if i remember correctly, he joins already after the greenhill liberation. it isn't dependent on how long you wait, you just have to reach a certain point in the game.
So if you talk to him for the first after the greenhill liberation and ask him about money, you won't be able to recruit him?? From my test, he just seem to come back after a long time, just like kinisson, richmond(when asking for investigation) and maybe other
nfq wrote:
by the way, it might be faster and cheaper to use mostly only magic in the TAS, because sharpening weapons takes a lot of time and money.
Look like there are an even better way:
http://www.youtube.com/watch?v=ZWSg51Nm2Kk&feature=channel_page
it took me a long time to figure out... but this is actually 100% legit, since the probability is very low and I feel like it work on every monster/boss :D
At least, I can't use this combo on every battle(most time nanami take one slot for herself in the party, etc..). Thought I prefer watch a suikoden run like this, than a boring kindness rune glitch on every battle.
Should I go foward on this?
I thought Hans was hanging around Two River City and that he would appear after Chaco had the player's money stolen.
yeah there a lot of "pop up" thing like this that make the route quite complicated.. I prefer to check every aera if there something new, when the "event" of the game change.
nfq wrote:
i don't know if i said this earlier, but you should have asked Hans to pay you 300,000 potch to join you, because money can be important in this game. that way you wouldn't have to collect other things that give you money and you could have saved some time.
in fact you said more something like this :p
nfq wrote:
money isn't really important in this game.
Seriously, I prefer to don't ask anything to hans, because I know that it may took like 3-4 hour(in-game) to come back with money and if I manage to be eought fast to make my way to matilda liberation, before his come back... this could block me to get the best ending.
Actually, with the lua tool I've write some different kind of bot and now, I was interrested to brute forcing the mysterious RNG for critical hit..
here's a script that write all address and value in a file, every time that jowy deal a critical hit (at tenzan pass/beggining)
==> http://lua.pastey.net/120798-56m2
and the result...
==> http://omploader.org/vMjR0bg/jowy-criticalhit-all-address-value.zip
I thought with a diff program, I could get some way to find this RNG... but look like that this is a dead end, since there so many value that change.
I think I should try to create some kind of database and doing statistic from them, but I have no idea how to proceed.
okay. Honestly, I don't feel suited for this run... I fell like I need more knowledge that I don't have. I'll put this to the side for now, and go acquire that knowledge. If anyone wants it, they can have it.
I think you shouldn't fear about all the "knowledge stuff"... in fact this it's almost useless to getting "frameperfect" on the first run on a such complicated game, because my thought is that this is absolutly IMPOSSIBLE to get the best route, right at the first attempt. Even an expert player can't do this.
Go use the turbo, forget about all the lua stuff for now. Maybe just using frame by frame feature is enought... simply for getting a gameplay that look "enought TAS".
This isn't like that website was enought asshole to reject a such long game, while your gameplay look good and it beat any other non-TAS run, even if not 'perfect' at the first publication.
fuck your run is good, I'm sure that if it would be an avgn review... everyone would shoot a YES
Seriously, you might want to edit your submission to add these important information.
I liked your run and the ending music was quite good... ^_^
Recently, I've been very interested about writing/learning a brute force bot with Lua and after spended many hours around this, I understand that there mainly 2 way to get the right result:
1. Random input : example (thanks to arukado for pointing me this)
On each frame, the button have to win a test on a rand fonction, so basicly there should be about 50 % that a button is pressed or not... It should be very usefull for luck manipulation or for trying to optimize something on few frame.
2. Every Possibility Input: (I'm sure that there already that someone already think about this before myself, but I din't found any script..)
This Lua script would probably require a lot more attempt(and time!!) than the random script... but if it have the right algorithm, It should actually test every single button possibility for each frame. Thought, at least this isn't a never ending "random attempt" and after doing every possibility to reach the "predefined target"of the script... this should be totaly frame perfect. (maybe there could be *still* an some way to do some other runs more entertaining with the same framecount, but this is an another story).
Applying this from the power ON of a game with the right "memory target" and a "frame limit" number to help the bot... and you should get the best time, right???
This seem doing perfect sense and I've already trying to write the algorithm this... but after two attempt, I feel like I probably use the right way, so I'm asking you for helping me.(with pcsx)
But first, here is some statistic that could scare you:
Possibility for one frame with only one psx controller(L3/R3 enable, but not analog stick):
1 - idle
16 - 1button pressed
120 - 2button pressed
560 - 3button pressed
464 - 4button pressed
364 - 5button pressed
286 - 6button pressed
220 - 7button pressed
165 - 8button pressed
220 - 9button pressed
286 - 10button pressed
364 - 11button pressed
464 - 12button pressed
560 - 13button pressed
120 - 14button pressed
16 - 15button pressed
1 - 16 button pressed
total: on each frame there about 4227 button combination
(thought maybe these number are'nt exaclty right, but this is a lot for sure)
on a long game(large frame limit), that means a lot of possibility... and might took even some week of calculation, but I'm sure that some game would deserve it as well :)
Maybe that lua isn't the most appropriate language for this algorithm, but I'm sure that we can found a way to do it.
Here is my best attempt for now, but I suppose that I can do better
==> http://lua.pastey.net/120311-1q0h
You need to save something before loading a temp savestate. :P
oops... ok with the saving fonction that's work and rocks, now :)
btw, where is saved this "temp savestate" ? does there are a way to get it on our hard disk? (once again, after many brute force test... I suppose that this is best way to get a backup of his best attempt, right?)
select l3 r3 start up right down left l2 r2 l1 r1 triangle circle x square
I din't know that pcsx supported r3/l3... I might try this if there some interesthing glitch with this, but what about the analog stick??.
Also,
if I do something like:
local input = {}
input = joypad.read(1);
how is it stored in the array/memory of input[]?
I might want to do something like:
input = joypad.read(1);
input[4] = true --Pressing start with the right nil/true byte?
joypad.set(1, input);
getting these nil/true number insthead of these predefinined string in the array could help me for a brute force bot... (but, I'm also just simply curious how this is stored)
another thing.. I'm very unsure how I can do a savestate and reload it...
local i = 0
test = savestate.create();
while true do
pcsx.frameadvance();
if i > 600 then -- loading savestate after 600 frames
savestate.load(test);
i = 0
end
i = i + 1;
gui.text(0,50, "frame : " .. i)
end
I was looking for something like aegisub, at first I thought that it supported only the ssa format...(ssa is a cool format, some feature like karaoke and color can be interesthing) .. Only problem with this format is that for some raeson, mplayer/mencoder have to use the ass librairy and I still din't found how getting a "square background" with ssa and this library enable
anyway looks like that .sub format is definitly more basic than anything else... so let's just getting the job done with the shinydoofy solution:
mkv 59.05mb with vorbis:
==> http://www.mediafire.com/download.php?g5nmwm20uhj (still under 4mb/minute)
althought, at this point... I feel like these luascript might help to save few second on this run... but since this still very laggy, any improvement would look much better when pcsx-reload will get some TASing feature.....
edith: youtube: http://www.youtube.com/watch?v=1SPrQLNFIlY
Still stunned.. are you the god of C ? :D
now some question...
what are the correct button mapping for the joypad??
I've been trying to guess by looking here at "static const char *button_mappings[]"
button with d-pad seem working fine
joypad.set(1,{["up"]=true})
but not this one:
joypad.set(1,{["x"]=true})
damn... maybe I just got everything wrong, but I think that if you could write an "example lua script" for showing how work each command... it would be a big help for everyone !! (thought the command that work by the same way as other command probably doesn't need any example).
in anycase thanks mz :)
Obviously I'm doing something wrong. Most games just freeze right after the logo comes up. I can't even reach the title screen. The only game I'm having no problems with is Symphony of the Night.
For most games:
the Sony logo plays just fine
then I hear the Playstation logo, but it's a black screen
After that nothing happens.
I am using v 0.1.0 as well as the plugins suggested for use.
Did you downloaded the psx bios(google for more information) ?
Also, after that loaded the iso file with the following plugin.. make sure to choice "Run CD Thought Bios"
---
@mz I saw that you just commited a very large file about adding lua script feature and I'm about confusing about how far this tool can go...
This quite a noob question(hope to don't create any stupid debate), but do you think that it would be possible to write a script that would be able to use brute force from the power ON of a game(when pcsx-RR would be on his final version or something)... and then the script would be able generate every single input possibility on each frame and perform a test comparaison with an(or more) address(example looking on each frame for the address that contain the font "The end"/or animation address, from a specific game). Then the bot would run for few day to test every possibility. When he found a way to get thought the ending.. the bot could even try to optimize hiself, etc... some other feature like trying to switch disc/system reset on every frame would probably be removed from these brute force script to avoid any annoying problem, as well thought.
Is it how far these lua feature could go, or there something that doesn't work with my logic?
edith : by making some search on the forum, I've finaly found this old post and I feel like this would be possible with the correct algorithm/programming...
does this is the last wip ?
Seriously I think these project become so big, that getting a svn depot would the best... at least people aren't (a bit less) closed of new progress and the "knowledge guys" can look up how are thing and maybe commit one shoot or two, for helping.
I made my own test and look like your supposed to have a little "cutscene" when the airbuster is defeated(the camera zoom on the robot face for few second)
.. so you have to manipulate the game enought to see "this cutscene", else the game would think that you din't even beat him, I suppose.
Is there a different port of this game that would make for a better TAS?
There is ONE on sms and it seem to work fine on dega
But at this point I feel like the sir VG idea is very appropriate
edith: as long as black spy win this time!!!
I just tested the game and I remembered that doing something like A followed by B should make a little jump in a fight... I think there were even a way to do a backflip too.
Anyway, I would love that someone try this game again.
I made some experience with left+right.. and this give some weird result and sometime getting stuck in a wall.
I'm sure that this game have even more potential, but I'm working with an another TAS now...
As for mupen64plus: getting the last SVN revision to compile and run has been an exercise in pain. I did manage to get it to compile, but it crashed when loading Paper Mario, presumably because it did not have valid sound or RSP plugins loaded, and compiling the plugins is a continuation of said pain. It seems they have put off Windows compatibility for now. The best bet is to patch 1.3, as AFAIK it should compile in Windows (it even has an official release of sorts). Maybe someone can port the mupen64-rr patch to it.
First of all, good job! I've tested this previously on cygwin and getting some libs working(like SDL-image with the "correct" libpng and some other) is very long and painfull... from what I heard on irc, you manage to make qt4 working insthead of gtk??
Does some other rom are working on your build? (you may want to try the no-sound plugin)
==> http://suikoden2.googlecode.com/files/suiko2-greenhill_over.zip
-Greenhill quest is over
-I recruited Mukumuku and Meg+Gadjet
-I started the Hix&Tengaar sidequest
-Taki and tetsu recruited!
-Tai ho and yam koo recruited!!
-Stalion, Richmond and Clive recruited!!!
already 760K frame... it's about 3h26 time in game/ 3h:31 realtime.
I was confuse about bisqwit note at the bottom.. anyway I suppose that's now is about 30 second + 36 frames.
So, cool improvement and enjoyable graphic glitch at some boss stage. Nice one!
Now what about the the bonus stage? does there are something after completing until 15-10 ?
I don't know if this what you intended, but you still asking something ridiculous insthead of explaining what you want... you should see this. (use (s)mplayer to play this frame by frame). Does cloud is supposed to walk on this specifc frame?
edith: my best advice would be that you try some pcsx config(maybe the keyboard insthead of a joystick) and you try to see how to avoid laggy frame