Frame by frame testing in umk3

[Remz]

Hi, I've spent some time testing different aspects of umk3 using frame advance in MAME and I found a lot of interresting things that I didn't know.

Maybe some of you already know everything I'll post here but here are some of the results of what I have tested so far.


HOW STARDARD MOVES AND FRAME ADVANTAGE WORKS:


Frame advantage and throws:

Every standard attack in the game only connect on a single frame which is always the same no matter who have the frame advantage or no matter the distance between the two opponents. The opponent only need to block that frame to prevent the move to connect. If you do a jab (which always connect and have to be blocked on frame #11) the blood or the animation that shows the dammage (energy bar) will start 3 frames later (frame # 14) for the player who have the frame advantage and 4 frames later (frame #15) for the other player. However, both player will recover on the same frame. When both players do an attack that connect on the same frame, only the player with the frame advantage will connect and you will see only one player taking damage. Throws always have the priority over another attacks no matter who have the advantage. If both player perform a throw on the same frame, only the one who have the advantage will move. A throw always connect on the very first frame after the input. A weird thing is that the damage still occurs one frame later for the player without the advantage even if it doesn't matter anymore since you will never see both player being thowed at the same time. If a player without the advantage do an attack while pushing [Back] on the same frame of a throw, this will prevent the player from performing a throw and the attack will connect. However, if both players do a throw on the same frame while holding [Back], the player without the advantage won't be able to prevent the throw! (weird) I thought that the player with the advantage would connect a jab instead but it's not the case.

Uppercut / Ducking attacks:

Many people think that you need to press both [Down] and [HP/LP/HK/LK] at the same time but it not exacly the case.

You only have to register the [Down] key on a frame (frame #1) and THEN register the [HP/LP/HK/LK] key on frame #7.

If you are already ducking, you have to hold [Down] while pressing [HP/LP/HK/LK].


Round House / Sweep:

You have to register the [HK or LK] input on a frame THEN register the [Back] input on the next frame.


Blocking:

When you tap [Block] (frame #1), it become active only on the 4th frame.

Seing your character holding block takes 10 frames but there is no recovery after releasing [Block].


Ducking:

When you tap [Down], the ducking animation starts only on frame #4.

You are ready to perform a ducking attack on frame #7.

Frame #8 is when you are holding [Down].

If you release [Down], you will be ready to perform a standing attack on Frame #15.


Jumping upward:

The animation starts 7 frames after you press [Up].

Depending on the character you are at the peak after 24 to 27 frames.

It takes 45 frames to recover.


Jumping Sideways:

You have to hold both [Up] and [Right or Left] for two consecutive frames.

The animation starts on the 3rd frame.

It still takes 45 frames to recover.


CONNECTION FRAME: (it's the frame of animation that has the hitbox)

I have tested every single moves on every character since some moves like the KNEE always hit on the same frame except for one character (Nightwolf).

Throws : frame #2

HK - LK : frame #8

HP - LP : frame #11

Ducking LP : frame #11

Sweeps : frame #16



Character | Elbow | Knee | Round House | Ducking LK | Ducking HK

| Kitana......| 6 | 5 | 13 | 8 | 14 | (46)
| Reptile.....| 5 | 5 | 13 | 8 | 14 | (45)
| Sonya.......| 5 | 5 | 10 | 8 | 11 | (39)
| Jax.........| 6 | 5 | 13 | 10| 11 | (45)
| Nightwolf...| 4 | 6 | 13 | 10| 14 | (47)
| Jade........| 5 | 5 | 13 | 8 | 14 | (45)
| Scorpion....| 5 | 5 | 13 | 8 | 14 | (45)
| Kano........| 5 | 5 | 13 | 10| 14 | (47)
| Mileena.....| 5 | 5 | 13 | 8 | 14 | (45)
| Ermac.......| 5 | 5 | 13 | 8 | 14 | (45)
| C. Sub-zero.| 5 | 5 | 13 | 8 | 14 | (45)
| U. Sub-zero.| 5 | 5 | 16 | 8 | 14 | (48)
| Sektor......| 5 | 5 | 16 | 8 | 14 | (48)
| Sindel......| 5 | 5 | 16 | 10| 14 | (50)
| Stryker.....| 5 | 5 | 16 | 8 | 14 | (48)
| Cyrax.......| 5 | 5 | 16 | 8 | 14 | (48)
| Kung Lao....| 5 | 5 | 13 | 8 | 14 | (45)
| Kabal.......| 7 | 5 | 13 | 8 | 14 | (47)
| Sheeva......| 5 | 5 | 16 | 8 | 11 | (45)
| Shang Tsung.| 5 | 5 | 13 | 8 | 11 | (42)
| Liu Kang....| 5 | 5 | 13 | 8 | 14 | (45)
| R. Smoke....| 5 | 5 | 16 | 8 | 14 | (48)
| H. Smoke....| 5 | 5 | 13 | 8 | 14 | (45)


Sonya is the most sudden character, and Sindel is the less.

Update: (october 4th 2008)

Recovery test (# of the frame on which you are ready to act after doing the move)

* I tested every single moves with every characters (with frame advantage and without it) without any exception. For example, I tested everything even if I knew that HP/LP and HK/LK have all the same numbers of startup frames and recovery frames, and I found that this is true but with the exeption of Kano's LK, which for some reasons have 3 extra recovery frames!

Moves that have the same recovery for every characters:

HP/LP......... 19
HK ........... 34
Ducking LP ... 25

| Character | Elbow | Knee | LK | Round House | Sweep | Ducking LK | Ducking HK | Uppercut Advantage / No Advantage

| Kitana......| 35 | 26 | 34 | 34 | 39 | 25 | 45 | 67/68
| Reptile.....| 30 | 26 | 34 | 34 | 39 | 25 | 45 | 67/68
| Sonya.......| 26 | 23 | 34 | 39 | 33 | 25 | 42 | 67/68
| Jax.........| 31 | 26 | 34 | 34 | 33 | 29 | 39 | 67/68
| Nightwolf...| 25 | 31 | 34 | 34 | 39 | 29 | 45 | 67/68
| Jade........| 26 | 26 | 34 | 34 | 39 | 25 | 45 | 67/68
| Scorpion....| 26 | 26 | 34 | 34 | 39 | 25 | 45 | 67/68
| Kano........| 26 | 26 | 37 | 38 | 33 | 29 | 45 | 67/68
| Mileena.....| 26 | 26 | 34 | 34 | 39 | 25 | 45 | 67/68
| Ermac.......| 30 | 26 | 34 | 34 | 39 | 25 | 45 | 67/68
| C. Sub-zero.| 30 | 30 | 34 | 34 | 39 | 25 | 45 | 67/68
| U. Sub-zero.| 26 | 29 | 34 | 37 | 39 | 25 | 41 | 67/68
| Sektor......| 26 | 26 | 34 | 37 | 39 | 25 | 45 | 67/68
| Sindel......| 26 | 26 | 34 | 37 | 39 | 29 | 45 | 67/68
| Stryker.....| 23 | 26 | 34 | 37 | 39 | 25 | 45 | 67/68
| Cyrax.......| 26 | 26 | 34 | 37 | 39 | 25 | 45 | 67/68
| Kung Lao....| 26 | 26 | 34 | 34 | 39 | 25 | 45 | 67/68
| Kabal.......| 36 | 26 | 34 | 34 | 39 | 25 | 45 | 71/72
| Sheeva......| 26 | 30 | 34 | 37 | 39 | 25 | 39 | 71/72
| Shang Tsung.| 26 | 26 | 34 | 34 | 39 | 25 | 38 | 67/68
| Liu Kang....| 26 | 26 | 34 | 34 | 39 | 25 | 45 | 67/68
| R. Smoke....| 26 | 26 | 34 | 37 | 39 | 25 | 45 | 67/68
| H. Smoke....| 26 | 26 | 34 | 34 | 39 | 25 | 45 | 67/68


* On missed ducking attacks, add one extra recovery frame on every characters.

* Elbow / Knee / HP / LP / HK / LK and Round House were tested using the down key ( -4 frames when you see the animation starting). Sweep / Ducking attacks / Uppercut and Throws were tested using the jump key ( -8 frames when you see the character leaving the ground and -14 for Ducking HK)

Walking Speed / Recovery on Connected Throws

*Every characters run at the exact same speed. The walking speed numbers means the number of frames it takes to walk the distance of one frame of running.

| Character | Walking Speed | Throws recovery

| Kitana......| 2.3 | 49
| Reptile.....| 2.3 | 57
| Sonya.......| 2.5 | 65
| Jax.........| 2.5 | 54
| Nightwolf...| 2.3 | 52
| Jade........| 2.1 | 54
| Scorpion....| 2.3 | 63
| Kano........| 2.1 | 46
| Mileena.....| 2.3 | 44
| Ermac.......| 2.3 | 63
| C. Sub-zero.| 2.3 | 63
| U. Sub-zero.| 2.3 | 61
| Sektor......| 2.3 | 67
| Sindel......| 2.1 | 68
| Stryker.....| 2.5 | 65
| Cyrax.......| 2.3 | 64
| Kung Lao....| 2.3 | 57
| Kabal.......| 2.5 | 61
| Sheeva......| 2.5 | 62
| Shang Tsung.| 2.4 | 78
| Liu Kang....| 2.4 | 56
| R. Smoke....| 2.3 | 42
| H. Smoke....| 1.9 | 63

(c) Remz 2008

 

[Shock]

You have to retest some contact frames because of the frame advantage. I tested Nightwolf's knee attack, and P1 connected on 5 frames, P2 connected on 6, P1 Nightwolf elbow connected on frame 3, and P2 on frame 4. If you reset the game or finish the match and retry again, it might be the other way around. I also tested Sindel and cut her frame data down from 50 to 46. Test every character vs themself only and record the minimums. Excellent post overall and I hope to see all the numbers worked out.

 

[Remz]

I did that kind of test, I'm pretty sure that frame advantage have nothing to do on the blockable frame. The only way I found to truly test the blockable frame is by trying to block the attack by taping block on that frame only (actually 3 frames before). You can't test by looking at the animation because it depends on frame advantage.

Did you test it by blocking that frame only or looking at the animation/blood/energy bar?

 

[Onewhite1Gold]

I'm not sure if this is right

| R. Smoke....| 5 | 5 | 13 | 8 | 14 | (45)
| H. Smoke....| 5 | 5 | 16 | 8 | 14 | (48)

I think H.smokes RH is mixed up with R.Smoke....

Robos all have 16 and male ninjas all have 13....im assuming from this awesome post.

 

[Remz]

Yes Onewhite1gold, I re-tested it and I think I made a mistake in the retranscription of my results. I edited it, thanks a lot.

I found some pretty weird things concerning the male ninjas, because reptile and ermac have 4 extra recovery frames on elbow when not having the frame advantage, but not Scorpion, Sub-Zero and Human Smoke. I'll post some result later as I have a lot more testing to do.

 

[dreemernj]

Very cool stuff. I have done a lot with frame advancing (figuring out infs with 1 frame openings to get a repeating pattern) but never documented much. I'm very interested in looking at the extra recovery when a character doesn't have frame advantage. 4 frames of recovery being added is an unusual number and suggests its reveal a strange mechanism of the game.

One thing I'll point out though, in doing combos that require frame counting I encountered situations where frame counts appeared to differ between versions of MAME. I didn't keep track of them and forget exactly what sort of differences there were but I think its possible.

One other interesting thing, you mention how you can tap down and then tap HP to uppercut. If you hold block and tap down, when you tap HP on that frame you'll uppercut through the block. Tough to time since its only one frame, but potentially helpful (and a lot of folks already know about but thought I'd mention). Its popped up in tournies a few times that I'm aware of.

EDIT:
I can't seem to reproduce longer recovery times on the different colored ninjas the way you describe. I think the animation might be a bit longer but if you hold a direction while waiting for them to recover, it's a good indication of when they've recovered because they'll start to move even before the animation changes.

 

[Shock]

OK after an obscene amount of testing with Dreemer I finally figured out what the Hell is going on here. We figured out some pretty trite details about the gameplay in the process. Basically all these numbers are animation start to contact, + 1 for that weird frame where they do things like, move their sprite on the jump before changing the animation.

You can block an attack by pressing block for just 1 frame, but it has to 3 frames before contact, since there is a 2 frame delay. Block is active upon the first frame you hit the button, but isn't exactly applicable until 2 more frames have passed. So as long as there are 3 frames avaible upon pressing block, you will block it. I hope I'm explaining that correctly.

Here is an interesting example of somewhat insignificant data:

You can press HK, then on the very next frame press back, and it will do a RH. If there is an empty frame between HK and back, it'll do a HK, even if you hold HK for another frame. There is essentially a "null" frame available to activate something like a RH or sweep.

f = frame
f(B+HK) = HK (with step back)
2f(B+HK) = RH (with step back)
f(HK)+f(B) = RH (no step back)
2f(HK) +f(B) = HK (no step back)

 

[MKK hanzo]

Great thread and very informative.

You can block an attack by pressing block for just 1 frame, but it has to 3 frames before contact, since there is a 2 frame delay. Block is active upon the first frame you hit the button, but isn't exactly applicable until 2 more frames have passed. So as long as there are 3 frames avaible upon pressing block, you will block it. I hope I'm explaining that correctly.

That can explain why sometimes for example Kabals Dash goes thru a perfect timed karajab (Lp into Block, piano style)? I always wondered about that "delay" in certain situations.

You can press HK, then on the very next frame press back, and it will do a RH. If there is an empty frame between HK and back, it'll do a HK, even if you hold HK for another frame. There is essentially a "null" frame available to activate something like a RH or sweep.

True. This is the way to do the SUJK stuff in MK4 too. You have to leave an empty frame between the button pressing.

However, if both players do a throw on the same frame while

holding [Back], the player without the advantage won't be able to prevent the throw! (weird) I thought that the player with the advantage would

connect a jab instead but it's not the case.

This explains the "non escapable" throws i think?

 

[Remz]

Thanks for you interest and contribution to this thread, The next thing I'll do is to test recovery frames on all the moves using frame advance, holding jump on every frame and checking when the animation start not counting the frames it takes to actually see the effect of taping up. (This is how I've notice the extra "real" recovery frames on the elbow of some males ninjas like reptile).

Note: Just to be sure that the numbers are clear, I always count the frame on which I register the switch that create the move as being frame #1.

You can block an attack by pressing block for just 1 frame, but it has to 3 frames before contact, since there is a 2 frame delay. Block is active upon the first frame you hit the button, but isn't exactly applicable until 2 more frames have passed. So as long as there are 3 frames avaible upon pressing block, you will block it. I hope I'm explaining that correctly

This is exacly what I say when I say that block become active on frame #4. (frame #1 is pressing block, frame #2-3 are delay) maybe your way of explaining is better.

The next step will be to test and compare the special moves with numbers. Maybe I'll have to classify different types of moves in categories like "projectile" and "anti-air" to compare the moves on different criteria like the speed, the recovery, the range, priority...

A nice thing that I would also like to see is a classified database of screen shots with highlighted hit boxes like in one picture I saw in a post by Shock on two sub-zero doing jump kicks without hurting each other.

edit: Here is another example I found using the search function:

One thing I'll point out though, in doing combos that require frame counting I encountered situations where frame counts appeared to differ between versions of MAME.

Maybe I'll have to run some test with different version of MAME, but are you sure that the difference is not due to frame advantage? Because even if I'm pretty sure that frame advantage doesn't change the blockable frame, it seems to have an influence on recovery frames on very specific scenarios.

 

[Shock]

OK cool, we'll definitely get to the bottom of this and come up with a way to standarize the terminology and where to start counting the frames. I prefer to start with the animations because the start up is the same regardless for everything.

Just so you know, the collision box screen shots are from N64. The yellow outline I added in to show where the sprites actually are since the overlap causes an illusion. I think the one I made of Sub-zero was done manuallly by me iirc.

 

[dreemernj]

How did you test out the throws? I went through when I had a minute last night and tried out both players holding away and pressing LP at the same time but I never got any unusual results.

I tried having both players hold away and tap LP at the same time and both did LP's. I tried having both tap LP and then wait 1 through 4 frames before pressing back and got normal results with that as well. I then did holding away for 1 or 2 frames before pressing LP and just got standing LP's. Trying every configuration of frames, I couldn't get it to happen.

And as far as the differences in the different versions of MAME, I wouldn't completely rule out that frame advantage was playing tricks with me. I will look around all the weird versions of MAME I have and see if I can figure out what happened. It came up when I was trying to record a combo and I couldn't get any versions of MAME with video recording to work. I literally tried a dozen versions and one of them had an issue with attacks and screen shake being different from the other versions. But, chances are, it's an emu that very few people would use (and the difference was a frame or two, lol).

A database of hitboxes would be very cool. The places where characters can be hit would be a bit annoying to put together and it doesn't have a huge amount of value because of how formulaic those hitboxes are. But the hotboxes for the attacks could be interesting.

Sorry, this is very minor but just to nitpick to avoid confusion, when you say a move is blockable on a single frame, that's the first frame the move is blockable. It's a minor thing but as this stuff compiles that might confuse newcomers into thinking attacks only have a 1 tick hitbox. EDIT: Or maybe say it's the frame of animation that has the hitbox. Whatever conveys that the hitbox is usually out for a few frames.

 

[MKK hanzo]

A database of hitboxes would be very cool. The places where characters can be hit would be a bit annoying to put together and it doesn't have a huge amount of value because of how formulaic those hitboxes are. But the hotboxes for the attacks could be interesting.

Great. I wasin the begginning of something like that but right now im kind of busy to mess up with the emu and copy and paste process hehe

 

[Remz]

Sorry, this is very minor but just to nitpick to avoid confusion, when you say a move is blockable on a single frame, that's the first frame the move is blockable. It's a minor thing but as this stuff compiles that might confuse newcomers into thinking attacks only have a 1 tick hitbox. EDIT: Or maybe say it's the frame of animation that has the hitbox. Whatever conveys that the hitbox is usually out for a few frames.

I'm not sure if I understand exacly what you are saying (my english is so-so) but I'll try to explain myself as clearly as possible.

When I tested the attacks, the only thing I needed in order to block any attack was to press block on a single frame. I never encountered any occasion where holding block for more that one frame was necessary. I tested all the attacks trying to find on which single frame I had to tap block to parry the attack. Then, I tested it in reverse by holding block on all other frames to see if I get hit to confirm that I have the right frame. Do you say that you need to hold block for multiple frames in some cases?

...And I tested what I was saying (from memory) about different males ninjas. And I can confirm that Reptile-Ermac-C.Sub-Zero have 4 extra recovery frames on elbow than Scorpion and Human Smoke. But again, no matter who have the frame advantage.

I test recovery by holding down immediately after pressing the attack and I count the frames until I see the character starting to duck, then I substact 4 to the number (the time it takes to see the char starting to duck when you activate the down switch). That way I obtain the number of the frame on which you are ready to act. After that, I confirm using jump instead, and I wait until I see the character leaving the ground, then I substract 8 to confirm the same number. I takes actually 7 frames to see your char leaving the ground when you press up on frame 1 but jumping need one extra recovery frame (weird).

 

[Shock]

Simply recording all the collision boxes in the N64 MKT then converting them to UMK3 frames would give a general idea for most characters, even though I know for certain there are some collision box inconsistencies in N64 MKT. It would then be possible to determine the boxes for Sheeva and Sub-zero based on using say, the male ninjas standing HP or LP to figure out where the front collision is, and put it the exact same number of pixels away from the back end of the sprite. This would be tedious but it'd work for a complete set.

Remz, Dreemer means that some attacks, probably even most, have collision boxes that last more than one tick, or even move (like projectiles), in essence, a reason why you can run in and block Sindel's scream or Kitana's lift, mixed with the fact that the cold collision box for the running character drastically changes between the run frame, and the blocking frame. But basically, you could do the same for say, a LK, run in, LK, hit block during run and then possibly block it a frame after the collision box is initially active. What I was determing through my tests originally is just the very first frame where a collision is active, by holding block with the other character.

One of the best ways to set up combos is to get the hits on the very last frame the collision box is available, say for something like, a Double HK in the corner with Jax after his combo, or Stryker's riot gun cross screen inf, you want them to land on top of the gun fire, instead of being hit as soon as it's there, this optimizes recovery times.

 

[Remz]

Character | Elbow | Knee | Round House | Ducking LK | Ducking HK

| Kitana......| 6 | 5 | 13 | 8 | 14 | (46)
| Reptile.....| 5 | 5 | 13 | 8 | 14 | (45)
| Sonya.......| 5 | 5 | 10 | 8 | 11 | (39)
| Jax.........| 6 | 5 | 13 | 10| 11 | (45)
| Nightwolf...| 4 | 6 | 13 | 10| 14 | (47)
| Jade........| 5 | 5 | 13 | 8 | 14 | (45)
| Scorpion....| 5 | 5 | 13 | 8 | 14 | (45)
| Kano........| 5 | 5 | 13 | 10| 14 | (47)
| Mileena.....| 5 | 5 | 13 | 8 | 14 | (45)
| Ermac.......| 5 | 5 | 13 | 8 | 14 | (45)
| C. Sub-zero.| 5 | 5 | 13 | 8 | 14 | (45)
| U. Sub-zero.| 5 | 5 | 16 | 8 | 14 | (48)
| Sektor......| 5 | 5 | 16 | 8 | 14 | (48)
| Sindel......| 5 | 5 | 16 | 10| 14 | (50)
| Stryker.....| 5 | 5 | 16 | 8 | 14 | (48)
| Cyrax.......| 5 | 5 | 16 | 8 | 14 | (48)
| Kung Lao....| 5 | 5 | 13 | 8 | 14 | (45)
| Kabal.......| 7 | 5 | 13 | 8 | 14 | (47)
| Sheeva......| 5 | 5 | 16 | 8 | 11 | (45)
| Shang Tsung.| 5 | 5 | 13 | 8 | 11 | (42)
| Liu Kang....| 5 | 5 | 13 | 8 | 14 | (45)
| R. Smoke....| 5 | 5 | 16 | 8 | 14 | (48)
| H. Smoke....| 5 | 5 | 13 | 8 | 14 | (45)

Then those numbers mean the first frame where you need to press block in order to prevent the attack if I'm right.

Edit:

One of the best ways to set up combos is to get the hits on the very last frame the collision box is available, say for something like, a Double HK in the corner with Jax after his combo, or Stryker's riot gun cross screen inf, you want them to land on top of the gun fire, instead of being hit as soon as it's there, this optimizes recovery times.

It's very interesting, there is a lot of things I don't know about this game because I'm a relatively new comer in umk3. Thanks to you and DreemerNJ for pointing things like that because I don't want to induce misconception with the compilation I'm doing and the way I interpret or express the results.

 

[dreemernj]

EDIT: Shock already explained this fine, lol. Didn't notice his post. Posting from cell phones is annoying.

All I meant about the collision box lasting more that one tick is when you do a punch, your punch is out there active for more than one tick. Someone could run into the punch's hitbox after the first frame it can be blocked.

Whenever an attack touches an opponent, whether they block it or not, the hitbox deactivates and basically disappears. So if you do the attack right next to your opponent, it'll be blockable on the first frame the hotbox is out and if they block it or get hit, the hotbox deactivates and can't keep hitting them.

The death punch in MK2 happened because the hotbox didn't disable upon contact in some situations. So if someone jumped at you, you could aaHP them in a certain way that they landed on your aaHP after the first frame that it can hit them and the hotbox would stay active and in the course of about 50-75ms, it would do 3 or 4 hits to them. Each tick of the game (if you were in MAME pausing it, each frame advance) it would count another hit.

In UMK3, AFAIK, there are no times when the hitbox stays active after an initial collission.

A clearer example is a jump kick. When you do a jump kick there is a large window for when the kick can hit the opponent, not just a single tick.

 

[Shock]

Well, I think it needs to be broken down to something more viewable like the animation start up, since you can't see that your opponent pressed anything until that happens.

P1 Kitana (advantage), Knee Lift (button pressed), 1 frame (delay), 1frame (delay), 1 frame animation, 1 frame animation (block MUST be pressed upon this frame or it is not blockable), 1 frame animation , 1 frame animation (contact frame).

So to break that down with direections, it goes:

Press P, Hold shift, Hold P1 HK, press P, 1 (command is sent on this frame), 2 (delay), 3 (delay), 4 (animation), now you must Press P2 Block or it cannot be blocked which is started on the next frame, 5 animation, 6 animation, 7 animation+blocked.

So this is where the 5 comes from?

Also notice that the frame advantage Kitana displays two different frames of animation for block, and the dis Kitana performs 1. These are more ways to determine who has frame advantage or not without risking glitch canceling.

 

[Remz]

To Dreemer (before I saw Shock's post):
Thanks for your explanation, I have now a better understanding of how it works, it's very interresting.

I edited "CONNECTION FRAME: (it's the frame of animation that has the hitbox)" in the first post. Maybe I'll need a better term to name that first frame.

Ohh, and I forgot to tell that I ran a good amount of test trying to find out how the attribution of the frame advantage works and I found no pattern, but I can say that the frame advantage is not ALWAYS to the same player for the entire match. Switching from Robot to Human Smoke caused the advantage to switch sides the few times I did it as well as loosing or winning a round by time out.

 

[Remz]

Press P, Hold shift, Hold P1 HK, press P, 1 (command is sent on this frame), 2 (delay), 3 (delay), 4 (animation), now you must Press P2 Block or it cannot be blocked which is started on the next frame, 5 animation, 6 animation, 7 animation+blocked.

So this is where the 5 comes from?

Yes!

 

[dreemernj]

Wow, I never even thought to try frame advantage tests with robo smoke changing. That suggests shang could change it as well.

One bizarre thing though. I only tried this with scorp vs scorp, but for them, I found the one without advantage won if both characters went for elbow (or axe in his case). The one with advantage won out in everything else for me. I mentioned it to shock and I believe when he tried scorp vs scorp, there were other moves the character without advantage won with.

So it seems in some cases frame advantage is not universal to all of a character's moves, at least with scorpion vs scorpion.

Great thread by the way :-D

I haven't looked at frame stuff in a while. Forgot how much I have messing with it.

 

[MKK hanzo]

Simply recording all the collision boxes in the N64 MKT then converting them to UMK3 frames would give a general idea for most characters, even though I know for certain there are some collision box inconsistencies in N64 MKT.

Indeed. Overall in the HpHp or LpLp flurry. You can see the enourmous change in both BUT I noticed that if both are used as an aa (reptile does the perfect example) the hitbox for the one that takes the combo differs a LOT more from UMK3. Try just Rept inf on Jax and you will see it. I think that the attacker hitboxes are so so (pretty much the same as UMK) but on the player that receives combos/hits/specials etc are very screw up.

I think that only the attack hotboxes could help here.

Great thread by the way :-D

Very true

 

[Remz]

One bizarre thing though. I only tried this with scorp vs scorp, but for them, I found the one without advantage won if both characters went for elbow (or axe in his case). The one with advantage won out in everything else for me. I mentioned it to shock and I believe when he tried scorp vs scorp, there were other moves the character without advantage won with.

I'm happy that you talk about it because while I was running tests on frame advantage and on the extra recovery frames of Scorpion and Human Smoke's elbow move, I've tested Scorpion against all other ninjas and at some moment I came to the conclusion that this move have always the priority over other moves. But I found other cases in which everyting was normal. The only thing I know for the moment, is that there is someting weird about that particuliar move of Scorpion and Human Smoke as this is the same move that have a faster recovery than the other ninjas.

 

[MKK hanzo]

I'm happy that you talk about it because while I was running tests on frame advantage and on the extra recovery frames of Scorpion and Human Smoke's elbow move, I've tested Scorpion against all other ninjas and at some moment I came to the conclusion that this move have always the priority over other moves. But I found other cases in which everyting was normal. The only thing I know for the moment, is that there is someting weird about that particuliar move of Scorpion and Human Smoke as this is the same move that have a faster recovery than the other ninjas.

True. That happens when I fight with RZP (a friend) a lot. Also funny fact is that it pops out when he on reaction hold block but his hand accidentally pushes Hp and the axe comes out as a single hit of course but funny too is that shit throw us off balance so much causeit sometimes break a perfect frame accurated strat or rushdown. We always LOL at it and WTF cause it beats everything and its almost done at random.

 

[dreemernj]

You should do frame by frame testing on breaking the combos for each character

So everyone can see which part of the combo is best to break(least recovery time) to start run-jab(s)/mix-ups again.

(A lot of work, but fun if you are already interested in this stuff)

Great idea. That's probably one of the most useful things that can be figured out by counting frames.

Many of the autocombo hits share recovery animations so maybe it wouldn't be too hard to figure out where the fast break is. It's probably something like, 1 to 3 hits = X recovery, 4 to 7 hits = Y recovery.

 

[Remz]

Yes, this is a great idea, I'll work on that as soon as I finish testing recovery on normal moves.

By the way I've found that frame advantage affect the uppercut recovery frames on every character. It's the only move so far that is affected by frame advantage.