Trying to fix the combo counter in 20XX to be more accurate

[GopherHat]

Ok so I have been trying to fix the combo counter that the game uses and that 20XX uses to take into consideration tech chases and such. I've been trying to find out where the combo code is but I am having difficulties. I first went into dolphin and tracked down these addresses that hold the combocounter in the game.

Current number of combo (melee NTSC 1.02)
0x0047380c

20XX combo counter (combo counter used in the set timer to combo counter code)
0046B6C8

So then I set a memory breakpoint onto this address
8046B6C8

Which led me to find these instructions that the code breaks on

lines that the program breaks on
lhz r0, 0x002C (r4)

and
mr r27, r3

this is when the player has not started a combo

When a combo has been started I found this code to be at the breakpoint
addi r31, r4, 9424

and
mr r27, r3
again


I have absolutely no idea how to go forward at this point. Can someone point me in the right direction as to how to approach this?

 

[Achilles1515]

Don't use 20XX to investigate the combo counter. Use vanilla.

8046b6c8 is the address for the timer value.

 

[Sham Rock]

Not gonna pretend like I´ve read that, but I do have some little notes because dan asked me to find some stuff out about the combo counter for TE

8047380c the number displayed in training mode for the combo counter

804545cd
804545d1 the current combo counter , resets to 0 when the opponent leaves hitstun

80040f80 addi r0,r4,1 takes the current combo counter r4 and adds 1 to store below

8006ab54 replacing this bl with a nop will result in the game not resetting the combo statistic (counter and damage)

all this stuff still gets accessed in multiplayer, maybe it will be of assistance

 

[GopherHat]

Don't use 20XX to investigate the combo counter. Use vanilla.

8046b6c8 is the address for the timer value.

Ok I'll stop testing on 20XX. It looks like i was looking in the wrong place a little bit.

Not gonna pretend like I´ve read that, but I do have some little notes because dan asked me to find some stuff out about the combo counter for TE

8047380c the number displayed in training mode for the combo counter

804545cd
804545d1 the current combo counter , resets to 0 when the opponent leaves hitstun

80040f80 addi r0,r4,1 takes the current combo counter r4 and adds 1 to store below

8006ab54 replacing this bl with a nop will result in the game not resetting the combo statistic (counter and damage)

all this stuff still gets accessed in multiplayer, maybe it will be of assistance

This looks really useful. Thank you so much . I'll have a look at the asm at 80040f80 and see if there is any branches that link to it. Same with 8006ab54

 

[GopherHat]

I'll work on it more tomorrow. Running low on motivation.

 

[GopherHat]

I had this done a while ago but haven't thought to put it up until now.
There is a problem in the logic somewhere which causes it to crash unexpectedly

804545d1 combo counter

0x80453FC0 p2 pointer

follow the pointer to 80cb0ae0

action state at 0x70 offset from the pointer of the player

resultant address:80CB0B50

state when waiting: E

combo logic notes
comboee is jumpable within damage fall
wait1 and wait 2 are completely actionable
downboundD downboundwait

current combo check is basically
they just got hit!
    are they in hitstun
        yes
            combo continued add 1
        no?
            new combo started set to one (or just set to zero and add one)

combo is happening!
    are they in hitstun?
        yes
            carry on
        no?
            set the combo back to zero


address of the current combo 804545d4

current combo happening 804545d4
last combo (training mode thingk) 8047380c

r3:80473758

line where the game zeros the combo
80041210
lbz    r0, 0x0022 (r30)

line where the game checks if the combo is less than 999
80188984
cmpwi    r30, 999

line that I want to replace
80188980
lwz    r30, 0x010C (r31)

instead of loading whatever is at r31 (which from testing seems to be 8047380c somehow) just load the value at the current combo instead

proposed line:
80188980
lhz r30, 804545d4

gecko code time:

Insert ASM
C2______ NNNNNNNN
ZZZZZZZZ ZZZZZZZZ
ZZZZZZZZ ZZZZZZZZ
ZZZZZZZZ 00000000

This code will replace the instruction at ba+______ with a branch that will point to ZZZZZZZZ.
The replaced is not saved, the code creator must then put it in his code manualy (if needed).
The instruction MUST end with ONE 00000000, because the code handler will add a "b (ba+______)" instruction there.
If your asm code fills all the line, add a 60000000 00000000 under it (and count this line in NNNNNNNN).


To use po instead of ba, change the codetype from C2 to D2.
For values of ______ >= 0x01000000, add one to the codetype.

value of r31 at this point:80473700
the value of the memory address which contains the displayed combo counter in training mode: 0x8047380c

0x80473700 + x = 0x8047380c

x = 0x8047380c - 0x80473700

x = 0x010C

which explains why the 0x010C is there:
80188980
lwz    r30, 0x010C (r31)
(this really took me hours to figure out asm is really cryptic :( )

ok so now we need to change it to the value that I want: the current combo counter: 804545d4

0x80473700 + x = 0x804545d4

x = 0x804545d4 - 0x80473700

x = -1f12c

(doesnt seem possible in one line)

-0xF12D + -0x7fff = -1f12c

lwz r30, -0xF12D (r31)


now to subtract a furthur

0x1f12c - 0x7fff - 0x7fff - 0x7fff =
0x712F

to get: at 80188980
subfic r31,r31,0x00007fff
subfic r31,r31,0x00007fff
subfic r31,r31,0x00007fff
lwz r30,-0x712F(r31)

My code:
C2188980 00000003
23FF7FFF 23FF7FFF
23FF7FFF 83DF8ED1
60000000 00000000

(it works but I realise now that it was a little stupid since the counter seems to only update when a hit happens so the positioning was quite terrible)

ok so now to the real thing:


80041194
add    r3, r3, r4
//r4 seems to contain the current combo at this point or just zero
r3 has address 804545b8

8003820c//this seems to be the line that it all happens on (setting the value to the number of hits)
rlwinm.    r3, r27, 0, 16, 31 (0000ffff)
5763043f

80041210 (this seems to be where it clears)
lbz    r0, 0x0022 (r30)
881e0022



80041214 // the first hit?
rlwinm.    r0, r0, 25, 31, 31 (00000080)
5400cfff

current damage of the hit
804545c8

strange pointer?
80041204

->

c3e285a0


0x80453FC0 p2 pointer

follow the pointer to 80cb0ae0

action state at 23A0 offset from the pointer of the player

resultant address:0x80CB2E80
p2 hitstun address:0x80CB2E80

probably a better address for the combo
804545ce

main condition for adding
80041140
cmpwi    r0, 0


8004114c
bne-     ->0x80041194

need some way to check this: 80CB0B50

some possible registers:
r25:80cb4080
r10:80cb58c4

r30 looks clean: 0

80cb4080 - 80CB0B50 = x

x = 0x3530

//AT 8004114c
//load the action state of p2
lwz r30, 0x3530(r25)
//compare the action state with the one that is wait (wait is 0x000E)
cmpi 2,0,r30,0x000E

//80041194 is the address of where it goes
//r28 happens to be free :)
    //the program counter happens to be the closest value to 80041194
    //at this point the pc should be 6 + 80041194
    //lwz r28,6(PC)
    //PC cannot be used for that :( its probably for the best actually

//I think this is how slw is supposed to work
addi r28,0,0x7fff
addi r28,r28,0x5
slw r28,r28,0x4
addi r28,0,0x1194


//branch if they are not equal to the part of the code where the combo is added
//001zy is the value to branch if it is false
//2 for the equal comparison bit in the conditional register

//bc 001zy,2,r28
bc 8,2,r28




lwz r30, 0x3530(r25)

cmpi 2,0,r30,0x000E

addi r30,0,0x7fff
addi r30,r30,0x5
slw r30,r30,0x4
addi r30,0,0x1194

bc 8,2,r28


C204114C 00000004
83D93530 2D1E000E
3B807FFF 3B9C0005
7F9C2030 3B801194
4102001C 00000000

 

[Achilles1515]

I was going to write a code to make grabs not reset the combo counter, then found out this a bad idea and will slightly affect gameplay.

In Melee, there are in-game physics features to mitigate standing infinite combos.

If a player's combo count is greater than or equal to 5 while grounded, the player is pushed backwards by an invisible force at the rate of 0.025 meleemeters/frame. As the combo grows to 13 or more, the rate of pushback is increased to 0.070 meleemeters/frame.

So under certain circumstances, changing the combo count logic to include grabs/techchases could cause abnormal x-coordinate movements of the attacking player.

PlCo.dat Offsets:
0xa4a4 5 = combo count (greater than or equal to) necessary to use 0.025 TopN x-coord movement
0xa4a8 13 = combo count (greater than or equal to) necessary to use 0.07 TopN x-coord movement
0xa4b0 0.025 = TopN x-coord movement per frame if combo count greater than 5
0xa4b4 0.07 = TopN x-coord movement per frame if combo count greater than or equal to 13
0xa4b8 20 = number of frames to execute TopN x-coord movement

@tatatat0

 

[Stride]

I was going to write a code to make grabs not reset the combo counter, then found out this a bad idea and will slightly affect gameplay.

In Melee, there are in-game physics features to mitigate standing infinite combos.

If a player's combo count is greater than or equal to 5 while grounded, the player is pushed backwards by an invisible force at the rate of 0.025 meleemeters/frame. As the combo grows to 13 or more, the rate of pushback is increased to 0.070 meleemeters/frame.

So under certain circumstances, changing the combo count logic to include grabs/techchases could cause abnormal x-coordinate movements of the attacking player.

I had no idea about this. Are there any circumstances under which that anti-infinite feature could be activated? What grounded attacks are there that can true combo for 5 hits, if any?

 

[Achilles1515]

I had no idea about this. Are there any normal circumstances (i.e. in an unmodified game) under which that anti-infinite feature could be activated? What grounded attacks are there that can true combo for 5 hits, if any?

Jabs. Fox's dair, and general waveshine combos. Just any string of attacks that increments the combo counter greater than 5. The frame timer for implementing the TopN x-coord movement is reset to 20 after every hit past 5 for the combo counter. It's subtle, but there.

 

[Achilles1515]

Also, I think the combo counter will continue to increment if a player is hit within two frames of leaving hitstun. This is quite depressing and means we have been duped this entire time...