! Valegro, Damon Hill and Totilas must be favorites but Bella Rosa, Undercover and Parzival will be fighting to step up onto the podium too! Before Aachen, I don’t think we would be having this discussion, but after Aachen everything is possible. I wrote an article for the WDM magazine in February fantasizing about the impossible duel between Edward and Charlotte on Totilas and Valegro, but reality caught up with my article, the duel is real but it is Matthias who brings Totilas to this competition not Edward. I think Valegro is still the favorite to win, but lets look at the details and see how this can go. It won’t be a two-way competition, Damon Hill is a force to be reckoned with, and Edward’s new horse Undercover has some magic movements too.

As the table shows, each of these top horses has different strengths, no single one is the best in everything, Damon Hill comes out top on paces and submission and has the most spectacular walk; Valegro in trot, changes, impulsion and submission; Totilas in pirouettes and impulsion; Undercover in piaffe and transitions; Parzival in changes, Bella Rosa in any half-pass. We are used to the idea that there is a champion, but the reality is different, they are all very close at the top.

By the way, if one single horse had all these best qualities we would see a score of about 88.5.

The Top Performers for each of the Grand Prix Special figure types, where more than one name is shown they are equal top!

My final prediction for the Individual medals: Valegro, Damon Hill, Totilas in that order! The numbers come out that way and I personally appreciate the harmony that Helen and Damon Hill show, so I sort of hope it works out this way too.

Predicted Final Scores for the top 10 Individuals.

CDI Scores in 2012. How do you compare?

In this picture we show the distribution of all CDI Grand Prix scores in 2012. The blue bars show what fraction (percentage) of rides received a particular final score. You see the typical bell-shape curve with an average score of 66.28%. From this distribution we can produce the red curve, where we show what percentage of rides score above any given value.

The distribution of CDi Grand Prix scores in 2012. On the left axis and blue the percentage of all results for each % of final score. In red, and on the right axis, the percentage of rides scoring above a given score

You can use this to see what percentile your average or your latest score lies in. For example, if you score a 71.5% you can read off the red line on the right hand axis that about 12.5% of rides scores this or above, or looked at the other way around 87.5% of all rides scored lower than this.

(For those interested in statistics the standard deviation of this distribution is 4.8%)