Idea for a 3-turn clear of Chapter 1 in HM

[ruadath]

Don't worry, I'm not starting another playthrough or anything like that; I was just playing around with Binding Blade last night, and decided to write up a report of what I found out. I'm too lazy (and busy) to actually carry out any of the stuff I'm posting here, but hopefully someone else will do it.

In case you aren't familiar with it, the first chapter of FE6 can be completed in 3 turns via the "rescue-death exploit," in which suiciding a unit carrying another unit (usually Roy) on the PP allows you to gain an extra turn for the rescued unit. A video of this clear can be found below:

Since the map is pretty much the same in HM, it seems like it would be entirely straightforward to generalize the strategy used here to HM. However, the problem one runs into is that all of your characters suck Wolt and Roy are no longer capable of OHKOing enemies with critical hits, and Lance isn't able to either unless a fighter spawns with a very low Def, which is highly unlikely (and can't be rigged properly since this is the opening chapter of the game).

Taking that into consideration, let's review what is necessary for the 3-turn clear:

- First of all, someone needs to kill the boss. Marcus is the only person who can reach him in 3 turns, and fortunately still has the stats to whack him with a Silver Lance crit, so he has to do the deed. This requires him to full-move every turn, which means

- someone (not Marcus) has to kill the closest fighter on the first turn PP

- Marcus has to occupy the forest tile on turn 2 PP

- We need to have someone perform the "rescue-death exploit." Bors and Wolt are out of the question for obvious reasons (too little mov or no short-range attack), and Marcus is killing the boss, so it has to be either Alan or Lance. Examining the map, one realizes the in order for one of them to get Roy close enough to seize on turn 3, they need to be on or beyond the forest the Marcus has to occupy, which means

- they need to be rescue-dropped forward by Marcus, since Marcus is occupying the forest

- the space west of Marcus has to be clear of enemies on turn 2; since the brigand on the village will inevitable occupy this spot on EP1, someone (not Marcus) has to kill him

- again, since Marcus is occupying the spot, the brigand can't be attack from up close, so he has to be Javelin critted.

- In fact, Alan needs to Javelin crit him, since Lance doesn't have the Str to OHKO him regardless of his stats. This means that Alan has to be the one who is sacrificed, unlike in NM, where you have a choice between Alan or Lance. It also means:

- The brigand needs to spawn with 1 Luck so that Alan can crit him

- The fighter east of the boss has to spawn with 1 Luck so that Alan can crit him on turn 2 EP (no one else can target him on turn 3 PP besides Marcus)

- More importantly, Alan needs to not be rescued at the start of turn 2 PP. This means that in order for Marcus to rescue-drop him, Lance needs to be able to rescue him on turn 2.

- Last but not least, we have to get Roy to Alan on turn 3. This means that someone needs to rescue Roy on turn 2 PP; the cavs are busy and Bors is too slow, so it has to be Wolt.

- Since this means pretty much everyone is busy on turn 2, we have to make sure that no random enemies get in the way during this phase. In order to do that, it is convenient to

have Roy rescue-dropped east of Marcus on turn 1.

So putting all of that information together, here is what should be a turn by turn strat with screenshots (minus actually doing the RNG abuse):

First of all, play through the entire introductions scene (you can mash B) to get the enemies to spawn with the right stats. In particular the 2 people get 1 Luck, and the closest fighter gets 2 Def so that Lance can kill him.

Turn 1: Roy takes a step east and takes Lance's Javelin. Burn some RNs, then Lance crit-kills the fighter from below. Alan trades his Iron Sword for Roy's Javelin, rescues him, and goes around to convene with Lance, one space short of fullmoving. Wolt moves ahead of him and takes Roy, then Marcus does the same, and drops Roy east.

Turn 1 EP: Roy will get attacked by 2 fighters and an archer, and Lance by a fighter. Neither of them can crit-kill anything, but I recommend having Roy crit the bandit to his east (since I think he will attack Lance on turn 2 EP, not sure) for EXP gain later on, and rigging a Lance dodge. Roy can afford to get hit (he won't be in any danger for the rest of the chapter) as long as he doesn't die.

Turn 2 PP: Burn a lot of RNs. Alan Javelin crits the brigand (39% hit, 1% crit), Lance rescues him and takes his spot (taking the Javelin as well), Marcus moves onto the forest, takes Alan and drops him east. Alan should only have an Iron Lance if everything is done correctly. Roy moves 2 south of Lance, then Wolt moves above him and rescues him.

Turn 2 EP: This is going to be a real nightmare to actually carry out. Wolt (weighed down by Roy) has to survive attacks from 2 fighters and an archer, all of whom double and 2HKO him. If that wasn't bad enough, Alan has to get hit by the fighter who charges him, and then (49% hit, 1% crit) kill him as well. Lance probably kills the guy who attacks him if Roy critted him earlier, and should be OK to survive otherwise.

I don't have a screenshot for this because I didn't do it.

Turn 3 PP: If you somehow manage to accomplish the above, things should be easy from here. Wolt moves forward and gives Roy to Alan, Alan suicides on the other fighter near the boss, and Marcus crit-kills the boss with his Silver Lance. Before Roy seizes, Lance can crit-kill someone in order to hit level 2 (pick someone who got injured by Lance/Roy on a previous turn).

You might notice that I haven't mentioned Bors at all in this post. That's because he doesn't have to do anything in this strat! Which means that unlike the video posted above, he's free to head east and pick up the 5000G from the village (the fighters won't bother him)!

Hopefully this inspires someone to carry this out.

[dondon151]

okay that's very nice

[General Horace]

the reality is nobody is going to spend 2 weeks finding a string where this actually all works

[ruadath]

It actually shouldn't be terrible to do in the sense that the GBA RNG is determined by a 32-bit linear-feedback shift register, which can easily be calculated in advance (as you guys see with the lua script), so I can probably dump something like the first million or so RNs into an array (and I probably won't even need that many). After that, the first 3 actions that occur (after burning some number of RNs that I've yet to determine) in order are the fighter charging Alan, and then two fighters charging Wolt. Their paths are all straightforward, so only 1 RN is burnt between each move, which makes it fairly simple to predict. Then I can just right some simple code to compare the values and iterate through the array until I find something that works.

All in all, it's a process that shouldn't take longer than a few days (if even that) if I put my mind to it, but meh, effort. I'll probably get around to it eventually.

EDIT: And the next person to move is the archer. I'm not sure what path they'll take, but it's likely that it can be easily fit into this framework as well. And to prove my point further, I just opened up an Excel spreadsheet and generated the first 10000 numbers with like 5 minutes worth of effort so I'm confident I could carry this out if I actually wanted to.

Edited January 22, 2016 by ruadath

[Gradivus.]

You have 2 playthroughs running already and you dropped around 10 playthroughs before in a pretty short timespan, I feel you should be able to see that you're jumping on ideas before thinking them over by now

Continue those that are active first, an FE6 HM rescue death exploit run would mostly just be a harder to execute replica of beamcrash's run anyway.

Edited January 22, 2016 by Gradivus.

[Espinosa]

You gotta admit that an entire (unhacked) run done in that style achieving the absolute lowest turn counts on HM would be hella impressive and interesting in its execution. But yeah, you'd have to bust your ass to get it done, which nobody has had the patience to do yet.

[ruadath]

You have 2 playthroughs running already and you dropped around 10 playthroughs before in a pretty short timespan, I feel you should be able to see that you're jumping on ideas before thinking them over by now

Continue those that are active first, an FE6 HM rescue death exploit run would mostly just be a harder to execute replica of beamcrash's run anyway.

Like I said in the OP, I'm not doing this right now (if ever), I was just pointing out that this is doable and that anyone on this forum who wants to this should be able to carry it out without too much trouble.

You gotta admit that an entire (unhacked) run done in that style achieving the absolute lowest turn counts on HM would be hella impressive and interesting in its execution. But yeah, you'd have to bust your ass to get it done, which nobody has had the patience to do yet.

Lol, it would probably involve quite a few several minute long RN burns. As I've told you guys on the FE7 threads, GBA RN burning isn't really my cup of tea, but again just pointing out that this is possible.

Edited January 22, 2016 by ruadath

[Gradivus.]

Got your post wrong then, my bad.

@Espinosa: Wrt hacking, beamcrash is only said to have hacked the RNG and even that isn't proven iirc. My point was that the most creative part of these strats has been discovered already. The execution is only impeded by stats, which are most likely to be addressed by rigging ridiculous EPs instead, which would probably make it less intriguing for the player. I don't see any fundamental differences between an NM and HM death exploit run, the former only saves you a bunch of headaches while being strategically similar. I'm not denying that either run would be impressive.

Edited January 22, 2016 by Gradivus.

[ruadath]

Got your post wrong then, my bad.

Don't worry about it.

@Espinosa: Wrt hacking, beamcrash is only said to have hacked the RNG and even that isn't proven iirc. My point was that the most creative part of these strats has been discovered already. The execution is only impeded by stats, which are most likely to be addressed by rigging ridiculous EPs instead, which would probably make it less intriguing for the player. I don't see any fundamental differences between an NM and HM death exploit run, the former only saves you a bunch of headaches while being strategically similar.

I'm not really all that familiar with beamcrash's run (or FE6 LTC in general, though I am actually familiar with the game, unlike FE11 lol). One example of what could potentially be a key difference between NM/HM runs in this format was pointed out by Irysa: Shanna needs a Silver Lance to crit-kill Henning in HM, whereas a Steel Lance suffices in NM. I have no idea if that kind of level difference can be made up for solely by rigging EPs.

EDIT: Again, not really familiar with resource distribution in an FE6 LTC, but the extra 5000G might be put to good use towards an extra Boots or something late game?

Edited January 22, 2016 by ruadath

[Gradivus.]

She "only" needs 16 str/16 skl to OHKO HM!Henning with a steel lance crit according to FEWoD (can always rig his stats by resetting via suspend + restart). If that can be reached (level 11 + promo), in all likelihood it'll depend on notable EP rigging.

And yeah in FE6 LTC boots are the main goal of conserving as much money as possible.

Edited January 22, 2016 by Gradivus.

[dondon151]

@Espinosa: Wrt hacking, beamcrash is only said to have hacked the RNG and even that isn't proven iirc.

either he admitted it when asked about why he randomly keeps animations on, or toothache showed that it was unreasonable for beamcrash to get the RN strings that he got without burning any RNs.

in fact iirc he pretty much never does arrow retracing yet still gets insanely low% crits with amazing level ups, so while that's not 100.0% definitive proof, it's about as confident as i can get in this claim. that's not to say that his strategies aren't still possible without hacking the RNG.

EDIT: in his chapter 16 clear, saul and lilina both get perfect level ups (1 in 4.4 million chance), milady, lalum, and fae survive combat that should've killed them, and fae gets a double 9% crit, all without arrow retracing.

Edited January 23, 2016 by dondon151

[ruadath]

either he admitted it when asked about why he randomly keeps animations on, or toothache showed that it was unreasonable for beamcrash to get the RN strings that he got without burning any RNs.

This was something that really confused me the first time I heard it. What does keeping the animations on have to do with anything? I've never tried hacking the GBA RNG, but I imagine that one could just pause the game (you're probably playing on an emulator if you're hacking RNG), edit the memory after each battle with either gameshark codes or just manually, and then press play again. Assuming he was using VBArr, the recording also pauses when you pause the game, and no one would be able to notice anything. It seems like it would be a lot more tedious to cut and paste together a bunch of different clips.

Then again, I've never tried this, so perhaps it causes some unexpected glitch.

[dondon151]

he was probably recording his videos in a not-so-sophisticated way.

[Irysa]

She "only" needs 16 str/16 skl to OHKO HM!Henning with a steel lance crit according to FEWoD (can always rig his stats by resetting via suspend + restart). If that can be reached (level 11 + promo), in all likelihood it'll depend on notable EP rigging.

The Silver Lance thing was to oneshot a Myrm who blocks the way to the Throne otherwise, not strictly to kill Henning. On further examination you can rig a Slim Lance crit on EP against him instead to make sure he doesnt get in the way.

Also Henning has 13 Def by default, not 12, meaning she'd need 18 Str, but yeah if you're allowing for rigging boss stats then you can get it lower.

Edited January 23, 2016 by Irysa

[Horakthi]

It actually shouldn't be terrible to do in the sense that the GBA RNG is determined by a 32-bit linear-feedback shift register, which can easily be calculated in advance (as you guys see with the lua script), so I can probably dump something like the first million or so RNs into an array (and I probably won't even need that many). After that, the first 3 actions that occur (after burning some number of RNs that I've yet to determine) in order are the fighter charging Alan, and then two fighters charging Wolt. Their paths are all straightforward, so only 1 RN is burnt between each move, which makes it fairly simple to predict. Then I can just right some simple code to compare the values and iterate through the array until I find something that works.

All in all, it's a process that shouldn't take longer than a few days (if even that) if I put my mind to it, but meh, effort. I'll probably get around to it eventually.

EDIT: And the next person to move is the archer. I'm not sure what path they'll take, but it's likely that it can be easily fit into this framework as well. And to prove my point further, I just opened up an Excel spreadsheet and generated the first 10000 numbers with like 5 minutes worth of effort so I'm confident I could carry this out if I actually wanted to.

Wow, this would be really impressive. If it wouldn't take you much effort, could you please do this playthrough? I want to see it.

[ruadath]

I'm certainly not going to start another entire playthrough at the moment. While it wouldn't be too much work to get Chapter 1 done, I am a little bit burned out of Fire Emblem at the moment, and I'm also busy a couple of things in real life (flying back to college later tonight, then I have to set up some things there, etc.)

Sometime in the next week or two I might be motivated to write a script that searches the RNs for a string that works, and then test it out by hacking the RNG to that point (I'm not going to sit around for 15 minutes or however long it takes scrolling the arrow back and forth) to check if it works. I'll keep you guys updated if I do this at least, but no promises.

EDIT: One thing I just realized

Turn 1 EP: Roy will get attacked by 2 fighters and an archer, and Lance by a fighter. Neither of them can crit-kill anything, but I recommend having Roy crit the bandit to his east (since I think he will attack Lance on turn 2 EP, not sure) for EXP gain later on, and rigging a Lance dodge. Roy can afford to get hit (he won't be in any danger for the rest of the chapter) as long as he doesn't die.

If I do this, it will in all likelihood affect the enemy turn order since the fighter that gets critted will probably go into "critical health" mode or whatever. This probably means that they won't end up attacking Lance anyway (since they'll just go for Wolt), but I'll have to think on this further.

Edited January 23, 2016 by ruadath

[ruadath]

She "only" needs 16 str/16 skl to OHKO HM!Henning with a steel lance crit according to FEWoD (can always rig his stats by resetting via suspend + restart). If that can be reached (level 11 + promo), in all likelihood it'll depend on notable EP rigging.

I completely forgot about this; does it work in FE7 too? If one decides to use this during the playthrough (and I see no reason not to), then some of the points in the analysis in the OP fall through, since you can actually get the brigand to spawn with 0 Luck.

In particular, it ends up being the case that you can sacrifice Lance instead of Alan. I imagine that this is a good idea since

1) Alan usually has a better starting position than Wolt in the 7 or so chapters before you get the preparations screen

2) Alan's higher Str is more immediately useful for reasons Irysa pointed out, plus the fact that he actually has enough Str to crit-kill things on EPs probably means he can gain more EXP in the long run and make up his Skl/Spd deficits (though the Skl one may hurt him, unclear)

However, as far as I can tell, one can't get this to work out without using Bors, which means no 5000G. I'm no expert on FE6 LTC, so I wouldn't know which ends up being more valuable, but feel free to discuss. I should also note that in addition to rigging a couple of 0 Luck enemies and a (2 Def, not 28 HP) fighter by the boss, this strat also ends up requiring even more RNG abuse on the turn 2 EP, since Wolt will end up surrounded by the 3 fighters and an archer all of whom double and 2HKO him, etc, and he needs to survive.

Edited January 24, 2016 by ruadath

[Gradivus.]

No idea, but FE7 has a very similar RNG to the best of my knowledge so you can probably try.

[Irysa]

Suspend maintains the RN string, so as long as you Suspend then don't actually reset, it works in FE7 and 8 too.

Also I forgot to mention the Horsekiller has 11 Mt so a 17 Str Shanna can critkill 13 Def Henning with it too instead of a Silver Lance.

Edited January 24, 2016 by Irysa

[ruadath]

No idea, but FE7 has a very similar RNG to the best of my knowledge so you can probably try.

As far as I am aware the difference is as follows. Both FE6 and FE7/FE8 use the exact same RNG to generate the RNs (which are actually 32 bit values, i.e. numbers from 0 to 65535), however the difference is that when converting it to numbers from 0-99, FE6 just divides by 656 and rounds down, while FE7/FE8 the increments vary between 655 and 656. This means that rolling 99s is considerably more difficult in FE6, and that the RNG between the two might be off by 1 every few hundred places, but it likely won't change anything important.

[Gradivus.]

Yeah that doesn't change anything about the RNG being very similar in its function. 100s being unable to miss or 99s harder to get (or whatever) has little to do with how stat rolls work.

Edited January 24, 2016 by Gradivus.

[ruadath]

Just tested it out, and indeed, it does work. Nevertheless I don't think I'll be incorporating it into my FE7 playthrough (and I don't think it will be necessary either).

As for this, I'm not planning to put any work into finding an RNG seed unless someone confirms which strategy they would like to see implemented. Especially in the case of the latter choice (sacrificing Lance over Alan), this would require a bit more work than I originally intended since it's not obvious to me how the game uses the RNs it burns at the beginning of the map to determine the stats of the enemy units.

[Vykan_12]

I'd be willing to give this a try using tools assistance.

P.S. I've been working on a heavily improved RNG lua script behind the scenes. It would make LTC shenanigans dramatically easier to pull off. If you're interested, I can PM you a link to the github repository.

[ruadath]

That sounds pretty cool, I'd love to take a look! I can share more details with you regarding the strats (don't know which one you are interested in implementing) via PM if you'd like.

Also, I don't know if you still do TAS stuff anymore, but I can perhaps share some stuff with you that I created for people interested in doing a RNG-abused segmented speedrun of FE10 HM (the idea inspired by your awesome NM run on that one website)

[ruadath]

Since I finally got around to setting things up to be able to easily search the GBA RNG for seeds, I decided to take a look at this once again and almost immediately noticed that I made an oversight in the main strategy:

Turn 2 EP: This is going to be a real nightmare to actually carry out. Wolt (weighed down by Roy) has to survive attacks from 2 fighters and an archer, all of whom double and 2HKO him. If that wasn't bad enough, Alan has to get hit by the fighter who charges him, and then (49% hit, 1% crit) kill him as well. Lance probably kills the guy who attacks him if Roy critted him earlier, and should be OK to survive otherwise.

In addition to all of that, the two southernmost fighters gang up on Bors, who has a chance of dying as long as either of them are able to double them. So rigging this might be even more difficult than anticipated...

EDIT: This could perhaps be remedied by having Roy crit one of them on the turn 1 EP so that they would get moved up in the turn order list and attack Wolt first, which would keep Bors safe. This would probably have a side effect of having Marcus get attacked by one more enemy, though that won't trouble him.

Edited March 4, 2016 by ruadath