Experienced Forum User, Published Author, Skilled player
(1892)
Joined: 11/14/2009
Posts: 1349
Location: 𝔐𝔞𝔤𝑦𝔞𝔯
mklip2001 wrote:
Impressive manipulation for a really clumsy-looking game. Indy's running stride looks really ridiculous. This makes level 1, in particular, look fairly imprecise, but it doesn't seem like the game allows better control.
I guess this is too short to be boring, but I'm not sure it's really that entertaining either (especially when the last level is just a luck-manipulated puzzle that lasts about 1/3 of the movie length!). I'm not sure whether I'm voting Yes or Meh yet.
I know what you are saying. Level 1 is actually 118 frames faster than Randil's original WIP, I optimized it by performing kills as fast as possible, better positioning, better movement, slight route change and "killing more enemy on the way instead of hunting them down at the final door". Regarding control... well, this game doesn't have it. If you hold L then hold U (for example moving up the ladder) you will still go to left if you didn't switched at 4th frame. Also you have to stop your movement to go up on a ladder. Now combining these just for the "ladder" scenario:
you need a good position to be able to use the ladder -> this needs a correct series of D's in a row to stop on the top of the ladder -> you have to time this perfectly, otherwise you will duck on the top of the ladder.
Regarding entertainment: I forgot that I did some "far" kills as much as I could (without wasting any frame) on stage 1. Not to mention the stage 3 kick madness.
edit: for those who didn't know this game: enemies need many-many kicks and punches to die. There's 3 techniques which does instakill: 1. running kick 2. whipping 3. spawnkill. That's why I used a lot of running kicks when I had to move along the way, and use whipping when I have to hold my position. The stage 3 kicking madness is actually "spawnkills" with rather crazy visuals (Indiana is nowhere to the enemy)
PhD in TASing 🎓 speedrun enthusiast ❤🚷🔥 white hat hacker ▓ black box tester ░ censorships and rules...
Experienced Forum User, Published Author, Skilled player
(1892)
Joined: 11/14/2009
Posts: 1349
Location: 𝔐𝔞𝔤𝑦𝔞𝔯
feos wrote:
will wait for possible movie editing. If it won't be edited, I'll do HD.
Since I couldn't find any better strategies to get the faster luck manipulation (spent many days just for this), the only thing what I could (but don't plan) is to implement a bit of entertainment. There's a really few spots where I could do anything which doesn't slows me through the way.
So movie editing depends on the audience/viewers vote/notes.
PhD in TASing 🎓 speedrun enthusiast ❤🚷🔥 white hat hacker ▓ black box tester ░ censorships and rules...
Experienced Forum User, Published Author, Skilled player
(1892)
Joined: 11/14/2009
Posts: 1349
Location: 𝔐𝔞𝔤𝑦𝔞𝔯
Guys...
MESHUGGAH wrote:
The developers @ sun already made the Java(TM) Platform Micro Edition SDK 3.0 which comes with an emulator.
Also every bigger manufacturer has their own "emulator".
Just like Nokia, Sony Ericsson, Samsung (their dev site is down at the moment for some hours), etc.[/url]
PhD in TASing 🎓 speedrun enthusiast ❤🚷🔥 white hat hacker ▓ black box tester ░ censorships and rules...
Experienced Forum User, Published Author, Skilled player
(1892)
Joined: 11/14/2009
Posts: 1349
Location: 𝔐𝔞𝔤𝑦𝔞𝔯
Welcome to TasVideos! Computer progamming skills is a good thing, especially when it comes to debug a game to perform a super-godlike speedrun. Also based on self-experience, Asperger-syndrome come in handy when it's about to think about various possibilities/tricks/glitches/ideas.
PhD in TASing 🎓 speedrun enthusiast ❤🚷🔥 white hat hacker ▓ black box tester ░ censorships and rules...
Experienced Forum User, Published Author, Skilled player
(1892)
Joined: 11/14/2009
Posts: 1349
Location: 𝔐𝔞𝔤𝑦𝔞𝔯
klmz wrote:
I doubt if this kind of emulators simulators can be "accurate as the real hardware" to any degree, since the "J2ME Virtual Machine" on which the code runs is not real hardware at all.
My post in this thread is outdated, at least refer to the latest Post #291342. My latest investigation shows that it's possible to make ~95% accuracy emulation (just like for NES), however this kind of tasing can be abusable as hell (custom resolutions, no source of valid ROM checksums, and many other things that already mentioned by you, others and me.).
PhD in TASing 🎓 speedrun enthusiast ❤🚷🔥 white hat hacker ▓ black box tester ░ censorships and rules...
Experienced Forum User, Published Author, Skilled player
(1892)
Joined: 11/14/2009
Posts: 1349
Location: 𝔐𝔞𝔤𝑦𝔞𝔯
DarkKobold wrote:
At this point, as you mentioned, people only use lua scripts for either short segments, displaying pertinent behind-the-scenes information, or brute-forcing certain parts. The reason is efficiency and ease. While an interesting concept, it is far too overwhelming, and is pretty much only theory.
I'm totally agree with this, something like this what I wanted to write down but I lack of proper expressing myself (just like this sentence).
PhD in TASing 🎓 speedrun enthusiast ❤🚷🔥 white hat hacker ▓ black box tester ░ censorships and rules...
Experienced Forum User, Published Author, Skilled player
(1892)
Joined: 11/14/2009
Posts: 1349
Location: 𝔐𝔞𝔤𝑦𝔞𝔯
scubed wrote:
AI stands for artificial intelligence.
Previously, people have thought about trying to make bigger, more complicated AIs
to complete the games.
But, those would be very big and complicated.
Likely, they would be far from the fastest known times
This is a typo, right? I don't understand, why would a complicated (I should say "more precise") AI be slower than a simple AI that checks fewer conditions.
I think I actually know what you are trying to say and what are you aiming for, but this really, REALLY depends on the situation. Various games needs vastly different approach. Some games are easy to debug, some games only needs some memory checking, there's also where only a simple LUA script needed that shows hitbox. I think that your kind of technique only required on games that has nearly the same physics and gameplay objects throughout the game. But I still didn't get the "higher level" stuff, except if it was what arukAdo explained me (higher level language, not higher level movies).
PhD in TASing 🎓 speedrun enthusiast ❤🚷🔥 white hat hacker ▓ black box tester ░ censorships and rules...
Experienced Forum User, Published Author, Skilled player
(1892)
Joined: 11/14/2009
Posts: 1349
Location: 𝔐𝔞𝔤𝑦𝔞𝔯
I don't understand what's your point. You are talking about making LUA scripts that uses artificial AI? Please explain it. Otherwise: Use the search function.
PhD in TASing 🎓 speedrun enthusiast ❤🚷🔥 white hat hacker ▓ black box tester ░ censorships and rules...
Experienced Forum User, Published Author, Skilled player
(1892)
Joined: 11/14/2009
Posts: 1349
Location: 𝔐𝔞𝔤𝑦𝔞𝔯
Please next time use the search function before creating new threads.
Thread #10324: Is it possible to TAS jar files (J2ME) (last post by me).
I actually did a bit more research after my latest post. What I found out already:
There's a huge difference between many phones, because of vastly different versions of KVM's and memory, cpu speed, screen size, implemented APIs, no frame based input, much harder TASing sessions (since JARs contains the source code, the TASers should MUST decompile them first and requires much more planning cause they have the ability to do this. Remember, we aiming for godlike playing, not trial and error).
I will try to contact one of the site admins later to get a solution for this, since I also think there could be many good additions to the site.
PhD in TASing 🎓 speedrun enthusiast ❤🚷🔥 white hat hacker ▓ black box tester ░ censorships and rules...
Experienced Forum User, Published Author, Skilled player
(1892)
Joined: 11/14/2009
Posts: 1349
Location: 𝔐𝔞𝔤𝑦𝔞𝔯
quvu wrote:
At least wallstrafe as its called in HL have different angle than the cs-wallrub. The original wallstrafe was "removed" from counterstrike back in 1.3 or so.
And I must admit im not sure about the details of standups.
In CS you loose height after a few bhops? However, Im pretty sure that is not the case in HL.
You are right, I forget that. Wallstrafe is different in HL.
quvu wrote:
I just have a "+moveright;+right..." script going constantly. And have an ascending/descending alias to manually alter the yawspeed (with 100 differentials maybe) as the velocity changes.
I'd be glad to hear if you find more info.
As I said about keybd_events: http://pastebin.com/KJxgPB8p. I didn't debugged HL so that's why I didn't implemented the actual framerate as a constant. Also this was only an experiment, it lacks the high frequency timer too and the dynamically changing yawspeed.
PhD in TASing 🎓 speedrun enthusiast ❤🚷🔥 white hat hacker ▓ black box tester ░ censorships and rules...
Experienced Forum User, Published Author, Skilled player
(1892)
Joined: 11/14/2009
Posts: 1349
Location: 𝔐𝔞𝔤𝑦𝔞𝔯
quvu wrote:
Hows it going meshuggah? anything new?
I mostly do/did other stuffs (rockin kat's @ DTC4, battletoads).
quvu wrote:
lol, i almost feel offended getting linked to a tuturial, i'm not that noob. Most of that other stuff is for CS only anyway (wallrub, standup bh ect.)
And using cj start in HL would only be applicable to get better distance/pattern to an obstacle i think.
Both wallrub and standup works in HL. Everything which shown there works in HL too. Counter-Strike only modified the maximum player speed and maybe the sideway speeds. You can test it by putting AMXMOD and downloading any ljstats for CS, it will show the correct values for HL too (I tested it long time ago, but if it doesn't work, I can write a plugin for you). I submitted a demo on SDA which shows the ability to jumpbug (which I did without tool assisting), IIRC it showed the speed values.
quvu wrote:
And in fact, I do surf in the pipes a bit, but the slope is not high enough, so surfing all way through would be slower. And I'm not sure if its even any slowdown in there.
It's hard to surf in 10aa (original, default value that both HL and CS uses by standard). But it's can be used for avoiding damage by surfing on even very small slopes to decrease the speed. Also it's faster to get maximum speed (if it's very long and you can move downward while surfing, therefore it's depends on the surface/part).
quvu wrote:
Now to the interesting part, frame by frame strafing.
Of all the clips on youtube of this kind of stuff, 99% is done with 100airaccelerate.
I've tried this and its not very sensitive about the yawspeed. Compared to default 10airaccelerate, which is extremely sensitive.
I would need some kind of yawspeed/ups ratio to make this perfect.
Yes, but it's works on 10aa too. The main difference between 10aa and 100aa is that you have to be perfect, otherwise you lose speed if
1. moving the mouse in the wrong way (pressing A and mouse to right/pressing D and mouse to left)
2. moving too fast with the mouse
3. pressing forward longer than it required (you should immediatly stop W when you jumped)
4. ducking and strafing (will decrease dramatically)
I still didn't calculated the correct ratio, but I'm sure that it's depends on the actual speed. You can only do small strafes when you have small speed, but you can do "higher" strafes (moving the mouse much more to the left/right) if you have high speed. Therefore, you have to dinamically increase the "distance" of your mouse movements as the speed increases.
quvu wrote:
Any ideas how to calculate this?
I already tried to calculate the HL's ratio of this, but it's ridiculous. The game does input polling many more times (more than 3 times) per frame, IIRC it polls mouse more often than the keyboard. I also couldn't have luck with the cl_yawspeed (getting out of sync because of the slightiest frame drop changed the turning). There are too many factors that changes the speed/distance, so I can't come up with correct values. I tried other techniques (making a plugin which reads an input file and interprets it from frame to frame, making an EXE that does some keybd_data) but all of them failed because of various frame drops.
PhD in TASing 🎓 speedrun enthusiast ❤🚷🔥 white hat hacker ▓ black box tester ░ censorships and rules...
Experienced Forum User, Published Author, Skilled player
(1892)
Joined: 11/14/2009
Posts: 1349
Location: 𝔐𝔞𝔤𝑦𝔞𝔯
Kumquat wrote:
To my knowledge, every boss takes one damage from every kind of attack, but thrown objects and the hammer weapon do three damage to normal enemies.
My team threw the first miniboss's eggs back at her to reduce lag (they would bounce across the screen otherwise). Grabbing the candle and throwing it at the boulder in channel 5 might actually waste time; I only did that because we were out of time in the contest and I thought it looked funny.
Incidentally, you can also pick up power ups (the hearts, money bags, and extra lives) and throw them at enemies.
Our run should drop below 16 minutes with all the optimizations we found at the end, so look forward to seeing that very soon!
Nice! I'm pretty sure that even sub 15 (maybe sub 14 with correct RNG calls) is possible with correct lag optimizing and throwing in every improvements/tricks/techniques.
PhD in TASing 🎓 speedrun enthusiast ❤🚷🔥 white hat hacker ▓ black box tester ░ censorships and rules...
Experienced Forum User, Published Author, Skilled player
(1892)
Joined: 11/14/2009
Posts: 1349
Location: 𝔐𝔞𝔤𝑦𝔞𝔯
AnS wrote:
MESHUGGAH wrote:
bad programming
Please stop throwing words you cannot prove.
Just count the number of subroutines on lag-frames and on non lag-frames. I don't even mention their extended length. Also of course I'm not throwing anything, I debugged this A LOT.
PhD in TASing 🎓 speedrun enthusiast ❤🚷🔥 white hat hacker ▓ black box tester ░ censorships and rules...
Experienced Forum User, Published Author, Skilled player
(1892)
Joined: 11/14/2009
Posts: 1349
Location: 𝔐𝔞𝔤𝑦𝔞𝔯
Truncated wrote:
... When managing lag, you do mostly random actions until the lag decreases...
Most of the lags are due to bad programming, also you don't need random actions just simply reducing/extending various inputs or inserting various input that doesn't change current state (x/y pos, subroutine call order, their length etc). I didn't debugged this too seriously only investigated a bit to verify it's an emulator issue.
PhD in TASing 🎓 speedrun enthusiast ❤🚷🔥 white hat hacker ▓ black box tester ░ censorships and rules...
Experienced Forum User, Published Author, Skilled player
(1892)
Joined: 11/14/2009
Posts: 1349
Location: 𝔐𝔞𝔤𝑦𝔞𝔯
edit: Thank you for the encode feos =)
Offtopic:
AnS wrote:
MESHUGGAH wrote:
FCEUX 2.1.4, FCEUX 2.1.5 and FCEUX 2.1.6 shares the same code that used for emulating games, it didn't changed.
Actually id did! And that's why I was surprised that Rockin' Kats sync on both 2.1.4 and 2.1.5.
Try that WIP of Battletoads you mentioned, it syncs on 2.1.4 and doesn't sync on 2.1.5, because in 2.1.5 there's nasty bug caused by mapper revision.
This bug was fixed in 2.1.6, so Battletoads (and probably all other runs made for 2.1.4) will sync in 2.1.6 (while not necessary syncing in 2.1.5).
Well actually I synced feos' Battletoads WIP 8 on both 2.1.5 and 2.1.6 (my improvement is made on 2.1.6 fceux/taseditor).
AnS wrote:
MESHUGGAH wrote:
...both games shares these kind of issues (cpu cycle dependency, and RTI in input polls). And I improved his upcoming WIP with 68 frames (the warping section) after a bit of trace log comparing...
..Nice job with improvement, but it's still far from what could be achieved by reverse-engineering that stuff I addressed there.
Huh, but where did you find RTI in input polls there?? I see standard poll procedure, nothing fancy. As for cpu cycle dependency, well, majority of NES games have this issue, don't blame RK for that.
I'm not at home currently, but I wrote it on ICQ to feos about that, ask him about subroutines and memory values that I posted to him. Or wait until tomorrow. Ask in PM, please don't make more offtopic posts as I did.
PhD in TASing 🎓 speedrun enthusiast ❤🚷🔥 white hat hacker ▓ black box tester ░ censorships and rules...
Experienced Forum User, Published Author, Skilled player
(1892)
Joined: 11/14/2009
Posts: 1349
Location: 𝔐𝔞𝔤𝑦𝔞𝔯
AnS wrote:
MESHUGGAH wrote:
It worths mention that the game has MANY emulator issues, which won't be fixed in the near future
Can you tell what are these issues in particular?
BTW, both contestants' runs sync well on 2.1.5 and 2.1.4a (and on 2.1.6 as well), using OldPPU of course.
FCEUX 2.1.4, FCEUX 2.1.5 and FCEUX 2.1.6 shares the same code that used for emulating games, it didn't changed. But newer versions which will have corrections/improvements/further developed code won't sync this run. The list of possible issues (didn't figured out really yet): incorrect PPU frame length (maybe 4 frames faster?), incorrect cpu cycles for sprite DMA copy (1 frames faster) and wrong RTI implementation (I'm not 100% sure about that).
There are several games that suffers from this game, check out this: Wiki: ConsoleVerificationTests.
Offtopic: last night feos contacted me about debugging Battletoads. I actually discovered that both games shares these kind of issues (cpu cycle dependency, and RTI in input polls). And I improved his upcoming WIP with 68 frames (the warping section) after a bit of trace log comparing and translating feos' notes from russian to english.
PhD in TASing 🎓 speedrun enthusiast ❤🚷🔥 white hat hacker ▓ black box tester ░ censorships and rules...
Experienced Forum User, Published Author, Skilled player
(1892)
Joined: 11/14/2009
Posts: 1349
Location: 𝔐𝔞𝔤𝑦𝔞𝔯
Aglar wrote:
Really nice RNG manipulation! I didn't even know there existed a series of 25 RNG numbers where only 3 of them were even (for the bull band), and you also got perfect RNG manipulation for the first ghost dog - everything without significant waiting in between the levels. If it's not too much to ask, I'd really like to know how the RNG planning process played out.
Thanks, it was also a result of teamworking.
The subroutine $FE60 is the "base RNG seed", and we watched the RNG calls @ 0x00F1.
PhD in TASing 🎓 speedrun enthusiast ❤🚷🔥 white hat hacker ▓ black box tester ░ censorships and rules...
Experienced Forum User, Published Author, Skilled player
(1892)
Joined: 11/14/2009
Posts: 1349
Location: 𝔐𝔞𝔤𝑦𝔞𝔯
Since our team finished the DTC4 submission, I have time to handle this.
I made a game resource page Wiki: GameResources/NES/IndianaJonesLastCrusade to avoid editing posts for multiple times.
Randil's WIP has 16 sideways on the holy grail map layout. Waiting for 4 frames at the title screen (and reoptimizing the syncs coming from different enemy spawn (very few frames lost)) changes the holy grail map layout into 11 sideways.
Comparison at holy grail map layout screen
Randil's WIP: 7644 frames / Randil's WIP 4 frame wait: 7665 frames (+21 frames)
edit: 97 RNG calls, 1 + 2001 for map generation / 106 RNG calls, 1 + 1974 for map generation.
However I will debug the $8BB9, so I can search the best RNG for minimizing the number of sideways.
edit: no need for debug, I made a lua script that shows the RNG calls and dumps the RNG numbers and another to make a screenshot with these values (thanks to Wiki: Kirkq's generous help!!). The state I used for the screenshot dumping was 7595 frames into Randil's WIP.
After some manual screenshot watching (took me some minutes): there's 9 layout * 6 ending tile = 54 possible maps can be generated by the game depending on which RNG it started. The ending tiles are always in the same order (heart, something, maybe an a, omega, bird, X, cup). We aiming for the fewest sideways. Now finding the fastest one in progress.
edit: Number of sideways of the maps: 15/12/11/16/7/11/11/11/4. No need to mention, we go for Map 9 which ends on a bird, which needs 90 18 more rng cycles than Randil's WIP. This sounds good, because we can adapt this to the title screen.
edit: the RNG cycles drives me crazy. Applying wait on title screen (to advance the rng cycle) changes the increments @ stage 2 Find Dad. This means, simply just waiting to get the exact 116 calls is "impossible" (without wasting many time). Of course it's possible with doing some route changes. Here's the most easiest (notice I used Randil's WIP): http://dehacked.2y.net/microstorage.php/info/98946327/FINALTAS.fm2
Also note that 116 rng calls is the most close for the "fastest route" (the earlier one is needs too few calls, the next one needs more calls).
Also there's a faster route at find dad which I didn't include this time (depending actual RNG, the faster route changes the RNG calls). I will throw out an update after I came out with the super optimized godlike TAS.
PhD in TASing 🎓 speedrun enthusiast ❤🚷🔥 white hat hacker ▓ black box tester ░ censorships and rules...
Experienced Forum User, Published Author, Skilled player
(1892)
Joined: 11/14/2009
Posts: 1349
Location: 𝔐𝔞𝔤𝑦𝔞𝔯
feos wrote:
All platforms can have "press right to win" games, and all can have interesting strategic (not meaning genre) games. If Rockin' Kats were dull straightforward game it would be frameperfected long ago.
The only competition is who can make more deliberately abuse the emulator, I never heard of something like this should be called as strategy.
PhD in TASing 🎓 speedrun enthusiast ❤🚷🔥 white hat hacker ▓ black box tester ░ censorships and rules...
However I will debug the $8BB9, so I can search the best RNG for minimizing the number of sideways.
edit: no need for debug, I made a lua script that shows the RNG calls and dumps the RNG numbers and another to make a screenshot with these values (thanks to Wiki: Kirkq's generous help!!). The state I used for the screenshot dumping was 7595 frames into Randil's WIP.
After some manual screenshot watching (took me some minutes): there's 9 layout * 6 ending tile = 54 possible maps can be generated by the game depending on which RNG it started. The ending tiles are always in the same order (heart, something, maybe an a, omega, bird, X, cup). We aiming for the fewest sideways. Now finding the fastest one in progress.
edit: Number of sideways of the maps: 15/12/11/16/7/11/11/11/4. No need to mention, we go for Map 9 which ends on a bird, which needs 
edit: the RNG cycles drives me crazy. Applying wait on title screen (to advance the rng cycle) changes the increments @ stage 2 Find Dad. This means, simply just waiting to get the exact 116 calls is "impossible" (without wasting many time). Of course it's possible with doing some route changes. Here's the most easiest (notice I used Randil's WIP): http://dehacked.2y.net/microstorage.php/info/98946327/FINALTAS.fm2
Also note that 116 rng calls is the most close for the "fastest route" (the earlier one is needs too few calls, the next one needs more calls).
Also there's a faster route at find dad which I didn't include this time (depending actual RNG, the faster route changes the RNG calls). I will throw out an update after I came out with the super optimized godlike TAS.