Library of Trantor

The plan needs to go on.

Welcome, wanderer of the Internet. You have stumbled upon the scribbles of one commonly referred to by fellows as shimmy1996, or rather, my blog. "Shimmy Xu" is just my mundane name.

Aside from HTTPS, you can also find this blog on the following alternate planes of existence:

ProtocolAddress
AtomFor timeline, posts, hoots, and comments
Gitgit://git.shimmy1996.com/blog.git
IPFS/ipns/shimmy1996.com or ipns://shimmy1996.com

Feel free to leave a comment or shoot me an email directly! Here's my PGP key and fingerprint:

5672 AC27 2669 A07A BD28 0896 ACC6 C791 312C F84D

If you are not in a hurry, consider paying a visit to my friends as well. If you are feeling extra adventurous, however, come surf along the IndieWeb webring: 🕸💍 !

All contents on this site are licensed under a Creative Commons Attribution-ShareAlike 4.0 International License (CC BY-SA 4.0).

TIReD: A Personal Rating System

As the pandemic gives me a chance to look through my backlog of movies, shows, and books (read: anime and manga), I started to consider establishing a personal rating system to ease up writing (hypothetical) reviews.

Guiding Principles

Typical rating scales feature 10 or more levels, which is in my opinion way too wide a range to choose from, not to mention those featuring a 100-point-scales. Even the most common 5-star system gets cumbersome fast as soon as we take half-stars into consideration. What exactly differentiates a 6 from a 7 or a 4.6 from a 5.1? Higher granularity could be useful in aggregated ratings, but not so much from an individual reviewer's perspective. I much prefer the approach s1vote took: give the users fewer but more distinctive levels to pick from.

My anecdotal evidences show that most online ratings converge around the 70% mark, a rating just as safe and useless as predicting a 40% success rate for anything. In other words, the lower half of most rating scales are underutilized: how often would you rate something one-and-a-half-star instead of just one? Besides, more often than not, I read ratings and reviews to find out about good shows, not the bad ones. It should be sufficient to only focus on "the better half": why would I sit through the entirety of a bad show and take the effort to give it a rating anyways? There is no -1 star in Michelin Guide, is there?

Summarizing the quality of anything with a single metric seems unfair. I want the rating system to be more expressive, capable of conveying the different aspects of a show that I find enjoyable. At the very minimum, an opinionated pick should be distinct from something with a more general appeal.

Rating Methodology

Enter the TIReD scale! The following uses anime/tv shows as the example here, but much of this methodology also applies to other art forms. A show is scored in the following categories, with sum of points forming the final rating:

CategoryRange
Tangible0-2
Intangible0-2
Revisit-ability0-1
Discretionary0-1

Tangible aspects of a show include visual style, animation, soundtrack, CG quality, special effect, etc. To put it simply, how physically well-made a show is. Starting from a score of 0, a show would be scored a

  • +1 if the show is overall attractive to watch and either has consistent high quality with very few shortcoming (perfection) or utilizes unique ideas/techniques to great effects (ingenious);
  • +2 if its physical quality/way of expression alone would be sufficient reason to watch the show, even if it gets a 0 in all other categories.

Intangible aspects include story, character building, plot pacing, cultural reference, etc. This quality should be relatively medium independent, i.e. I would enjoy a faithful recreation of the story in other art forms at least just as much. Criteria for scoring is similar except for remakes/adaptations with an clear intent to follow the original and when I have seen/read the source material: scoring would be based on the source material's intangible score adjusted downwards by 1 point, with at most extra 1 point adjustment based on quality/difficulty/effect of the remake/adaptation with in the range of 0-2. For instance, a mediocre retelling of a +2 story should only be awarded at most a +1. Remakes and adaptations probably have an easier starting point than original contents, so I wanted to adjust for "how good the show could have been", provide an answer to "should I still see this if I've seen the original", and pick out the "watch this instead of the original" or "transcended and elevated the original story" shows.

Revisit-ability, as the name indicates, represents whether I would want to revisit/rewatch the show later. This correlates more with my own taste or nostalgia: is this something that I would gladly jump into in an leisure afternoon. Longer shows tend to suffer a bit by this metric, so I would take into account of especially memorable segments/episodes. However, in event of remakes and adaptations, this point should generally only be rewarded to the best version of the work in my point of view.

Discretionary point should be awarded sparingly and only when a show doesn't already achieve full scores in all other categories, making the possible maximum score 5 instead of 6. This is used as an adjustment for shows that I feel the current rating system doesn't do it justice. Common situations where this applies include but are not limited to:

  • categorical superiority: best of its kind;
  • a tight coupling between tangible and intangible aspects of the work: it simply won't be the same without one another;
  • quality in spite of objective limitations, especially for older shows or those with a tight budget.

Format

A TIReD rating is recorded as X=T/I/Re[+D]. For instance:

  • a show scoring 1 in tangible, 2 in intangible, 0 in revisit-ability, and 0 in discretionary would be recorded as 3=1/2/0;
  • a show scoring 1 in tangible, 0 in intangible, 0 in revisit-ability, and 1 in discretionary would be recorded as 2=1/0/0+1.

Shows that I abandoned halfway, meaning I won't be able to give a rating, will be marked as DNF (did not finish).

Self Q&A

Some fragments of thoughts that I came across when designing TIReD.

Q: How should tangible points for books be awarded?

A: I'd say it's how good the writing is at face value, i.e. is it "literature" worthy. While I not really confident in my ability of identifying great works, but a +2 should at least be something better than Harry Potter.

Q: How should world settings built up in previous/related works affect the rating?

A: World building actually fits into both revisit-ability (if the system/world is interesting and makes me want to read more about it) and intangible quality (whether the character actions are justified).

Q: How was the rule for discretionary point determined?

A: The best shows should always get full score regardless of the exact scale, so awarding them discretionary points is meaningless. However, there are seemingly not-so-impressive works that really show the passion/devotion/love/good faith of the production team/author and shows whose existence alone is a boon for its fans. I want to express my enjoyment in a way that still allows me to assess the tangible and intangible aspects of a show on an absolute scale, as any further complication can be taken account of as discretionary point.

Q: What happens to ratings for a remake before and after you watch the original?

A: I'll adjust score for the remake now that I have experienced the original.

Q: A lot of details could be lost in translation. How to deal with translated works?

A: For now I will treat these the same way as remakes: adjust the rating if someday I came across the original.

Q: How did you come up with the name "TIReD" (and name for the categories)?

A: The first category to have a concrete name is revisit-ability. From there on it's mostly just playing around with words and initials. I almost settled on "TIRD" thanks to Urban Dictionary. Well, not everything is sh*t. 😜

Took a detour on the way home today for an hour-long ride on my bike. I was able to cover so much more ground (quite literally) than if I were to spend the same time running. I finally know what's lies beyond my normal running route now. Judging by the fact that most pain is coming from my neck instead of my legs now, I guess I've grew more used to riding.

I finally pulled the trigger on Suunto 9 during the sale a few weeks back. It took a while to arrive and was actually lighter than what I expected when comparing the specs against my Pebble Time Round. With news about the Garmin hack going on, can't say that I didn't feel smug about my choice. One thing I worried about Garmin (Suunto 7 as well), is that they are too smartwatch-like. No, not the Pebble-era of smartwatch-ness, but the Apple-Watch-era of smartwatch-ness, a.k.a. the bad kind that strangles you in an eco system. What I wanted was a dedicated device that is the running equivalent of a bike computer and I think Suunto 9 fits the bill better than most other choices. Pebble will remain my choice for daily routines and notifications. I still haven't logged any runs on the watch due to COVID though, but I did tracked the past few days of sleep on it - doesn't sleeping also count as a sport though?

I wish wiki pages have tagged releases so that I don't have to get all the editing done in one go.

Somehow the comic version of Hototo in Appare-Rannman's transition cuts looks like Nagase-san.

Lunch on the day of workplace satisfaction survey felt particularly delicious: flat iron steak laced with just the right amount of fat and smashed potatos baked to perfection. Yum! Even the box of fruit had the half-translucent-almost-jello-like kind of sweet melon - the kind you would see Suneo enjoying in a hot Summer afternoon, fresh out of the fridge.

Interpreting the ultra in ultramarathon as that in Ultraman immediately renders the sport and related events much more futuristic and appealing.

Tried out the Kensington SlimBlade trackball over the weekends and I liked it a lot.

My index finger got sore from excessive scroll-wheeling and mouse-clicking recently and I shopped around for more ergonomic alternatives. It's a shame that pointing devices as futuristic as trackballs didn't gain more traction: I blame the prevalence of flat design. To be fair, not all of the trackball designs seem that confortable to use. I didn't like any of the recent Logitech offerings (all thumb-operated/wireless trackballs). While it's nice to have open source offering in this space, Ploopy's scroll-wheel implementation seemed no better than an ordinary mouse. The SlimBlade seems to be one of the few trackballs has integrated scroll-wheel functionality as opposed to actually adding a scroll wheel. The CST/X-keys L-Trac was the only other contender with very well received free-spinning scroll wheel, but I preferred the size, design, and lower profile of the SlimBlade.

Back to SlimBlade: I did need to apply some Vaseline to get the ball to spin freely in its socket at the beginning, as the one I received does show some age. Instead of storing settings on-board, SlimBlade relies on it's OS-side driver software, which means on Linux we need to go through X-org/libinput for customization: not that it needed much at all though. The scroll-wheel function utilizes the vertical axis of the trackball and produces a very satisfying click when activated. The official demonstration features the scroll-wheel function with a holding-a-knob gesture; some user reviews I read indicated another common way is to hold trackball in place with one finger and pivot it using another; my preferred way is to rest one finger on the metallic ring outside the trackball and scroll away. Despite the SlimBlade's low profile, I find an old ErgoDox wrist rest to be just the perfect fit for an even more comfortable hold. I'm really happy with the experience so far - looking to replicate this at work if I feel tangible long term improvement in comfort.