Author Topic: Meter Statistics (Really really long post)  (Read 926 times)

Hobojim

  • New Challenger
  • *
  • Posts: 30
    • View Profile
Meter Statistics (Really really long post)
« on: November 17, 2015, 04:18:53 AM »
Alright, so this is going to be a long-winded and somewhat odd post, bear with me. It does have a point, and poses a question at the end, so go ahead and skip to there if you aren't interested in rambling points about spreadsheets and resource analysis.

When I was coming up with combos for some characters when I first started this game, I became somewhat obsessed with the efficiency of a combo with regards to meter usage. I did this because I was having fun sitting in training mode for hours on end trying to come up with the most damaging combo for any given amount of meter and drive, and also it was a low-stress way to work on execution that didn't require any winning or losing, etc. It was basically how I figured out how to play the game, and as those of you who play me regularly know, it was undoubtedly the WRONG way to figure out how to play the game.

But I digress.

When I was coming up with combos and how best to use resources with a particular character, I wanted an easy-to-use method of figuring out if a combo was "worth it" and gave a good amount of damage for it's resources. Being somewhat proficient with spreadsheets and statistical analysis (but far from an expert, more on that later) I came up with a simple formula to express the amount of damage a combo does in terms of resources used and measure it against a statistical baseline. I called the number that this formula produces the "efficiency factor" of the combo, and calculated it for every possible combination for meter and drive, corner and mid-screen, for my three characters. I used this number to determine which combos would be useful and which are a waste of meter. It also provides some insight as to how to best utilize a character, for example, Andy does a ton of extra damage when he's got someone in a corner, and has meterless options to get the opponent there. Kyo does more damage in the corner, but it's not as drastic a change, and Shen doesn't really care one way or another, he just does big man punches for big man damage.
Obviously this doesn't take in to account silly things like "normals," "mix-ups," and "pokes." I did mention this was the wrong way to learn, right?

Anyway, my "efficiency factor" (EF for short) is calculated as follows: Develop combos at each level of meter and drive for corner and mid-screen that start from the same normal. Mid-screen combos and corner combos are calculated separately, as there isn't much point in lumping them in with each other; it's almost universally better to do corner combos if possible and it would dilute the results from the mid-screen. In my case, I chose close C as the combo starter for my characters, which seemed logical. It might not be for other characters. No jump-ins, because you want a simple baseline that's common among characters and different characters have different usable jump-ins for combos, etc., and you want to keep changes to a minimum.

Your meterless, driveless combo will be your baseline. For each combo in that category (mid-screen or corner), you can take the difference in damage between that combo and the baseline. If this is done for each non-drive combo, you can calculate the amount of damage added per unit of stock by taking the "above baseline" value and dividing it by the amount of meter used. You can do this for drive as well, by comparing a combo that uses drive to the combo that uses the SAME amount of stock but NO drive. From these numbers, an average amount of damage added from stock and from drive for that character is simple to calculate.

Still with me? Probably not. I'd be surprised if anybody read all of this, honestly, but I'll keep going anyway.

Okay, so now you have the average amount of damage added to a combo per unit of stock and drive. The more data points you have, the better the average will represent your combos.

From our average numbers, we can calculate the "expected" amount of damage a combo will do by taking the baseline and adding the average numbers for stock and/or drive. To use Andy as an example: For the combos I use, mid-screen Andy does 237 damage baseline, adds 79.2 damage per stock and 66.0 damage per half-drive. If I wanted to know how efficient Andy's combo for 1 bar and half-drive was, I could look at the "expected" value based on these averages:

Expected Value = (Baseline) + (Average bar damage)*(# of bars used) + (Average half-drive damage)*(# drive used)

In our case, to plug in numbers:

Expected Value = 237 + (79.2*1) + (66.0*1) = 382.2 damage, or 382 after rounding.

As it turns out, my combo for Andy mid-screen with 1 bar and half-drive is exactly equal to 382. This means that it isn't a particularly efficient combo, but isn't inefficient either. The Efficiency Factor of this combo is 1.

EF = (Actual Combo Damage) / (Expected Combo Damage)

The higher the EF, the better off you are using that combo in terms of resource efficiency. To go back to Andy, mid-screen, his most efficient combo (of mine, anyway) is for 2 bars, no drive: Cl.C, f+A, hcf+BD, qcb,hcf+P. (You can get exactly 1 more damage by doing kuuhadan-break into EXDM, but I screw that one up more, so eh.)
The expected value here is 237 + (79.2*2) + (66.0*0) = 395 damage. The actual amount it does is 409. The efficiency factor, then is (409/395)=1.09.

This is still not a great number, but that's why Andy's so much better in the corner. His corner combos range from an EF of 1.02 on the low end to 1.38 on the high end. Shen's 1-stock no-drive combo has an EF of 1.93, which is the highest I've got so far (outside of HD, that is).

Looking at all of the numbers, I can see where characters are best and with what resources. Andy, on average, adds almost 100 extra damage for half-drive in the corner. That's a lot. Kyo adds just under 5 (Yes, 5) extra damage, on average, for half of the drive meter mid-screen. From this we can conclude not to use Kyo's or Andy's drive meter mid-screen if you're aiming for efficiency.

Universally, for the characters I've tested (All four of them, admittedly not a large sample size), combos are structured in a front-loaded way with respect to meter, by which I mean combos using 1 or 2 bars are more efficient than combos using more than that. That doesn't mean you shouldn't use heavier combos; a kill's a kill, and if you can forego another set-up in order to take them out now, losing out on conceptual efficiency is not at all important. You can also use these numbers to "decouple" drive and meter, that is, figure out the best combo to use if you don't care how efficient you are with drive but want to maintain some stock efficiency, etc., but that's probably not a useful metric.

It's important to note that HD combos were left out of all calculations, or calculated separately. They're kind of a statistical anomaly, in that they do a ridiculously large amount of damage compared to the non-HD combos, and EF values of 2-3 are not unheard of. I'm also bad at them, so it's harder to do calculations because it takes me a lot longer to get the data.  ;)

Okay. So that's WAY more words than I had intended to write about this possibly useless concept.



TL;DR - Unemployed guy with a wordy disposition and a background in science got bored and figured out what constitutes resource efficiency in combos.

My question to anybody still reading this; I'm new to fighting games in general, and resource management is a pretty foreign concept to me. This was my way of trying to make some sense of it. How do "good" (or at least experienced) players think about meter management? Is it just something that you glance at now and then and do whatever you want? Do they just use as much as meter as you can every time if it won't kill, and tone it back if it will? Do you plan ahead and go for specific combos/set-ups that use pre-planned amounts of meter? Save everything for HD combos? I'd love to know.

If anybody has questions about this process (as if I haven't typed enough about it) I'm willing to answer. Honestly I'd be impressed if anybody read the whole thing, but maybe somebody will get something out of it.

Thanks!