UPSB v3
Tricks & Combos / [topic][1.26.1] Formal Analysis
-
Date: Tue, Jan 5 2010 21:48:58
Formal analysis is defined as a systematic analysis of a combo (video or breakdown) in order to extract certain absolute properties, or relative properties if compared to other combos.
The process is as follow:
INPUT -> ANALYSIS -> OUTPUT.
As input, we feed the analysis video(s) or breakdown(s). The analysis studies the input and provides an output, which is the results of the analysis.
Possible applications:
- By deducing properties in an objective manner, we are able to provide objective judging criteria based on the desirability of those properties. For instance, if an analysis is able to discern the relative difficulty between two combos, we can use this as a judging criteria. Spam is considered a negative element in combos, if an analysis can tell us the level of spam of a combo, then this can serve as a basis for assessing the value of a combo.
- By analyzing multiple combos, we can discover metrics to better understand and define notions. Example: What is smoothness? By analyzing combos most consider smooth, we can discover a formula or metric for evaluating smoothness.
- By associating certain properties with certain trends, we can exploit those trends to generate new combos. For instance, if we want to generate a very fluid combo and we know of several sequences that tend to be included in fluid combos, the generator will base itself on those sequences.
- Analyzing combos can prove or refute certain hypothesis. For instance, is it true that a fast combo has a higher rate of tricks per second?
- Certain analysis can provide interesting results which can inspire new hypothesis or research questions. For instance, suppose we conduct an analysis on the frequency of the use of each finger in a certain representative sample compared to another. If there is a great variance, how can we attempt to explain the phenomenon? (maybe style is related to the relative use of each finger)
Formal analysis can also concretely define certain impressions or opinions people feel when they watch combos that they cannot put into words.
Mechanization
Formal analysis does not necessarily require automation. They can be done by hand by reading breakdowns and keeping track of certain data. However, it will be more efficient if the analysis is automated by a computer program. To do so however, we need a formal grammar so that computers can understand the language of breakdown. This is a on-going project in the FPSB lab.
The next step
So what can we do for now? There are two types analysis we can do: on videos and on breakdowns.
- Collecting videos is a fairly simple task, but analyzing without a breakdown is fairly difficult because we can miss out a lot of details. Some of the analysis we can do concern length of the combo or some of its parts ("is there a relationship between combo length and ____?") and camera angle.
- Collecting breakdowns is fairly time intensive. We do have a breakdown team but it is inactive atm. However breakdowns provide a lot of information that we can analyze, such as use of fingerslots and fingers, tricks type, modifiers, etc. Here are some of the hypothesis we can test:
-> Is there a relationship between the number of hybrids and spam? Take a sample of combos considered spammy, and another sample of non-spam, with their breakdown, and evaluate the ratio of hybrids used in each sample.
-> What are the main trick families used by japanese compared to korean style spinners?
-> Is there a relation between direction changes and smoothness? Count the number of direction changes in combos that considered smooth and unsmooth.
-> Collect a metric called "modified deviation" which counts the number of modifiers in a breakdown (i.e. fingerless, reverse, twisted, etc.) Try to relate with properties. -
Date: Tue, Jan 5 2010 22:06:51
hmm, it seems that in order for this to be successful, before the analysis is possible, lots of intangibles have to try to be quantified. For example, you mention analyzing spam content. But first "spam" itself has to be objectified. Which leads to "easy" vs "hard" tricks. which is extremely subjective, and difficult to quantify.
One aspect of this I really like is the breakdown analysis. Maybe it'd be easier at that point to try to observe spam, because you look for repetition of tricks/trick families. If someone has say 15 passes in their combo, but they're spread out and used as linkages, it could be ignored, but if there are 15 passes in a combo, and 9 of which are within a 12 trick sequence, then it could safely be marked as "spammy"
Though in re-reading your "The next step" section, it sounds that instead of simply analyzing the combo and coming to conclusions of our own, you're suggesting that we instead look at people's reactions to combos, and then analyzing why people came to that conclusion.
This really interests me. If I'm understanding correctly, it sounds like you're saying do a long-term study. Look at released combos/videos, take not of the reactions the video gets, then analyze over the course of say a hundred videos, and see what's similar between all of the "smooth" videos, or between all the "spammy" videos, etc.
Very interesting (though IMO, wholly pointless really). I'd be in on working on this project -
Date: Tue, Jan 5 2010 22:17:58QUOTE (strat1227 @ Jan 5 2010, 05:06 PM) <{POST_SNAPBACK}>hmm, it seems that in order for this to be successful, before the analysis is possible, lots of intangibles have to try to be quantified. For example, you mention analyzing spam content. But first "spam" itself has to be objectified. Which leads to "easy" vs "hard" tricks. which is extremely subjective, and difficult to quantify.
Well at its core, analysis rests on subject principles. This is simply because the goal of pen spinning is to spin beautifully for humans. So the fondation of pen spinning is subjective.QUOTEOne aspect of this I really like is the breakdown analysis. Maybe it'd be easier at that point to try to observe spam, because you look for repetition of tricks/tr[[ick families. If someone has say 15 passes in their combo, but they're spread out and used as linkages, it could be ignored, but if there are 15 passes in a combo, and 9 of which are within a 12 trick sequence, then it could safely be marked as "spammy"
Though in re-reading your "The next step" section, it sounds that instead of simply analyzing the combo and coming to conclusions of our own, you're suggesting that we instead look at people's reactions to combos, and then analyzing why people came to that conclusion.
Yes, one application of analysis is to explain why people hold certain opinions when watching certain combos. As I wrote: "Formal analysis can also concretely define certain impressions or opinions people feel when they watch combos that they cannot put into words."
So yes most analysis are related to explaining opinions, but there are other usages. You can for example try to find relationships between objective properties, such as relating "number of different fingerslots used" with "maximum rotation of a single trick" (my hypothesis is that the more slots you use, the less likely you have a high rotation trick, but they could be unrelated). But those analysis are not really meaningful because they don't tell us anything about the perception of the combo. It doesn't tell guide us into making better combos or how to judge them or to classify combos into styles. -
Date: Tue, Jan 5 2010 22:21:12QUOTEwe need a formal grammar so that computers can understand the language of breakdown. This is a on-going project in the FPSB lab.
Meaning they have this program already? Or are working on creating one?
Ok, so what I describe in the latter part of my post would be like a subsection of this project? (Like 1.26.2 or something?) -
Date: Tue, Jan 5 2010 22:23:32
individual analysis would be 1.26.1.1,2,3,4, etc if they get too big we can make new project for them.
they dont have a problem but they have the stub of a grammar. I will post about it, not sure if it belongs here or in NC -
Date: Tue, Jan 5 2010 22:34:08
Hmm ok, well if we decide to open a project regarding what I was talking about, I'd definitely be interested in working on it.