KO percentage charts

[RexMaster]

I propose a useful addition to potential Character FAQ's: KO percentage charts
Every time I take up a new character, I spend some time making up a chart in training to find out the 'absolute KO point' of any tilt, grab, aerial, or smash on a non-DIing human opponent, in training.
For most, the knowledge of 'what move kills, when' is quite natural, however it could be useful to have some specifics on when different moves guarantee the win. Similarly, moves that require extremely high-%'s to KO are great indicators of potential juggles.
These sorts of lists obviously wouldn't be of much use to a "professional", who's done it all and seen it all, but I still find them useful from time to time.

For example, Plairnkk's Guide to playing Sheik! has a nice section denoting when Sheik's Uair KO's certain characters:
Bowser: 136%
Kirby: 97%
Mario: 119%
Pikachu: 112%
Peach: 107%
Ness: 118%
Samus: 113%

etc.
However, all my charts use Mario as the benchmark, since he is the mathematically-average weighted character. If we could find out rough #'s to both other extremes, that'd be great too.
Here is what I propose a formal list to look like:

Falco (again, Vs. a Mario)
Neut A
550
Dtilt
115
Utilt
160
Ftilt
200
Dsmash
125
Usmash
130
Fsmash
85
Dash Attack
?
Nair
145
Fair
195
Bair
120
Dair
225
Uair
165
Fthrow
250
Uthrow
390
Bthrow
420
Dthrow
---
Shine: 250
(note: May not be applicable. we don't expect falco's Dthrow to KO etc, applies to all throws)

I would like to make a full compilation of lists like these, so here are my Scientific Methodologies, so that they may be replicated.
-Mode: Training
-CPU: Human ('cept no one actually controlling)
-DI: none
-Stage: Final Destination
-Position: the Spawn point immediately after KO'ing the CPU (far left of FD)
-Direction: to the left (towards nearest 'wall')
-CPU should not be nudged from the position it spawns from, and should remain standing. if nudged or not KO'd when testing, KO and wait for spawn.
-KO percentage is judged from the % BEFORE being hit, make sure it's replicable.
-A KO is determined at the lowest possible percentage that a single hit sends the CPU straight out the side/top, avec fancy explosion effects. If you set it to standing CPU and it uses it's 2nd jump, the % is too low

B-moves can be optionally included, because we all know how 'useful' some of them are (looking at you, Sheik)

More charts will be posted later.

DISCUSS.

EDIT: NO DISCUSSION. TOOMAI IS KING. And has done a lot of work already. Woo~!

 

[Toomai]

Oh I am way ahead of you on this one. Just gimme a few minutes to collect my findings.

EDIT: Okay, here're my findings so far. I wanted to release them all at once, but you started this topic, so I thought "hey, I'll release them now!". Here are my parameters:

• Traning mode with 2P as human.
• Stage: Final Destination.
• For horizontal KOs, 2P rolls towards edge before hit.
• For vertical KOs, 2P stands anywhere.
• I used Mario for all testing (slight floatiness counteracts FD's high ceiling).
• Always used sweetspot of attack.
• Standing attack counted for first hit that actually does some distance.
• Uair was used at peak of Mario's first jump.

Now for the table I've collected:

Note: n/a means the move doesn't exist, "-" means the move needs more than 999% damage to KO.

 

[Zodiac]

This should be good.

 

[dguy6789]

I am looking forward to looking at some of these statistics. One question though. How will you account for DI in your findings? I think it would be better to include DI in your findings as that will make the benchmarks more valid, as any real match should have at least some degree of DI. However, it would also be interesting to include both DI and non DI. See how much DI helps each character against each move.

 

[RexMaster]

Good Lord, Major points for the community of replicable SCIENCE!
In my opinion, Non-DI would be correct for charts like this. It would be nice to know DI, but I also want to know the percentages that will more-or-less just KO somebody. DI and recovery can get you back from amazing distances, and would just be way to complex to calculate.
However, I agree that including DI would make it more "Valid" in terms of real gameplay, but it's simply wouldn't be practicable or reliable.

One Point of minor disagreement with methods for the fantastic table you have set up:
-Why make mario jump before testing Uair's? Are there any that you *cannot* hit while the CPU is standing?
This isn't a huge issue, however, as long as it's consistant.

Also, it'd be much appreciated if you explained what the first four parameters mean. (ie: "walk")
I can't wait until you're able to release the whole file! Also: NICE TIMING.

(how long have you been working on this, anyways??)

 

[petre]

i think walk is f-tilt, rise is u-tilt, and crouch is d-tilt. you should rename them to tilts so people dont get counfused (at first i was like hah he missed u-tilt, because i thought rise was a standing up attack after being knocked down...). but i agree you should leave the compy on the ground for doing uairs, push them to the edge and do a ledgehopped uair if you cant make it hit otherwise...i dont think it would be too hard with anyone.

question: whats with all the different shades of color? if youre just trying to seperate the boxes, id just stick with two alternating shades of each color, instead so many different ones. for some reason i think the shades have a purpose though...

 

[Toomai]

Okay, some answers to questions about my table:

1. I started this before I called them tilts, and in my rush to put up the results I forgot to change them.

2. Originally, this project was to see if a character's KO ratings matched their tier. I set up colours so that lighter colours are power moves, while darker colours are moves that just don't KO well. (The blank columns are light because the computer sees blank cells as 0.) I divided the colours into 4 categories: <75%, 75%-125%, 125%-200%, and 300%+.

3. I made Mario jump for the uairs to add that extra amount of consistency, plus there are some uairs that just don't hit on the ground (Yoshi's comes to mind).

Give me a day or two and I'll change the table with better names and less confusing colours.

By they way, I've seen some interesting results in my data already: Link can't KO anyone under 89% and Ganondorf has the most <75% moves.

 

[RexMaster]

Toomai, this is just fantastic stuff. I can't wait to see the updated version.

Is there any chance that you will eventually distribute the main file? I'm kinda curious to see what kind of statistical manipulations I can do to that stuff in SPSS.

I like the basic idea of color-coding, bringing out the different ranges with different shades. Great way to make everything stand out as it should- Good KO moves at one end of the spectrum, and good potential juggles at the other.

As for the intervals of those shades, I have one recommendation- bring the last group from 300%+ to 175%+, perhaps. This is how I envision it, with an extra category perhaps:

<75%, 75%-125%, 125%-175%, 175%-300%, and 300%+

I recommend that from a utility standpoint- it still helps identify the 'non-killers', but also makes the 125-175 interval equal (50%), and honestly, I don't think in any serious match that ANYBODY should ever get above 175%. Just my two cents, but if you have to re-code it or do it manually, I wouldn't worry too much about it.
Also: Bowser Charged-Fsmash KO @ 55% FTW!

One more thing: Ok, so, who else think it'd be neat to also include actual get-up attacks, ie: pressing A while laying Prone front/back after tumbling into the ground. I'll bet Bowser, and Marth have the strongest! (no actual verification, I'm just guessing. RANDOMLY)

 

[Toomai]

Updated version: Click here for pic, it's quite big

Changes made:

1. More stats: averages, 25% of char's moves are in this category, etc.
2. Less colours: I still have the original version, but I think it should be saved until the end. I bolded moves in the <75% category.

 

[RexMaster]

Since you have some missing, Here's the data I've collected previously. Though I use a slightly different system, you can use these as benchmarks for figuring out the #'s according to your ways. They're also a bit rough, but it'll make your end job easier. I've already posted Falco, so here's Fox and Bowser

Fox (again, Vs. a Mario)
Neut A

Dtilt
150
Utilt
115
Ftilt
200
Dsmash
170
Usmash
89
Fsmash
115
Dash Attack
?
Nair
150
Fair
240
Bair
125
Dair
-
Uair
115
Fthrow
?
Uthrow
?
Bthrow
?
Dthrow
?

Bowser
Neut A
210
Dtilt
135
Utilt
105
Ftilt
91
Dsmash
100
Usmash
95
Fsmash
80?
Dash Attack
?
Nair
170
Fair
140
Bair
145
Dair
-
Uair
90
Fthrow
?
Uthrow
?
Bthrow
?
Dthrow
?

I can't wait to see how Koopa's charged smashes look. MM-mmm!

 

[Toomai]

New version with Dr. Mario and Bowser added: This link!

 

[RexMaster]

Just a question: can you explain in detail what you count as a KO? I understand that you roll towards the edge before using side-KO'ing attacks, what at what point is it an official KO? When they exit the screen without being able to 2nd jump? When a computer can't recover? Perhaps when a HUMAN can't recover?

Wooo~, Bowser! Most L2's!

 

[Toomai]

For horizontal trajectories, a KO is when the no-input Mario dummy hits the side blast line instead of the one at the bottom.

 

[Adi]

I'm making a marth one right now, I'll post it here when I'm done (it'll include tippers ;D)

 

[RexMaster]

Marth, eh? I wanna see how CF stacks up compared to the rest!
As a sort of an aside, though these %'s are all artificially low compared to actual gameplay, they are pretty much all accurate if you imagine the opponent having used their 2nd jump already, with a few exceptions (multi-jumps, Samus, Etc). Hooray Validity!

 

[Banks]

holy crap toomai nice graph, must have took you about forever.

 

[Doraki]

ok I have to say a few things, if people want to get the most of these things. dunno if it was said or not but whatever.

Always suicide between each try, because if you don't, the move will weaken.
I personally find it better to indicate the % after the attack also for this reason, because the weakening only lower the % damage of the attack and the knockback you have depends on your damage after the hit.


* knockback formula : the speed at which you're sent by a hit is always

S0 + S1 * (your % damage)*200/(100 + Weight) where S0 and S1 are 2 constants depending on the move ; Weight is the weight of your character as measured with the MKI scale.
It doesn't depend on DI.
exceptions are throws, where the weight doesn't matter, and fox's shine, as far as I know.
I didn't study what happen when you bounce off the ground when being hit by a spike, nor the effects of CCing.

* Scream %age : Your character will do 2 different screams when he's sent at certain speeds. Those 2 speeds are the same for every character.

With this you can compare the strengths of every move in the game and somehow compute S0 and S1 of every move. (I did it for most of fox and falco's moves)
I don't remember the 2 speeds corresponding to those right now.
Also I'd like to measure the starting angles of the trajectories but I don't have a very reliable way to do that.

* Teching speed : if you're sent faster than this speed you can tech.
Approximately 42 feet/sec.

* Ground speed : if the vertical component of your speed is less than that,
you can ASDI down and stay grounded. It's (1/2. Smash DI distance)/frame, so (I think) 60 feet/sec (1 Smash DI = 2 feet)

The speeds are calculated in the game in foot/sec or meter/sec depending on the language you selected.
(see top speed and fastest pitch at the end of a match).

* Nana : she has a defense ratio and screws up everything.

* Upwards KOs : your character his accelerating down during his trajectory, this acceleration is dependant of the character.
I saw some weird things with fox's upthrow KO %ages on different stages, there might be some other factors.

* Side KOs : They depend on both the stun time and the speed the character is sent at.
I know that the heavier you are, the less stun time you have, but I don't know if it follows the same kind of formula the knockback speed has.
And again for throws, stun doesn't depend on weight.

If you're jigglypuff, your horizontal speed will instantly stop after the stun ends, so the only way you can be KO'd by the side is when you're still stunned, unlike the other chars, who can try to double jump just before dying (but will die anyway).

Maybe with all of this, with a lot of work we could measure the boundaries of every stage and compute every KO %age.

 

[RexMaster]

Alright, to reiterate: WOW, that is some intense possibilities you have there. Congrats to the community for fostering that kind of refinement.

That being said, I am still interested in the completed tables first of all. Has there been any progress at all, or can we help with any characters in particular?

 

[t0m0]

thanks for putting in the time to create this thing.

 

[Toomai]

Has there been any progress at all, or can we help with any characters in particular?

I've been busy a lot, so there's been no real progress. You can go ahead and do some testing if you want, but I'll only add the results to my data if they're under the same conditions.

Oh yeah, and anyone have any ideas on Marth and Roy? Do we need to do them twice (sweetspot/non-sweetspot) or what?

 

[RexMaster]

I would say that you should always sweetspot. In every other circumstance, the rule has been to measure the %age at KO "on First or Strongest hit". Therefore, the "best" hit is a sweetspot in every case.

I'll let you know if I can get any basic templates of %'ages to help your progress

~Rex