Tech - Shinnok Some Mathematical Analysis I did with Shinnok...

[STB Damashi]

So basically, I went into Kustom Kombat mode with two controllers, and then enabled Juggle Kombat in order to juggle the opponent for far longer than usual. I did this in order to repeat the same move over and over again, then I took a look at the damage info in the Match Replay section of the game. I looked at his B3 (in the corner) up until 25 hits, his 4 (I made the opponent Neutral Jump, then pressed 4 to keep him in the air), and his 1st Hell Spark.

B3: y = 5.5531e^(-0.105x) , x is greater than or equal to 1

4: y = 7.3711e^(-0.051x) , x is greater than or equal to 1

1st Hellspark: y = 3.5395e^(-0.163x) , x is greater than or equal to 1

With a graph of the three different moves that I gathered the information on, I was now able to extrapolate the damage per hit of each attack by plugging in the Hit number into x, and then I was able to predict the damage of whatever hit (for example hit #28 or whatnot). I still can't predict the damage though if I combine different attacks, like instead of the string B3, B3, B3, what if I did the string (in the corner), B3, 4~Hell Spark? Do any of you guys know how to predict the damage of each hit in a combo combining different types of attack?

I went into practice mode to checkout the damage of each attack if I combine them into the string B3, 4, Hellspark, and I got 5%, 6.3%, and 2.56% with the string. If the damage was not scaled, the damage for each attack would be 5%, 7%, and 3%, but the attacks are affected by the previous attack and I am unable to predict the damage. I would really like to discover a way to maximize combo damage but I just can't seem to find out what to do with the equations.

Edit: Here are some graphs with the values that I got off the damage I looked at - http://imgur.com/a/XnPgZ

Here is a video of how I got my results

Edit: Here are some more formulas -

D1: y = 2.108e^(-0.051x) http://imgur.com/qhzqj3u
B1: y = 8.6687e^(-0.222) http://imgur.com/qNBR3NP for all natural integers (hits) greater than 22, damage = 0.07%

 

[MrSunshine]

What

 

[Derptile]

Location: Compton.

 

[STB Damashi]

What

Basically, found an equations that match the damage of the string B3, B3, B3, B3..., 4, 4, 4, 4, 4..., and 1st Hell Spark, 1st Hell Spark, 1st Hell Spark, 1st Hell Spark... I am trying to maximize my damage output though and I can't figure it out.

 

[errormacro7]

I hate math.

 

[Derwood1979]

So basically, I went into Kustom Kombat mode with two controllers, and then enabled Juggle Kombat in order to juggle the opponent for far longer than usual. I did this in order to repeat the same move over and over again, then I took a look at the damage info in the Match Replay section of the game. I looked at his B3 (in the corner) up until 25 hits, his 4 (I made the opponent Neutral Jump, then pressed 4 to keep him in the air), and his 1st Hell Spark.

B3: y = 5.5531e-0.105x

4: y = 7.3711e-0.051x

1st Hellspark: y = 3.5395e-0.163x

With a graph of the three different moves that I gathered the information on, I was now able to extrapolate the damage per hit of each attack by plugging in the Hit number into x, and then I was able to predict the damage of whatever hit (for example hit #28 or whatnot). I still can't predict the damage though if I combine different attacks, like instead of the string B3, B3, B3, what if I did the string (in the corner), B3, 4~Hell Spark? Do any of you guys know how to predict the damage of each hit in a combo combining different types of attack?

I went into practice mode to checkout the damage of each attack if I combine them into the string B3, 4, Hellspark, and I got 5%, 6.3%, and 2.56% with the string. If the damage was not scaled, the damage for each attack would be 5%, 7%, and 3%, but the attacks are affected by the previous attack and I am unable to predict the damage. I would really like to discover a way to maximize combo damage but I just can't seem to find out what to do with the equations.

Ow my head

 

[GNG Iniquity]

Basically, found an equations that match the damage of the string B3, B3, B3, B3..., 4, 4, 4, 4, 4..., and 1st Hell Spark, 1st Hell Spark, 1st Hell Spark, 1st Hell Spark... I am trying to maximize my damage output though and I can't figure it out.

...You should try the combo threads that other people post?

 

[Biggs]

If the damage scailing isnt as simple as percentages you might have to use some calculus

 

[STB Damashi]

If the damage scailing isnt as simple as percentages you might have to use some calculus

I am taking Calc BC right now and I already know a bit of AB right now but I can't figure out what to do, sure we get word problems but I don't know what to do here with the equations

 

[True Grave]

Just do combos and whatever the highest number is, remember that. Then try to do every variant within that combo and see if you can increase the damage percentage. That's how you optimize combos to max damage.

 

[Biggs]

I am taking Calc BC right now and I already know a bit of AB right now but I can't figure out what to do, sure we get word problems but I don't know what to do here with the equations

You probably need more info on his other moves. Your models seem to be exponential decay... Maybe rewrite them as a natural log so you can take a derivative to find the slope. See if the slopes are similar

 

[DavS13]

There's no point to any of this, you can just attempt different normals in combos to see what can maximize your damage. Second, you have no proof the damage exponentially decays with the amount of hits of the same move so your equations are probably totally incorrect.

 

[Icy Black Deep]

This looks too complicated. Why such odd numbers (so many decimal places)? If the scaling is based on percentages, it should be in the form 10^x rather than e^x.

My curiosity is piqued though. Can you post your raw data? (And data for the mixed combos you can't figure out?)

 

[STB Damashi]

There's no point to any of this, you can just attempt different normals in combos to see what can maximize your damage. Second, you have no proof the damage exponentially decays with the amount of hits of the same move so your equations are probably totally incorrect.

Nah man, they are definitely correct, trust me. I would definitely not have posted this unless I was completely certain.

 

[Biggs]

This looks too complicated. Why such odd numbers (so many decimal places)? If the scaling is based on percentages, it should be in the form 10^x rather than e^x.

My curiosity is piqued though. Can you post your raw data? (And data for the mixed combos you can't figure out?)

He probably used a program such as tracker to come up with a model based on his data

 

[STB Damashi]

This looks too complicated. Why such odd numbers (so many decimal places)? If the scaling is based on percentages, it should be in the form 10^x rather than e^x.

My curiosity is piqued though. Can you post your raw data? (And data for the mixed combos you can't figure out?)

I'll definitely post them when I get home in a graph and all, I really want to dive more into this.

 

[marietta1200]

...jk math's awesome.

 

[Doombawkz]

Seems too hard.

It's easier to do a move and then follow up with something like b.3 and then multiply the number by 2 and divide it by 10.

For example, if you did b.3 b.3 and it came up as 5%, and then 4.50%, then you'd multiply that by 2 and divide it by 10.

That's 9%, then 0.9%.

So the move b.3 scales all other mov a after it by a factor of 10. I.e. If you did b.3 and d.2, d.2 would do 12.6% instead of 14%.

From there you can predict how much moves wil scale based on their scaling measurements added in the way that say...


B.3 b.3 b.3 would scale the move 10%, then an additional 9%. So the numbers should read say... 5%, 4.5%, 4.05% or something to that effect.

 

[Doombawkz]

I did this with ferra torr at one point, our uppercut scales 35%, and our d.1 scales 10%, where as our f.3 scales 5%.

Using that I was able to optimize our combos based on adding the larger scaling at the end and the smaller at the beginning. Also found out which moves are less efficient that their damage provides in benefit.

 

[STB Damashi]

This looks too complicated. Why such odd numbers (so many decimal places)? If the scaling is based on percentages, it should be in the form 10^x rather than e^x.

My curiosity is piqued though. Can you post your raw data? (And data for the mixed combos you can't figure out?)

Here are the graphs/x and y values I got during the sessions http://imgur.com/a/XnPgZ as for the B3, 4~Hellspark String though, I just got 5% with B3, then when I followed up with 4 I got 6.3%, and then when I canceled into a single Hell Spark hitting the opponent, I got 2.56%. I am trying to found out how to predict these types of numbers with the equations above though. I was hoping that the equations would be useful somehow, I haven't delved deep into this as much as I'd like to.

 

[Vigilante24]

Omg guys im just a music major! Stahp plz, im not ready.

 

[STB Damashi]

You probably need more info on his other moves. Your models seem to be exponential decay... Maybe rewrite them as a natural log so you can take a derivative to find the slope. See if the slopes are similar

Hypothetically speaking, lets say I do rewrite them as a natural log and find the derivative, what would I do then?

 

[Biggs]

Hypothetically speaking, lets say I do rewrite them as a natural log and find the derivative, what would I do then?

Look to see how similar the slopes of each model are. Maybe you will find out if it really is a percentage

 

[STB Damashi]

Look to see how similar the slopes of each model are. Maybe you will find out if it really is a percentage

Nice Toby pic btw, and Ah ok, I see what you mean. Hmm, I'll definitely see and try it out.

 

[Icy Black Deep]

So thinking a little more:
Assuming a percentage scaling so 1 hit = x, 2 hits = x + p*x, 3 hits = x + p*x + p*p*x, etc.
For n hits damage = x*sum[i=0..n-1](p^i) which is equal to x*(1-p^n)/(1-p)
x is the damage from a single hit, so that's easy. I'm not sure how you would go about trying to find p other than trial and error though.