dwangoAC and I have both been busy, but what progress there's been has been debating how to handle Astral.
We may resort to the figurine method after all. It only has a 1 in 9000 chance of working, but those odds aren't completely out of reach for a TAS, and luckily we have two degrees of freedom to line it up correctly (the RNG seed, and the turn we obtain the figurine). So the next step would be figuring out how to set it up, and then testing that it works.
More thoughts about Water and Astral:
If we can't use conflict on Astral, then bringing a tame monster through the portal with us isn't useful, so we won't need so many eggs on Water; one is all we'd need, and only to be able to drown faster. (We're using safe_teleds to aim for the portal, not enexto.) Having a peaceful or tame monster around would help with luck manipulation, but doesn't seem necessary otherwise.
We could make the existing strategy work using the monster-push technique (which we've proved is possible but mindboggingly hard to manipulate) in order to give us a free action on Astral before the turn boundary. That way, we could put on a ring of conflict (we'd be in natural form at the time, so no problems with lack of hands), and still get to the altar by the end of turn 2015.
Alternatively, we could dispense with the existing Astral strategy and use a figurine to create the following monster who'll attack us that we need. This would make monster-pushing impossible, AFAICT (unless we can do it off a tame egg-spawned monster who isn't used to conflict-attack us, or a Rodney summon); but it wouldn't matter so much, as we wouldn't need to spend the last action of turn 2014 putting on a ring and we could use it to just manually step next to a wall.
Finally, is the possibility of triggering enexto off a random spawn on Astral, rather than dragging one from Water. We are, after all, spending a turn boundary to hatch the eggs, which should theoretically be possible to charge the monsters. I thought this was impossible because none could charge fast enough, but that actually seems not to be the case; vampire lords are both a legal spawn upon Astral level generation, and can legitimately appear arbitrarily near the player. I just thought of this today; there's something to be said for using the "crazy luck" approach rather than the "crazy preparation" approach. I haven't had time to test it yet, but it's looking like quite a serious possibility.
Joined: 7/12/2009
Posts: 181
Location: São Paulo, Brazil
Just reporting that I'm looking forward to seeing this TAS, and just reading all this insight on what you're planning/doing and the turnbyturn.txt is enough to vote yes ten times.
Good luck with the luck manipulation!
We hit dlvl 2 at the start of turn 2012. Dlvl 2 and 1 each take 1 action, Earth takes 1 (enexto setup), turning back to air-E form after enexto takes 1, that's turn 2012. Then Air takes 1 action + the turn boundary, and we spend 2 actions (plus 1 recovering from the enexto on Air) on Fire; that's turn 2013. Then Water takes the turn boundary + 2 actions - 1 action (because we glitch an extra action into the turn), but we have to use a polyform that only gets 3 actions per turn; upshot is that we have 2 actions left in turn 2014 as we arrive on Astral.
And so, we end up on Astral with 2 actions left, in natural form. There's no way I know of for completing Astral in 2 actions without help from a turn boundary, so that means winning on turn 2015. What we do on turns 2014 and 2015, though, is currently up in the air.
It's been silent. How goes the breaking of things?
dwangoAC and I have mostly been busy, but there's been a bit of progress (unfortunately, mostly negative results).
The good news is that we found another spare action on Water (getting the bubble to move from beneath us at turn boundary, which we need to spend anyway, in order to save an action moving out from it manually); it takes a bit of manipulation but it's not unreasonable. This gives us at least four or five viable ways to do Astral, via giving us a bit of headroom before the frame rule there. (We still don't know which is fastest, but because of the frame rule, it doesn't matter.)
The bad news is that we've been trying to work out how to do the egg setup, and it turns out that there's no way to adjust the timer of an egg once it's been set. (We were testing iceboxes; by NetHack logic they should slow the timer, or at least kill the egg almost immediately like they would in real life, but instead they act identically to chests, i.e. not really useful at all.) So this means that we have to manipulate all the timers to be correct, the instant they're created. At the moment the only real idea we have for this is "stand next to a wall and create the timers one at a time, manipulating in between." There are a couple of ways to create the timers. The fastest in gametime is to create them at the same time that we set the egg species, but that requires manipulating two unlikely events with no chance to manipulate in between, so is probably not viable. Alternatively, we could create the egg, then reset the timer on a later turn (we might have to kill the timer in between to stop it hatching, but that isn't difficult); that takes twice as long but is easier to manipulate.
If anyone does know a way to manipulate egg timers (other than via rerolling them with undead turning, which we already know about), please let us know in the thread.
In case you ever need to do it again, Xorns can manipulate luck to some extent without using time by kicking at empty space.
Neat idea. Reordering actions seemed to work well enough, but if we have to slip through walls in the future, that might come in useful too. (Also, what we were doing the first time we were xorning was AI manipulation, which is mostly based on your movements, not the RNG.)
More turnbyturn.txt editing:
turn 814: "to manipluate it for us."
822: " the distribution of timeouts is utterly bizarre " Ah, good old rnz.
Still interested in progress, and what kind of ring the Xorn ate.
Is 823 where the actual TAS progress is?
Moderator, Senior Ambassador, Experienced player
(898)
Joined: 9/14/2008
Posts: 1007
eternaljwh wrote:
More turnbyturn.txt editing:
turn 814: "to manipluate it for us."
822: " the distribution of timeouts is utterly bizarre " Ah, good old rnz.
I'll try to get back to this soon. Both ais523 and I have been busy as of late (and distracted by SGDQ :) but we hope to make some more progress soon. For the record, we expect to finish more than 12 months from now based on how much work is left to do. This is such a complicated game...
On the upside, we're on the 4th "server" computer we're collaborating on since we started the run three years ago. :)
A.C.
******
It's been 6 months since we've had a post here, so I just wanted to remind you that some of us are still excited and hopeful about this! We know there's still a load of work to do, but here's cheering you on!
I'm not really here. Pressing Up+Down too many times glitched your graphics. And audio. Yeah.
Thanks for the encouragement.
dwangoAC was busy with AGDQ, but isn't any more. Meanwhile, I'm busier, which means I have more time to work on the TAS (when I have lots of free time, I tend to work on things that require large blocks of free time and can be done offline; things that require online access, like the NetHack TAS setup we have, thus tend to be more activity for me when I'm busier).
The next step is still working out how to get a very large number of egg timers to work up (most likely, this will be done as or soon after we create the eggs). It's one of the remaining absolutely vital steps that hasn't been rehearsed.
Moderator, Senior Ambassador, Experienced player
(898)
Joined: 9/14/2008
Posts: 1007
What ais523 said. :) I'm in a "crunch time" at work but can make some time short-term. One of the scheduling challenges for ais523 and I is the fact that we're 7 to 8 timezones apart (depending on DST; ais523 is in the UK and I am in California).
As ais523 noted, the egg setup is the only thing we haven't "rehearsed" (tried in wizard mode with all of the benefits thereof). There's plenty of other things we've rehearsed that we'll have to remember how we did when it comes time to do it the "right" way outside of wizmode but I think we'll be OK. We're still tracking for a 2015 completion date, although that'll be at risk of slipping if we don't start making progress soon.
Thanks for the continued interest!
A.C.
******
Joined: 3/31/2010
Posts: 1466
Location: Not playing Puyo Tetris
Do I even want to know what you are doing? I haven't played enough Nethack to know if I should be scared for the game or not. Still, it's great seeing how broke things have been so far.
When TAS does Quake 1, SDA will declare war.
The Prince doth arrive he doth please.
Moderator, Senior Ambassador, Experienced player
(898)
Joined: 9/14/2008
Posts: 1007
hegyak wrote:
Do I even want to know what you are doing? I haven't played enough Nethack to know if I should be scared for the game or not. Still, it's great seeing how broke things have been so far.
Do I even want to know what you are doing? I haven't played enough Nethack to know if I should be scared for the game or not. Still, it's great seeing how broke things have been so far.
The game's badly bent but (at least in this category) not completely broken. Think of it as "almost everything makes sense in a twisted way" (apart from what we do with gold, that doesn't really), rather than "this run makes no sense at all". It could be done faster using total control, but we're categorizing that out (partly because it's a pain to optimize and we don't fully understand the setup, and partly because this run is more entertaining).
For instance, we get mindbogglingly high stats, but none of the individual steps in that are nonsensical; we gain stats when we level up, lose them when we level down, and it's just that the amounts don't match. We do things completely out of the standard order, but there isn't a fixed order (and the game has code to account for the order in we actually do things; it's intended to be possible, even though it's bad strategy outside a speedrun). We always have appropriate magical abilities to hand without needing to carry around items with the purpose, but that's just the result of an ability that causes them to change at random combined with huge amounts of luck manipulation; nothing about that is impossible or even unintended, it's just very unlikely. We have perfect knowledge of items, enemies and the map, but a sufficiently lucky player could simply do that by guessing.
So apart from the gold glitches, and the enexto glitch (which will come up heavily later on but hasn't been used in the run yet), there's nothing in the run that outright makes no sense. It's simply just using a route that can't be replicated on console because it requires far too much luck, and exploiting some interactions between mechanics that individually make sense.
It's been 6 months since we've had a post here, so I just wanted to remind you that some of us are still excited and hopeful about this! We know there's still a load of work to do, but here's cheering you on!
I was about to do this...I see that turnbyturn is still at 822 where it was 6 months ago.
t2000timings.txt wrote:
2001:49 Read the Book. (This costs 24 movement energy.)
2001:M1 Air elemental steps on the polytrap, becomes a cockatrice.
2001:M2 Cockatrice moves adjacent to us.
2001:25 Levport to Sanctum.
2001:M3 Cockatrice oneshots HPoM.
2001:13 Step on HPoM's corpse, autopickup amulet.
Okay, uh...just to make sure, levport into Sanctum is to be manipulated to land you with the High Priest? There's no "landing zone" like VoD?
The Sanctum has no from-above teleport region. Not only have we tested those turns explicitly to ensure it works (it does), but this strategy is routinely tried realtime (via using a large stack of c?oT and repeatedly teleporting between the Sanctum and VS levels until the teleport arrives inside the inner Sanctum; it's actually pretty likely, because it's open space and most of that level is packed with monsters).
We actually made some progress on the run yesterday. No turn played, although we rejected a range of RNG seeds for the next move. We calculated the efficiency of the egg conversion engine to calculate how many scrolls of earth we needed. If the answer was along the lines of "50", we'd have gone and raided a scroll shop, but it turns out that we only need 3 or 4, so we're going to raid Sokoban:6 instead, which has two guaranteed scrolls of Earth, loads of boulders, and one guaranteed wand (having a guaranteed wand on a level, when normally they're rare, means that we get a very good opportunity to manipulate a wand of wishing, which is good because our current wand has only the one charge left).
Now, Sokoban:6 is pretty much decided as our next destination, but the problem is that we have to go via the currently non-generated level Dungeons:7, and what's holding us up at the moment is to find something interesting to do when generating it. We were considering generating 7 candles naturally, but it's unclear how exactly to manipulate that, at the moment. If anyone has good ideas for something to do with the level (ideally in terms of speed rather than entertainment), let us know. They can't be too unlikely; about a 1 in 2000 chance is viable, anything less likely should be avoided.
turnbyturn.txt is a little behind the actual run atm, but I hope to update it some time in the next few days.
One other problem we've run into: the current WIP ignores two leashes on dungeon level 16, when we should have picked them up last time we were there. We may well be returning to the level anyway later on (it's designed to have a convenient RNG environment), but it seems very sloppy to just leave items on the ground that we later need; it shows a lack of planning. I'm going to review the run to see how close to them we came; it may well be that it's faster to collect them on a separate trip anyway, in which case we have an argument for avoiding them last time. (It's not like any particular location on D:16 is better than any other for RNG purposes.)
The obvious thing to generate on D:7 is more Scrolls of Earth.
Generating a staircase/trapdoor/hole/level teleport trap adjacent to the Sokoban staircase would also be flashy.
Trying 127.0.0.1...
telnet: connect to address 127.0.0.1: Connection refused
telnet: Unable to connect to remote host
I think our quest for "which item to generate" has been solved; looking at the TAS progress so far, I see that dwangoAC has generated a D:7 which has a shop selling two wands of wishing. Not sure we're going to manage much better than that. (So much for my idea of generating one in Sokoban, although I guess we can do that too.) This gives us 2 wishes remaining from our existing wand, a total of 12 from the two in the shop = 14, and we could get up to another 14 from a wand of wishing in Sokoban (trivially easy to manipulate compared to some of the things we've been doing) and a hitherto unrevealed wish source. I'm wondering if having this many wishes might make things a bit repetitive, actually.
We should probably use this D:7 as a strong demonstration of what a TAS can do, even if the wishes subsequently end up irrelevant and we ascend with the wands unused. It's not like there are even 28 non-randomly-generating items we need. EDIT: We could wish for cursed teleport, I guess; we always seem to need more of that, and although it generates randomly, you don't get very good value because the main method to force items to generate randomly is cursed teleport…
Incidentally, I'm wondering about creative ways to extract them from the shop. Grab and teleport works, but some way to keep the shopkeeper happy would be more fun. The standard "wish for gold" trick doesn't work because of the antigold we're using for carry capacity, but it's been established that dropping antigold in shops does weird things, so we may be able to overflow the debt counter and get the items for free that way.
This is all moot if the seed turns out to require too much manipulation to hit, of course.
In other news, I've been looking into the leash situation. The conclusion: we left two leashes on the ground earlier that we need, and we could pick both of them up and yet still save three turns, due to a massive oversight in that section of the run (turns out polyform speed affects spellbook reading, something that I really should have checked rather than relying on an imperfect memory of the code). The problem is that that's about halfway through the TAS, and it makes the monster AI desync. I'd love to go back and fix it, but resyncing the rest of the TAS is going to be very difficult, if not impossible, meaning we might have to redo half of the TAS again. We can, however, line up level generation and most things like that reasonably accurately; it's just the monster AI that desyncs pretty much irretrivably. Thus, particularly lucky turns that we got before are still obtainable and won't have to be manipulated.
I'm genuinely unsure on whether to redo from halfway through the TAS or not. It looks sloppy, and it is sloppy, so I don't really want it in the run, but it might be a lot of effort to get everything lining up like it should. I guess one option is to see whether we can get the gold duplication section to sync again via moving random seeds around (without changing our movement pattern); if we can keep the same movement pattern, everything should start syncing again after a while.
For shop stealing--don't you have at least 433 rocks by now? It would be a lot of typing, but this sounds like a great opportunity to show off that particular exploit.
Perhaps the best plan with the leashes is to look for a particularly funny circumstance to retrieve them under, so that it looks more like a playaround spot than a sloppy mistake.
