Header Menu

Monday, August 20, 2018

Feature: Combo Scoring Visualized

We are returning to the basics with Feature posts like this, for our newer audience hailing from the glorious return of Taiko no Tatsujin to the western circles. If you want us to go deeper on certain aspects that we took the knowledge for granted, let us know!


Recently there is a surgence of social media posts in which newer players question "why does the other player have a higher score, when I have a better Good/OK/Bad ratio and drumroll counts", especially with PS4 Session de Dodon ga Don/Drum Session where ranked battles are a signature mode.

To me who has been with Taiko no Tatsujin for over 8 years, the concept of Combo Scoring may be the obvious answer, but not to a lot of the said new players. While you can definitely read up on our Beginners Guide (in desperate need for update, even though) for the information, what better way is there to visualize the concepts with the power of graphs? MATHS!

Combo Scoring, a Recap
This "Combo Scoring" system basically boils down to the higher you can maintain a combo, the more points you can get. In the 2nd generation the score-per-note increases once every 10 combo up til 100 combo, while in the 3rd generation (AC0 and 3DS1 onwards) it increases only 4 times at 10, 30, 50 and 100 combo but at varying rates (1x, 2x, 4x and 8x the increment respectively). Because of how score-per-note used to work like arithmetic sequences, Donders has since used actual maths terms "first term" (初項) and "common difference" (公差) to refer to the characteristics of the scoring scheme. It has been found that in most recent scoring schemes first terms and common differences have been controlled to be in an approximate 4:1 ratio.

Let take Densetsu no Matsuri (伝説の祭り) for an example:
  • In Wii4 (2nd gen.), the score-per-note starts from 370 and then increases to 450, 530, 610 and so on, up to 1170 points (=370+80×10) per note after 100 combo. The first term is 370 and the common difference is 80.
  • In AC0 (3rd gen.), the score-per-note starts from 380 and then increases to 470 (=380+90×1) after 10 combo, 560 (=380+90×2) after 30 combo, 740 (=380+90×4) after 50 combo and up to 1100 (=380+90×8) after 100 combo. The first term is 380 and the common difference is 90.
You might have noticed that all scores in Taiko no Tatsujin are multiples of ten, but in the 3rd gen. there can exist common differences with odd units digit, in which case they will be rounded down to the last 10. VERTeX, for example, has the scores-per-note of 430/530/640/860/1290 in AC0, and the common difference is worked out to be 108.

If you don't want to have Combo Scoring and want the score to focus more on overall accuracy (like in tournament settings), consider switching to the Shin-uchi (Mode) which gives a flat rate in score-per-note regardless of combo and also removes certain bonuses. See our separate article for more details.

The Visualization
Having laid the basics out again, let's use graphs to better showcase the essence. For now we'll just focus on the Combo Scoring and neglect all other drumroll notes and bonuses like Large Notes or Go-go Time.

For illustration, let us plot all the results of each note (minus bonuses) as strips of width 1, so that the area of each strip is the points you get for that note:

Spanning that across the whole song and you should get a graph whose area under the curve is the total score you get from normal notes (again, minus bonuses) like this: (the steps before 100 combo is smoothed out for easier discussions later)

Back to the Question
With that visualization schema in place, we can now answer the question that started off the post - A main reason why better Good/OK/Bad ratios and drumroll counts sometimes still yield less score is the location of those OKs and Bads.

Under combo scoring schemes, every Bad you get will break the combo and reset the score-per-note rates to the initial value. That removes quite a big chunk from the total score you get, as you can see from the graph:

Imagine then if you would, that you missed 3 notes in 3 separate places, and someone else missed 10 notes at one place (say, they just gave up on that one deathstream). The 3 separate "chunks" you lost can still be greater than the one bigger "chunk" they lost, leaving you with a lower total score:


This "chunk theory" also opens up explanations for other questions that you might (or might not) have. Say if you have to miss one note no matter what and (for some reason) you can control where it happens, try to put it as early or as late into the song as possible to minimize the "chunk" lost. Tell us how this representation helped you in your Taiko life, and we might try to help you make the graphs to prove it.

Of course, other reasons also exists that will affect your score, like whether you have your OKs/Bads in or out of Go-go Time (flames on the marker, giving 20% more points overall), and whether your drumroll hits go to finishing your balloons and Kusudama/Kozuchi notes (which give additional points).