Umamusume Observations: April 2021

Just jotting down some thoughts about the game. Take it as you will. A hint of a spoiler for S2 of the anime down there but it’s not much of one.

Also this is one of those annoying articles with an image. in. between. every. paragraph.

The Umamusume game is just short of 2 months old as of this writing. It’s doing really well as far as organic growth and other KPI for a mobile F2P video game. Five million downloads is a lot of downloads! Localizations in Korean and Simplified Chinese were already announced and are in progress. The second season of the anime concluded and looks like the Blu-rays will sell well, if only due to trainers buying them for the in-game codes, as well as the announced live concerts in July and August.

I don’t follow JRA and Japanese horse racing, but thanks to the popularity of the game, folks on my radar are starting to take more notice of it. There was always a small otaku niche for this particular … vertical, as well as a more general awareness in Japan pop psyche along with other gambling sports (in the traditional sense, like boxing or whatever) that horse racing occupies. However thanks to Twitter and our very online social life, the hype behind Umamusume is amplifying those voices. The Japan Racing Association has partnered with the game/Cygames to also promote horse racing, as well as the game itself, via its traditional and online channels. On that note, April has 4 big races in Japan and the game is promoting it, albeit casually. Fans of Umamusume are obviously a lot more into these April races it than the companies are. I mean, now I know what they are and what they mean. All the games are live streamed on Youtube so they do make the trend on Twitter (nationally).

There is a very active theorycrafting scene for the Umamusume game largely because the math is opaque. Generally due to inability to really test, and be able to trust the results, different folks have approached reverse engineering and putting the results together. There have been some comprehensive testing on the effect of raw stats on overall track time in A/B manner, with some standard deviation you can put on. There has been enough reverse engineering to know that the raising scenario races have different parameters than races outside of it, and this is besides souped up opponents. The cause-and-effect of procs, kakari, and other buff/debuffs IRT the wisdom/int/brains stat is known. The track selection distribution (for any given length type) is more or less known, as they are set and data-minable. Oh yeah, this is largely data-minable except obviously the server side stuff, which includes the core racing logic.

Which is to say, many of these factors are only known because the game exposes them to us in the racing visualization. I am probably extreme in saying it, but given how racing results is returned in the press of a button that takes a hundred milliseconds or whatever, when you skip the race, you wonder what the racing module extrapolates from winner and losers. This is the “visualization” camp of thought.

Many people take the racing much more literally. It is probably safe to clock a race. It’s a hard data point. Some go further and break down a race (say, a 2000m race into 20m/s rate) using frames or pixels. It’s more important to say these are all just guesses. There are evidence why these guesses are compelling though, so they are fairly educated guesses. I still don’t trust them since, as they will all say, they are incomplete. It’s one thing to trust randos on the internet, it’s another to trust randos that trust visualizations.

With that said, the data modeling and sampling, both anecdotal and harvested from Twitter, sites, and other forums, tend to converge in some possible truths. Raw data, when collected properly, is always good. For example, while you can have too much stamina, you really can’t have too much stamina (or rather, too much stamina means you don’t have enough speed). Nobody ever doubt the value of brains, but nobody really knows for sure how guts work. Power will help you accelerate and break past crowds, and speed is fast.

Somewhat controversial (but all evidence points towards) is the notion of a fuel economy where the horse’s racing energy depletes faster at faster speed. It may also pertain to certain racing strategy. Youtubers can be convincing, I guess. There are some multi-variable equations to be optimized here. I’ll link to more stuff in the postscripts.

In corollary or similarly speaking, this means in PVP races there is a rock-paper-scissor effect where horses with high brains can proc a lot of skills, but horses with higher speed and stamina actually win the race, and scoring accounts for both skills procced and winning. In addition there are attack-based skills that drains opponents of their energy that come into the mix.

Personally, I play this game for the fun of it. The raising mode is fun, but it’s because it is multi-dimensional. It’s a bit of a story mode where you and your horse go through various check points, racing successes and failures, and deal with what the story dish out with the goal to clear URA. You could be emotionally invested in getting that classic triple crown. But the meta goals are always there. You might be raising a horse for progeny purposes. You might want to raise a horse for PVP. You might want to raise a horse for a legend race. You might be trying for some achievements. You might be just farming fans or money. There are many ways to play the raising mode, and depends on what you’re doing you may be running the game quite differently.

This is a very different approach for me compared to my IDOLM@STER games, which are more like content delivery platforms. I use a game like Theater Days to engage the franchise daily, to receive new information, and to interact with the music, the characters, and the stories within. In Umamusume, I’m literally playing a game where the winner has the biggest numbers after some RNG and choices I make. There is a performance mode where you can customize the characters dancing and singing, but I don’t even touch it. Somehow, this is really fun and engaging even so.

I thought about the whole stage aspect of Umamusume a lot since season one of the anime. I think it is a mistake to say it is a mistake to include this, similar to how it’s a mistake to make all the horses women (?). I think it’s easy to say they are channeling existing frameworks and tropes to tap into an known quality for their potential base of players. But I think it is also brilliant in that they are bringing in something new to the table. The only real drawbacks are people who are thinking this is just another idol game or people who wants to bait people who are woke or something stupid. Umamusume is actually the most woke form of post-modern Japanese anthropomorphism. It is both really basic but at the same time really advanced. Because ultimately the spirit of the series glorifies horses, not the characters that are now bearing the horses’ names. It is literally a vehicle, and it is the most honest, and most earnest version of it. It isn’t turning a war machine into a pretty lady. It’s exposing the charms of majestic animals through their human-ish counterparts. Having them sing and dance is icing on top, but also apt commentary in this kind of a meta narrative about, well, what are idols.

The season 2 episode related to Tokai Teio on stage (a minor spoiler) is really the most brilliant idol anime concept ever, but without spoiling any more I will just say that I’m glad it is there, and it worked wondrously to create one of the best scenes I’ve seen in any idol anime. Umamusume isn’t even an idol anime, so it was all the more amusing.

After playing a lot of it since launch, Umamusume continues to be fun, although the grind can be a pain. By pain I really mean it takes a while to do a raising run–upwards of 45 minutes personally. And you kinda want to do at least 5 per day. That adds up. Umamusume game’s sudden popularity also is seen in other games–I can say personally that IDOLM@STER games are hit by it, and even Cygames’s own games as well. Those games are taking some effort to make the play easier to accommodate by giving us more QOL and shortcuts (but they should do this anyways). Priconne recently added the dungeon skip feature, which makes basically the game non-grindy at all other than the new content drops that show up a few times a month. Theater Days allow you to do 2 lives and jobs at once even during the intermission “events.” In order to have staying power, the daily grind has to be able to fit in the lives of busy people, I think.

PS1. I laced pics (more like tweets embedded) because, I don’t know, Umamusume does not have a space between Uma and Musume. Whenever I see a space I feel like the thought is interrupted. If Cygames cares enough to remove the strokes from the horse kanji, you ought to be caring enough to spell it correctly. I get why there used to be a space when people didn’t know any better, but not anymore.

PS2. Umagenics from Tin. Which is probably an end point rather than a start point, but there are links in the docs to take you upsource. If you want beginner guides, I think mine still is largely on target.


Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.