An Intro to Horse Evaluation Part 3: Visualizing Speed Data
By Lambert Chu
It’s time for a new chapter of horse evaluation guides - but this time with an extra-spicy dose of data!
As Ryan explained extensively in the first two parts of the series, we can evaluate a ZED horse by its maximum ability and ability deviation, which can be measured using speed data.
In this post, we’ll revisit these concepts with the help of data visualizations. We’ll also dive into the speed data of a few interesting horses, which will hopefully provide useful insights for your own horse evaluations.
Let’s start with a refresher - why do we even need to consider speed data at all? ZED already provides odds for each horse in every race, which is super informative.
However, a horse’s odds can change dramatically depending on the other horses in the field - just look at how soft the competition can be in griffin races! Ideally, we’d like to evaluate a horse against all other horses in the game, not just against other horses they’re actively competing against in their class or prize pool.
This is where finish times come in. They are synonymous with horse speed, since faster horses have lower finish times. Finish times are the common denominator in all ZED races, and they allow us to compare horses across different race classes and prize pools (e.g. free vs. paid races), even ones that have never raced against each other before.
For this to work, we have to prove that finish times are consistent across different classes and levels of competition in ZED. But in the interest of keeping this blog post digestible, we’ll save the deep dive into finish time consistency for a future post.
The “Average” Horse
Let’s take a closer look at finish times we can expect from horses in ZED.
To start, let’s plot ALL finish times that have ever been recorded, grouped by distance. By looking at these plots, we can quickly see how finish times are distributed.
Check out these clean bell curves! The finish times follow normal distributions, so the average outcome is also the most common outcome. All outcomes are symmetric around the mean, which means a horse is just as likely to finish 1 second faster than average as it is to finish 1 second slower than average.
Since these plots use every race time ever recorded, we can consider them to be the plots for an “average” horse in ZED. This will be a key reference point for analyzing individual horses.
Major Key Alert: Ability Deviation
If you’re not first (or second or third), you’re last! As you all probably know, only the top three positions win cash prizes in ZED. On top of that, finishing in the last four positions (9 through 12) subtracts class points from your horse, which could allow it to move down in race class and face weaker competition. It’s clear that you want your horse to finish at the very front or very back of the field and NOT in the middle.
For this reason, ability deviation is a key factor for winning races. We can quantify that by looking at the standard deviation of finish times. The higher the standard deviation, the more likely a horse will have an abnormally fast or abnormally slow time - which leads to more finishes in first place (amazing result) and last place (not too bad of a result).
Check out the finish time distributions for the KYH boys' very own Vanilla Bean:
Vanilla Bean has a “wider” and “flatter” distribution than the average horse, which means she’s more likely to run way faster or way slower than the average horse. The standard deviation of her finish times is way higher than the average horse’s, which leads to massive ability deviation and very impressive maximum ability. This is why she has so many 1st and 12th place finishes but rarely finishes in the middle of the pack, making her the prototypical “U-shaped” horse.
Is Deviation the Only Key to Winning?
Vanilla Bean’s high ability deviation is a major key to success, but can a horse with lower ability deviation still win races?
One of my horses, Quest Land, looks like she might fit the bill of a winning low-variation horse. She actually finishes faster than average 71.4% of the time in 2200m races.
At that rate, you’d think she’d pull great odds and be a very consistent winner… and she actually was in free races! But unfortunately, once I started entering her in paid races and at higher classes, she was quickly outcompeted. Quest Land went from a positive ROI in lower classes to a negative ROI in Class 3. 😩
Why is Quest Land no longer winning consistently? When you compare its finish time distribution to Vanilla Bean’s, the explanation becomes more obvious.
Check out the left tail of the plot. Vanilla Bean (orange line) has a much higher likelihood of finishing quickly compared to Quest Land and the average horse. To win consistently at high levels of competition, your horse has to be able to hit that level of speed.
Vanilla Bean is faster than average only 46.7% of the time, but because of her outstanding ability deviation, she’s way more likely to place in the top three. Although Quest Land is more likely to finish ahead of Vanilla Bean on average, she isn’t more likely to win the race.
In fact, Quest Land has never finished with a time outside of two population standard deviations from the population mean time. On the other hand, Vanilla Bean has finished faster than two population standard deviations 9% of the time.
On the flip side, let’s look at another elite horse in the ZED world, The Crimson Chin. Crimson also finishes above-average 71% of the time just like Quest Land, but why is she so much more dominant?
In her preferred distance of 1000m, Crimson has an outstanding average time of 57.17 seconds, which is significantly faster than the average time of 57.7 seconds.
Crimson’s times also have a fairly high standard deviation of 1.06 - which is higher than the global average of 0.88, but NOT as high as Vanilla Bean’s 1.31.
Despite a high-but-not-insanely-high ability deviation, Crimson’s maximum ability is still one of the best in the game for 1000m races. The key is that she finishes faster than two standard deviations from the population mean at an absurd 11% of the time.
Therefore, we can’t say that a horse is a winner based on its average finish time alone or its standard deviation alone. Instead, we have to put both factors together to determine its ability to finish consistently in the top percentile of times across all horses in the game.
Deviation vs. Odds
Now we’ll take a broader look at how odds are related to finish times.
Let’s split all horses into two groups: “beasts” and “donkeys.” We’ll consider a beast to be any horse that pulls odds less than 8 in a race, while a donkey is any horse that pulls odds greater than 40 in a race. Then, we’ll record the finish time that horse ended up with for that given race.
Here’s a plot of these two groups (along with the average curve) for 2200m races:
As you would expect, the beasts (orange line) are far more likely to have fast finish times. The donkeys (green line) do NOT necessarily run slower than average! Instead, they’re just way more likely to have an average time and finish in the middle of the pack. Their ability deviation is tiny!
For reference, we can see the same exact effect in 1000m races (and other distances):
Hopefully, this blog post got you thinking about how finish times and speed data can be useful for evaluating horses!
There is one major drawback of speed to keep in mind: you’ll need a decent sample size of races before drawing any conclusions. With odds, you’ll only need a few races at each distance to know whether or not your horse is a consistently profitable winner. With finish times as the only source of data, I personally wouldn’t be comfortable drawing conclusions until there are at least 20 races for a given distance.
All of this is still a working theory for how ZED RUN functions and should be taken with a grain of salt. But we look forward to collaborating with the community to understand just how this game works. Stay tuned for more analysis, and keep your eyes peeled for new tools on the horizon!