Rayman 3 scores

[The Jonster]

I guess someone could try to analyse his code in Raymap? I might have a look tonight...

Ooh, whenever you do, lemme know what your results bring up!

Just seen this video. I think I could actually peform this combo myself (even though it depends on the fish). It looks easier than me attempting to blow 30 minutes in the PC version trying to get to the back of the ship with the blue can (I'm probably still doing something wrong lol)

[Steo]

That's a very interesting way of doing this actually, probably very difficult but also means you don't have to rely on the glitch to get to the top with the lockjaw.

[Maz]

If you don't want to use the Launching Glitch, I'd recommend this approach instead (combo at 0:55):



With this one, you have some time to get into a good starting position, and having all yellow gems accessible also makes the combo a lot easier. The newer version is a combo which was optimised purely for the sake of being optimised, you don't actually -need- to play that garbage it's much easier to find an additional one or two thousand points in part 2 than subject yourself to this abomination of an approach to part 1.

[Steo]

It's amazing though that even 20 years later, people are still just looking into finding new combos for the sake of it, just for fun. I think it's really cool to see that some are this dedicated to it.

I feel like I won't ever bother to really try get more than I have right now, but it's amazing to see 900k being broken by people.

[PluMGMK]

I guess someone could try to analyse his code in Raymap? I might have a look tonight...

Ooh, whenever you do, lemme know what your results bring up!

Thanks for reminding me, I would have completely forgotten otherwise!

Basically, it is NOT random: Razoff will always choose the direction that moves him furthest away from Rayman. So because Cut jumped to the left before shooting, Razoff ran to the right. The rest of the post is technical detail that you don't have to read if you don't want to…
EDIT: I guess it's kinda obvious really, and Cut probably knew this already But hey, at least now we can see it for definite in the code



OK, so I realized that simply reading the code wasn't going to get me very far, since there was basically no way for me to find out the relevant script entries So I spent much of the afternoon trying to figure out how to get a live report of what behaviour he was in while playing the level. I don't know if it's because I'm a little rusty at reverse engineering / debugging, or just because the levels of indirection in this engine always tie my mind up in knots, but I feel it took way longer than it should have to come up with the right GDB commands to do this

Code: Select all

# In GDB, to watch the comport of superobject 0x050B6BBC, do
print/x *(*(*(*(0x050B6BBC+4)+0xc))+4)+8 # This establishes the location of the pointer to the current behaviour
watch *(int*)$ # This sets a watchpoint at that location to alert the debugger when the behaviour changes
commands
silent # This suppresses the usual notification message when the watchpoint is hit
printf "Now in behaviour %d\n", (*(*(*(*(*(0x050B6BBC+4)+0xc))+4)+8) - *(*(*(*(*(*(*(0x050B6BBC+4)+0xc))+4)))))/0xc # This prints our own message, subtracting the behaviour base pointer (from the AI model) from the current behaviour and dividing by 12 (the size of the behaviour structure)
c # This automatically continues the game so it is not interrupted after the message is printed
end

From this I determined that behaviour 13 is the one where Razoff stands in the corridor shooting at Rayman, and behaviour 5 is the one where he runs away. So, the relevant code would appear to be the transition from 13 to 5. I found the relevant script using Raymap fairly quickly, and here it is with a few comments I added myself:

Code: Select all

// Waypoint_27 = next waypoint, Waypoint_28 = current waypoint, Waypoint_30 = previous waypoint
ChangePersoSighting(((YLT_RaymanModel)GetPerso("Rayman")).Position() - Position());
Float_12 = AbsoluteValue(GetNormalSlopeVectorAndAngle(Orientation(), ((YLT_RaymanModel)GetPerso("Rayman")).Position() - Position(), 0));
RotateModuleAroundX(unknown, Float_12);
Float_34 = DistanceXYToPerso(((YLT_RaymanModel)GetPerso("Rayman")));
if (IsInAction(Razoff.Action[13]) && CheckActionEnd() || Float_34 < 15f)
{
  ChangeAction(Razoff.Action[4]);
  Int_37 = NetworkAllocateGraphToMSWay(Graph_0, CrossProduct(5), WayPoint_28, 2);
  WayPoint_27 = NetworkNextWPOfType(Graph_0, 0, CrossProduct(5), Nowhere, Nowhere, -Orientation()); // MINUS "Orientation()" => run AWAY!
  ChangePersoSighting(GetWPAbsolutePosition(WayPoint_27) - Position());
  WayPoint_30 = WayPoint_28;
  ReinitCollider();
  Int_16 = GetTime();
  Float_43 = 10;
  Int_21 = (GetTime() - 1500);
  SetActionReturn(1);
  ChangeMyComport(NIN_m_Razoff.Intelligence[5]); // Start "running away" behaviour
}

This is the first script entry in Razoff's behaviour 13, and it handles both making him face Rayman continuously, and making him run away if Rayman hurts him or gets within 15 length units (centimetres?) of him.

The first line of code, "ChangePersoSighting", ensures that Razoff is always facing Rayman while this behaviour is running, which is important.

The "If" statement handles making him run away if Rayman is too close. The line beginning with "WayPoint_27" is the one where he chooses which node to run to. I'm not 100% sure what the arguments represent, but it is important to note that one of them is "-Orientation()". The minus sign means that this gives a vector which is the exact opposite direction to the one he's facing in. Because of the "ChangePersoSighting" line, this means it always points directly away from Rayman.
Then the game's code goes through the nearest nodes on Graph_0. As far as I can tell, the relevant nodes are the three highlighted in orange here:

Presumably, it gets the vector distance from Razoff to each possible node, and picks whichever one makes the smallest angle with the "-Orientation()" vector. If Rayman is to his left, that will always be the node to the right, and vice versa. It will never be the node at the bottom of the picture, because that would involve running towards Rayman!

Anyway, I'm a bit tired at the moment and frazzled from trying to figure out what code to look at (as mentioned above) so this post may not be the most coherent Let me know if anything is unclear / makes no sense…

[Maz]

I feel like I won't ever bother to really try get more than I have right now, but it's amazing to see 900k being broken by people.

We're still holding out hope that we'll crack the million some day. It would never be more than a theoretical value, since at that point the combos would be so brutal that no sane person would try for all of them, but even then it would be really cool to get there

[dr_st]

What about TAS Rayman3? AI Rayman3?

[Maz]

I don't know if you're talking about scoring specifically, but TAS is something that we (as in, a couple of scoring people) have recently started considering, for purely academic purposes of course. In the years before, we always tried to find better stuff for the purpose of actually playing for points, so TAS was not something we ever looked into up until now.

There was a person named Lance on this forum a couple of years ago who did TAS scoring, but I don't remember what the final scores were.

Edit: I went back and checked Lance's posts, but the only thing I could find was a mention of a 903k total (seems like he stopped playing before a lot of major improvements were found), which is not the best benchmark for what TAS would make possible.

[Steo]

I feel like I won't ever bother to really try get more than I have right now, but it's amazing to see 900k being broken by people.

We're still holding out hope that we'll crack the million some day. It would never be more than a theoretical value, since at that point the combos would be so brutal that no sane person would try for all of them, but even then it would be really cool to get there

Yeah even if it's theoretically possible it's just cool to know. At the moment, I don't even have a hold of my save, since I used to play on Xbox. My Xbox is back in Ireland and I hope it even still has the save, though it gives me a reason to install it on PC and give it another go. I don't see myself reaching that score again, but it would be fun to break out the game again nonetheless.

[Reese Riverson]

Yeah even if it's theoretically possible it's just cool to know. At the moment, I don't even have a hold of my save, since I used to play on Xbox. My Xbox is back in Ireland and I hope it even still has the save, though it gives me a reason to install it on PC and give it another go. I don't see myself reaching that score again, but it would be fun to break out the game again nonetheless.

You should definitely give this game a good replay and try to beat your old score or at least come close to it.

Master and I played through it yesterday with you watching us, and while I didn't break 400k with mine, but getting close to it, I know it was a lot higher than my first playthrough when I beat it as a kid.

[The Jonster]

Yeah, that was a lot of fun to watch! I do wish Steo's gf had stuck around a little bit longer though, she missed out on all of her appearances in the game!
I hope once I give another go for a good score, I'll have some sort of remembrance of all the skills and combos I had done to get that 700K score.

[Maz]

Yeah even if it's theoretically possible it's just cool to know. At the moment, I don't even have a hold of my save, since I used to play on Xbox. My Xbox is back in Ireland and I hope it even still has the save, though it gives me a reason to install it on PC and give it another go. I don't see myself reaching that score again, but it would be fun to break out the game again nonetheless.

No need to be so pessimistic Even if the road to 800k+ seems difficult in hindsight, it will be a lot easier to get back there now that you've done it once and know what to do. Trust me, I speak from experience

You'll have to find time and motivation for scoring (which is often the real issue with R3), but if your biggest worry is the gameplay, then I have no doubt that you'd get back to ~800k with no issues!

[Steo]

Yeah, that was a lot of fun to watch! I do wish Steo's gf had stuck around a little bit longer though, she missed out on all of her appearances in the game!
I hope once I give another go for a good score, I'll have some sort of remembrance of all the skills and combos I had done to get that 700K score.

Haha yeah, too bad she missed that part.

I'm sure you'll break 700k again no problem.

No need to be so pessimistic Even if the road to 800k+ seems difficult in hindsight, it will be a lot easier to get back there now that you've done it once and know what to do. Trust me, I speak from experience

You'll have to find time and motivation for scoring (which is often the real issue with R3), but if your biggest worry is the gameplay, then I have no doubt that you'd get back to ~800k with no issues!

Thanks for the advice, and I suppose that's true, that if you done something once then surely it's possible again. I really dread the Bog of Murk.

Finding the time to dedicate to it without feeling like I'm not doing anything productive is what gets me, especially being in a relationship etc. Of course, I can still make time to try play it here and there, so long as I can manage my time well.

[Hunchman801]

https://www.youtube.com/watch?v=Sh6joEwiu8o

theoretical maxmium score in BOM is now officially over 110.000

Very interesting! I'm still catching up on posts from August and only reading this now, but this sounds like a large step from the previous maximums, right? The last one I remember is 100,899 (do the Razoff triple glitch in part 5, hit him five times in part 6 and be sure to get those 100 extra points from knocking him of the ball), and 101,299 was theoretically attainable by hitting him six times but I can't remember if anyone ever managed this. And then, you guys started improving combos in earlier sections, with the goal of finishing the level in part 5, ideally with a Razoff triple glitch, but I don't think anyone managed that either?

So, which final combo is supposed to yield that 110,000+ maximum? Doesn't sound like the Razoff triple glitch would be enough, or am I mistaken?

Anyway, based on the description you wrote for the video, looks like the odds of getting this AND a Razoff triple glitch in the same run are close to zero.

unlucky and unwise is he who risks entering my domain

worth +1280 points.

Hitting Razoff like that and to getting him in a combo is frame-perfect. If you are looking to try this, you must nail the technique to correctly jump down, otherwise you will waste your time with attempts that looked very close but were never going to make it (like I have). Study the jumping point and the trajectory. If you get twice +100 points the majority of the time, you are doing very well, all you need then is to wait for the golden attempt in which you pressed shoot in the correct frame.

[Cut]

For all of you who had "reaching 57.000 points in Fairy Council" as a new years resolution: I have good news!

https://www.youtube.com/watch?v=VyLo0IVt6ik

So, which final combo is supposed to yield that 110,000+ maximum? Doesn't sound like the Razoff triple glitch would be enough, or am I mistaken?

The triple glitch is indeed enough. Maz' walkthrough already yields more than 101k with only a double glitch. There is an incredibly stupid combo in BOM1, there is +840 points in BOM4, so combine that with the new green gem combo and a triple glitch, which is worth 6000 points, and you get 110.000 in total.

Anyway, based on the description you wrote for the video, looks like the odds of getting this AND a Razoff triple glitch in the same run are close to zero. :fou2

The combo is incredibly technical, and it will take a couple of days to learn and understand the way that actually works. With more practice, I am confident that you would be able to get down to around 30 minutes per successful attempt, maybe even 15 minutes or less after some time (one attempt is exactly one minute, including the time it takes to reset). That is not great, but compared to the green gem combo in DOTK3 for example, it's also not the worst thing this game has ever seen.

[Hunchman801]

For all of you who had "reaching 57.000 points in Fairy Council" as a new years resolution: I have good news!

https://www.youtube.com/watch?v=VyLo0IVt6ik

So you're taking both rooms in combo? How do you get to the second one without completing the first one though? Don't you need to find André to open the door?

The triple glitch is indeed enough. Maz' walkthrough already yields more than 101k with only a double glitch. There is an incredibly stupid combo in BOM1, there is +840 points in BOM4, so combine that with the new green gem combo and a triple glitch, which is worth 6000 points, and you get 110.000 in total.

I get that there's a shit ton of points now (I watched that combo in part 1 by the way and I have absolutely no idea what's going on there ) but my question was about the final combo to get over 100,000: if it's the triple glitch and it yields 6000, then surely it won't be enough to reach 110,000?

[Maz]

I think there's a bit of a misunderstanding going on here

"The triple glitch yields 6.000 points" means that it gains 6.000 points compared to the double glitch, not that the final combo is "only" worth 6.000 Points; the final combo is worth at least 15.320 points (I haven't tested this myself, but in theory, there might be a way to get an even bigger combo). This is not only enough for 110k, it also means that we are still several thousands of points short of a proper maximum in BOM5.

Regarding FC4, there's this to skip the first room, and this to skip the second room. This whole part is a glitch-y mess

[Hunchman801]

Thanks for the clarification, this makes perfect sense now! Clearly, the levels where every point counts are those where people are finding more and more ways to break the game.

[Cut]

We're doing a challenge on the scoring Discord, but you don't have to join to participate!

The rules are: Try to achieve as many points as possible in the first two areas of LOTLD 2 (the outer area with the Hoodlums and the room that houses the Matuvu and the Tribelle), including the Grim Keeper but nothing after it, so not the secret room or anything on the upper platform. You can only use one charge of the Vortex and the Lockjaw each. Glitches besides PCM and IPG are allowed. If you want to keep your ideas to yourself, upload an unlisted video to Youtube and share it after the deadline is ended.

The challenge ends on Friday, January 26th at midnight GMT.

[hoodlumsworld]

Nice challenge idea, maybe those could be synced with COTM?