1 2 3 4 5 6
Posted: 10/27/2013 12:24 AM
Quote
Tristal
Joined: 7/10/2007
Posts: 280
She doesn't especially need levels to proc the Para kn, though. Might be more useful to have Midia die sometimes so the rest of the party get more XP. Not necessarily going out of your way to do so, just not minding if she does at times.
Posted: 10/27/2013 3:05 AM
Quote
janus
Active player (285)
Joined: 9/1/2008
Posts: 900
really? I would have to test that. http://dehacked.2y.net/microstorage.php/info/781167373/Paladin%27s%20Quest%20%28U%29.smv back at the equator with 13k less frames. I would need to test if not giving Midia magic (she would have strength up so Chenzi can attack Guardian more efficiently) is worth it. If not, I would get the magic in Misuto
Posted: 10/27/2013 3:43 PM
Quote
Tristal
Joined: 7/10/2007
Posts: 280
Why did you lose frames between the last two WIPs? Or am I misunderstanding "saved 13k in this segment" instead of 13k overall
Posted: 10/27/2013 8:56 PM
Quote
janus
Active player (285)
Joined: 9/1/2008
Posts: 900
the loss in frames is due to differences in combats. I have to find the ideal level of magic/experience to defeat Kaymat http://dehacked.2y.net/microstorage.php/info/1782619014/Paladin%27s%20Quest%20%28U%29.smv now about a 16k frames advance in Karon. Chenzi is at maximum fire (127), and sp is not far behind.
Posted: 10/27/2013 11:05 PM
Quote
Tristal
Joined: 7/10/2007
Posts: 280
Having to walk around people in Karon looks really bad; does this at least avoid a battle by abusing the 0 boundary when you leave town? Otherwise, I would definitely try to find a way to manipulate the townsfolk movements.
Posted: 10/28/2013 1:13 AM
Quote
janus
Active player (285)
Joined: 9/1/2008
Posts: 900
you're probably talking about Conshiuto (Karon is the city of the specters). Yes, I need to review that. It's easy coming in; but going out is hell. I will need to work on it on the rerun. Other than that, how's the run?
Posted: 10/28/2013 3:11 AM
Quote
Tristal
Joined: 7/10/2007
Posts: 280
It'll be easier for me to make comments on the route strategy when I see the final fights. As you're making the TAS you get to take things one frame at a time, but as a route planner I sort of need to work backwards from the end. I still worry about how much we need to level spells, I get the feeling we don't need to do that much. The only differences I've caught so far are Fritz's fight, killing of Midia in the Jarmlx fight and NPC pathing in Conshiuto. I'm sure there are improvable fights throughout the run, it's just very tough to find everything. Actually, I think a lot of the fights themselves are fine, it's more about what ideal monster formation we should be pursuing. There's the added problem of health management here too. EDIT: Clarified wording
Posted: 10/28/2013 6:53 AM
Quote
TheAxeMan
Ambassador, Experienced player (709)
Joined: 7/17/2004
Posts: 985
Location: The FLOATING CASTLE
To help Janus out I have disassembled the encounter algorithm. It's complicated but it should be possible to manipulate. There are several parts so I'll try to break it down a bit here. This happens at the end of a battle when the screen collapses to black. There is something different for setting steps on leaving town though it might use some of this code. First the addresses involved. I'm not sure exactly how snes assembly addressing works but it seems that the high byte of the address can be 7e or whatever. I'll just show the low 2 bytes. 1868 is the steps counter that ultimately gets set here 1867 is an encounter rate that is set based on the encounter rate of an area. Typical values seem to be 7-12. 1859 is a counter that increments every frame 0023 is a one-byte battle RNG 0024 is a one-byte map RNG. It seems to change while on map but not in battle. 0025 is a two-byte encounter helper that is the most important factor 0048 is temporary storage for an adjustment value that tweaks the result a little The $0048 adjustment is loaded through a DMA call of some sort. I'm not so familiar with this aspect of SNES but that is what it looks like. Address is set in $4202 then 4 nops and a read from $4216. The high byte of the address is the encounter rate and the low byte is the $0024 RNG value. Battle RNG $0023 gets updated by adding the low 3 bits of $1859 and then multiplying that value by 5. There is no multiply so this is actually done with 2 left shifts and an add. This only affects the step calculation because the carry from the last add sticks around. Encounter helper $0025 gets updated by adding the low 3 bits of $1859 with the leftover carry bit if set. You can ignore the last paragraph and just know that this means adding a value from 0-8. Then take that result and multiply by 5. For this part everything is 16-bit operations so we are actually updating $0025 (low byte) and $0026 (high byte). Now we put everything together and get the steps. Take one byte from $0026 (the high byte of our two-byte value) and clear the high bit. Depending on that value (ranging from #$00 to #$7F) we will get radically different step counts: If result < #$33 then steps = (2 * rate) + adjustment - 2 If #$33 <= result < #$59 then steps = (3 * rate) + adjustment - 2 If #$59 <= result < #$65 then steps = (4 * rate) + adjustment - 2 If #$65 <= result < #$78 then steps = rate + adjustment - 2 If #$78 <= result then steps = adjustment + 1 (This usually means the next fight comes quickly!) Hex to decimal translation: #$33 = 51, #$59 = 89, #$65 = 101, #$78 = 120 So if you consider this value random, the chances of each case are 51/128, 28/128, 12/128, 19/128 and 7/128. Fairly low odds to get the highest number of steps but also low to get a battle right away. How to use this information? Dismiss the message at the end of each battle to manipulate the 0-8 addition to $0025. Because of how it works, the manipulation from one battle will actually make a difference 2-3 battles down the line when those bits shift to the high byte, making this an interesting mathematical problem. If possible you could get a few more steps by tweaking the value of $0024 on entering battle. Would need to figure out how to get the adjustment from that value though.
Posted: 10/28/2013 12:53 PM
Quote
Tristal
Joined: 7/10/2007
Posts: 280
>Because of how it works, the manipulation from one battle will actually make a difference 2-3 battles down the line when those bits shift to the high byte, making this an interesting mathematical problem. I was worried about this, but even ignoring that, being able to memory watch for best route will go a long way. Thanks so much for your help!
Posted: 10/28/2013 2:11 PM
Quote
janus
Active player (285)
Joined: 9/1/2008
Posts: 900
I was worried that influence in battles was long-term. let's try to influence it from now on, especially starting at the water temple, where I started not fighting regular monsters
Posted: 10/29/2013 2:37 AM
Quote
janus
Active player (285)
Joined: 9/1/2008
Posts: 900
http://dehacked.2y.net/microstorage.php/info/743170298/Paladin%27s%20Quest%20%28U%29.smv going back to my hometown. The road is not optimized, but I was able to save 23.3 k frames. Crit. on Noi Green actually takes more frames
Posted: 10/29/2013 2:56 PM
Quote
Tristal
Joined: 7/10/2007
Posts: 280
Is it worth it to have Chezni equip a weapon and use that in the Bubbley fight?
Posted: 10/29/2013 4:08 PM
Quote
janus
Active player (285)
Joined: 9/1/2008
Posts: 900
I need to calculate it, but I doubt it. His fire for this run is at maximum strength and does 88 damage, whereas sword damage in the previous run did at most 60. I also think I won't need to give Midia strength up since Fiery has it (against Noi Green). If she does need it, she can get in Ratursk of the past (at the same time as attack back). However, I need to calculate if she can withstand the final bosses, as she if 7 levels under
Posted: 10/30/2013 4:47 AM
Quote
janus
Active player (285)
Joined: 9/1/2008
Posts: 900
http://dehacked.2y.net/microstorage.php/info/960601646/Paladin%27s%20Quest%20%28U%29.smv completed the game with 28.1 k less frames. I'm now reviewing in detail to see areas of improvement (battles, mainly). I may not even need cards, since Fire gets strong so quickly
Posted: 10/30/2013 1:39 PM
Quote
Tristal
Joined: 7/10/2007
Posts: 280
Does the Defense Down spell work on Gabnid? You attack a lot, so it'd probably save time.
Posted: 10/30/2013 7:27 PM
Quote
janus
Active player (285)
Joined: 9/1/2008
Posts: 900
nope, it failed after at least 20 tries. I could try to do strength up, since I can get every magic I will in the city
Posted: 10/30/2013 9:30 PM
Quote
Tristal
Joined: 7/10/2007
Posts: 280
If it works on Msl it's worth it, don't know if it'd also be worth it for Mouth but probably. Almost certainly not for all three, the spell animation is a bit long
Posted: 10/30/2013 9:31 PM
Quote
Tristal
Joined: 7/10/2007
Posts: 280
Also, even if it takes you 100 frames to manipulate Defense Down hitting, that'd be worth it no doubt.
Posted: 10/31/2013 7:06 PM
Quote
janus
Active player (285)
Joined: 9/1/2008
Posts: 900
I might give it a try. http://dehacked.2y.net/microstorage.php/info/51437893/Paladin%27s%20Quest%20%28U%29.smv I saved Midia nearly 5k frames earlier. The Goblin fight is optimized, as the extra attack keeps a goblin from attacking (-50 frames), and the Alonzo fight was only 43 frames slower than the test run
Posted: 11/1/2013 3:17 AM
Quote
janus
Active player (285)
Joined: 9/1/2008
Posts: 900
http://dehacked.2y.net/microstorage.php/info/1885761556/Paladin%27s%20Quest%20%28U%29.smv back to Jurayn 15.7k frames earlier than the test run (20.7 from the complete run). I hope I'm not too weak; the fights were simply terrible
Posted: 11/1/2013 6:50 PM
Quote
Tristal
Joined: 7/10/2007
Posts: 280
You're right about those fights being terrible. They LOOK bad (even if they're faster) - but there are also noticeably less of them to begin with... Still, Purpbird attacking and failing a run (ouch!) make it seem like there's a lot of room for improvement here...
Posted: 11/2/2013 5:56 AM
Quote
TheAxeMan
Ambassador, Experienced player (709)
Joined: 7/17/2004
Posts: 985
Location: The FLOATING CASTLE
It would be nice if the fights were better. It looks like you are getting longer between random fights but sometimes you still only get the 1x or 2x multiplier. Maybe on purpose though? Seems that you don't always want to get the most steps between fights since having 0 at certain points such as entering a cave resets it (This is why he takes extra steps before entering island cave). With a little work you could plan out these spots and figure out what multiplier you want after each battle. Would require some complicated but interesting math. There is another way to abuse the encounter mechanic when you move from an area with lower rate to one with a higher rate without resetting the steps. For example on entering the Hagubo caves where the rate is 7 compared to 12 in the outer world. If you had an encounter right outside you might get 45 steps inside before your next fight when inside you cannot do as well. I guess if it doesn't cut the number of fights it doesn't matter though and this means you can't use 0-step reset trick either. One more thing, if you are maxing out fire power you could use FireG once each on Alonzo and Jarmlu to get 1 fire/1 sph instead of 2 fire.
Posted: 11/2/2013 1:20 PM
Quote
janus
Active player (285)
Joined: 9/1/2008
Posts: 900
since I might restart I will think about it. but for now, fireg on alonzo would prolonge the fight "If you had an encounter right outside you might get 45 steps inside before your next fight when inside you cannot do as well" worth trying also, for some reason, I can't freeze both renegades although sp is as strong as the published run (56). could it be related in a way to fire magic even though it doesn't use the element?
Posted: 11/2/2013 4:35 PM
Quote
janus
Active player (285)
Joined: 9/1/2008
Posts: 900
http://dehacked.2y.net/microstorage.php/info/1187410064/Paladin%27s%20Quest%20%28U%29.smv after many tries and backtracking (sp needs to be at least at 54 to freeze both renegades AND I needed enough money to get both mercenaries), I am now 13.8 ahead of the previous test (19.5k from the complete run).
Posted: 11/3/2013 4:51 PM
Quote
janus
Active player (285)
Joined: 9/1/2008
Posts: 900
http://dehacked.2y.net/microstorage.php/info/311799145/Paladin%27s%20Quest%20%28U%29.smv back in Karon with 43.4k frames advance from the test/59.2k from the complete run. I might fight a little more in Conshiuto if needed; getting experience is very easy
1 2 3 4 5 6