Hatkirby on October 29th, 2023 at 12:36:03pm
Hey all, I'm here to talk about The Witness again. When it comes to speedrunning video games, it's uncommon for players to play the game just as the developers intended. Speedruns usually incorporate tricks that allow you to do things faster than you're supposed to or that allow you to entirely skip doing something ordinarily required. And very often, these tricks take advantage of glitches in the game, to achieve something that goes past the boundary of the laws of the game.
The Witness, for a very long time, was almost entirely glitchless. The only real exception was Windmill Cycle Skip, which was discovered within the first couple of weeks of the game's release, and saved some time in 100% and 99.8%. I'll talk about it more when I talk about Windmill. But other than that, we were really just playing the game as intended, as quickly as possible.
There were a few tricks we used that seemed like glitches, like Jungle Wall Skip and panel sniping. Even Latch Skip is arguably not a glitch, and rather just an exploit of the game's error correcting functionality, although people will argue you on this one. People had definitely discovered things that were glitches, some of which would become very important later on (e.g. the Theatre Pause Glitch), but at the time they just seemed like random bugs with no real use.
And then everything changed when we found out that walls aren't real.
Enter: Panel Sniping
To solve a puzzle in The Witness, you have to enter solve mode, done by left clicking or pressing space. This freezes the player in place. Mouse movement no longer turns the camera, and instead moves around a cursor that has appeared in the middle of the screen. You can use this cursor to click on the start point of any puzzle within your locked view, after which the game allows you to input a puzzle solution.
The key thing here is "the start point of any puzzle within your locked view". Solve mode basically flattens the 3D world before you into a 2D plane. Anything within your view that is unobstructed by something closer becomes part of that plane, and can be interacted with. As a result, with very few exceptions, The Witness allows you to solve puzzles from any distance. We call this trick "sniping".
Sniping is old. Sniping in all likelihood was discovered the day the game came out, if not earlier, during beta testing. Swamp snipe (pictured above) was known as early as February 1st, 2016, less than a week after the game came out. It's also the gift that keeps on giving. Whenever we want to see if we can skip something, the first thought is often "can we snipe it?" And the answer is often yes.
There was one huge limitation when it came to snipes, though. The panel you're sniping has to be unoccluded in your view. It doesn't have to be visible; there are snipes involving panels so far away that they aren't rendered. But if there's any kind of wall or object between you and that panel, it blocks your cursor. So a lot of snipe-crafting involves finding specific positions you can stand in that allow a panel to become completely unoccluded in your view. It's much harder than it sounds, and often a potential snipe will remain barely out of reach thanks to random pieces of geometry.
Now that you know that, let's talk about how these rules got turned on their head.
Keep Hedges 2 from Mountaintop
The beginning of the end was not signaled by raging fires or swarms of locusts. It was a day like any other.
In July of 2021, I was routing out a ridiculous category called "Touch All Audio Logs No Mountain Low%". Let me break this down.
- There are 58 recorders scattered around the island, which we called audio logs.
- Clicking on them (or "touching", as the category title refers to it) makes them play recorded messages.
- Some of these audio logs are located within or under the mountain, which means you'd have to activate all lasers in order to get them. This would make the run basically "All Lasers + some minor extra stuff", so to allow for a more unique route, we specify that only the 45 audio logs obtainable without entering mountain are needed.
- "Low%" is a modifier that usually means "do the category's main objective, while also doing the least of something else". In this case, that "something else" is solving panels. In this category, you are required to solve as few panels as possible while touching all audio logs outside of mountain, even if solving more panels would allow you to go faster.
I've done a lot of arbitrary and frankly unnecessary things in this game. I like it.
In the front of the Keep, there are four hedge mazes. There's an audio log in the front wall's ramparts, which can be accessed after completing the second hedge maze puzzle. Usually, you can only get to the second hedge maze by solving the first one, mandating 2 puzzles solves in order to access this audio log. However, there is a shortcut door out of the rampart area that opens when you solve Hedges 2. I figured, if I could find a way to solve Hedges 2 without doing Hedges 1, I'd be able to get into the ramparts and reduce my solve count by 1.
On July 18th, I documented my Low% route, including a discovery I'd made. It is possible to snipe Hedges 2 while standing on the top of the mountain. The description of the trick included the following picture and caption:
The panel is completely invisible because it is blocked by the castle wall -- however, the wall has no collision so the panel is still solvable.
This statement is false.
It was an understandable assumption to make. There's no reason why that wall needed to have collision coded into it, since it wouldn't be occluding anything in normal gameplay. It's common for game developers to cut corners in places that don't really matter. But it remains an incorrect assumption.
I posted about my Touch All Audio Logs No Mountain Low% route in the speedrunning community's Discord server, and I mentioned the snipe I'd found. No one acknowledged it. It was never watched by a moderator, since it was not a category with a leaderboard. I am likely the only person to have ever run it. Both the route and the snipe were soon forgotten.
Keep Back Laser from Mountaintop
A few weeks passed. I was still interested in Low%, but I'd set my sights on a loftier goal: 7 Lasers Low%. 7 Lasers is the main way to beat the game, so 7 Lasers Low% is just beating the game while solving as few puzzles as possible. One of the lasers you get in 7 Lasers Low% is Keep, since it only requires three solves:
- Hedges 4, which activates the Front Laser panel.
- Yellow Pressure Plates, the first puzzle in the back of Keep. Solving this allows you to get into the back wall's ramparts, from which you can snipe the Tower Shortcut panel, allowing you to go up the Keep tower without solving either the rest of the hedge mazes or the rest of the pressure plates. The Tower Shortcut is notably a "non-counting" panel; solving it does not add to your solve count, so we are freely allowed to use it.
- Front Laser.
I was interested in whether there was a way to reduce the required solve count in Keep. The first idea involved sniping Front Laser1, which would allow us to skip doing Yellow Pressure Plates. Unfortunately, Front Laser faces away from the rest of the island (so there's no high place to stand that would face it), and it's angled upward, which makes sniping it from below infeasible. There is a location near Quarry that you can stand where it is possible to click into Keep Front Laser, but it is not possible to draw the full solution path.
A second idea involved somehow doing Back Laser instead of Front Laser. Back Laser does in fact face the rest of the island, including the mountain. The problem is that even if you could snipe Back Laser, there is pretty much no situation in which it would be useful. Back Laser only turns on when you solve Blue Pressure Plates, and if you've solved Blue Pressure Plates, you can just walk up the tower to get Back Laser. Adding the snipe only adds complexity to your route. On top of that, you wouldn't even be able to get to Blue Pressure Plates without also solving Yellow Pressure Plates and Green Pressure Plates, which gives us a solve count of 4, as compared to the Front Laser strat which only gets 3.
I kept playing with it, however, and on August 5th, I posted the following messages in the Discord:
btw i think i found the actual most useless snipe in the game: keep back laser from mountaintop. 1) no route would ever want to use it, not even for a low% category. and 2) it's so precise to input that i haven't even managed to do it at increased FOV
it looks like you're blocked by a wall but i guess it doesn't have collision at this distance? i can feel the whole grid, including the breaks and the endpoint, if i move my cursor around. i just haven't managed to input the right solution yet because it is so long and so wiggly and so dumb
This trick may be completely useless, and it was indeed again immediately discarded by the speedrunning community, but there's an important insight in that message that wasn't there last time:
i guess it doesn't have collision at this distance
Another piece of the still-unfolding puzzle.
Desert Laser Redirect from Keep Tower
This is when things start to pick up. Only four days later, while continuing to mess around in Keep, I made another discovery, and posted the following video in the chat:
The Desert laser is sort of "broken", in that it doesn't point at the mountaintop like the other lasers. Instead it points at Town, directly passing by a mirror on a rotary. There is a panel next to this mirror that controls the angle the mirror is facing at. You need to solve this panel in a certain way in order to redirect the Desert laser to the mountain.
In this video, I stand at the top of Keep tower, I solve Front Laser, and then I snipe Desert Laser Redirect through the wall of the Town tower. This, finally, seemed to have an actual use. The standard 7 Lasers route does Town as one of the lasers, but 7 Lasers Low% replaces Town with Desert because it is fewer solves, even though it's slower since it's farther away from everything else. Since 7 Lasers Low% did Desert, you had to solve Redirect at some point. Until then, we'd just detoured into Town after doing Keep, just to solve Redirect, before heading into Monastery. But, with this new tech, we could solve Redirect while still in Keep and then walk directly to Monastery, without going into Town at all. If you could perform the snipe quickly, you could save perhaps 7 seconds.
I, again, assumed that the tower wall just didn't have collision for some reason, but through further testing I found something odd. The Witness has three preset graphics configurations: High, Medium, and Low. I usually play on Low because it gets the best performance, but there's much more visual detail at High. Since this trick actually had an application in a mainboard category (albeit perhaps the least popular one), I wanted to find a good lineup for this snipe, so I switched to High graphics settings, hoping that the increased visual detail would provide some better landmarks.
And the snipe didn't work.
Thus the next piece of the puzzle fell into place: through testing, I found that you could only snipe through the Town tower wall on Low or Medium graphics settings. Something about High graphics prevented you from being able to click in. This lined up in an interesting way with my previous assertion that the Keep tower wall didn't have collision at a certain distance. It is not uncommon for video games, when configured for lower-performance environments, to cut corners when it comes to objects that are far away.
Despite these things coming together, the true ramifications of this still hadn't been realized. ADGOD and I traded the 7 Lasers Low% world record back and forth a bit, both using this trick to save a few seconds in the middle of the run. Because this was a real category, as opposed to Touch All Audio Logs No Mountain Low%, the leaderboard moderators had to watch and verify these runs. But the trick was not considered controversial at all, really. It only saved a couple of seconds, and besides, it's not like it was a glitch or anything. The wall just didn't have collision coded into it.
Mountain Discard from Treehouse
In mid-October, two months later, I was in a voice call with ADGOD, and we were talking about the Desert Laser Redirect from Keep Tower snipe. He asked me "do you think it would be possible to snipe Mountain Discard from outside Mountain?" I immediately said "Nawwwwwwwww."
Once again, I was wrong.
There's a category extension2 called "All Discarded Panels". There are 14 gray puzzles with triangle symbols on them scattered across the island. They're peculiar in that they're just strewn on the ground, instead of mounted on a wall like pretty much any other puzzle. They don't do anything when solved, other than hopefully teach you what the triangles mean. The quest to solve them takes you to several, but not all, of the major parts of the island.
While most of them do not have very strict requirements to access them, there is one in particular that practically defines the entire route: the one within the Mountain. Entering Mountain at all requires seven lasers, so now, not only do you have to find and solve the 14 discards, you also have to complete seven lasers along the way. ADGOD's question about sniping Mountain Discard is important because being able to do so would cut minutes off the run's time.
On October 28th, I was hanging out with my speedrunning friends, and the topic of the snipe once again came up. After I dismissed the idea, I went to sleep, and filled my head with the charming lies of the night. Meanwhile, in the real world, ADGOD got to work. It wasn't long before he proved his hypothesis correct: there was a very precise position in Shipwreck from which you could solve the Mountain Discard through the wall of the mountain. It just goes to show that you can't discourage a dragon who's made up his mind.
And this wasn't even the end of it. Later that same night, IHNN found another position; one in Treehouse, with a much more consistent lineup.
You walk up to the first Yellow Bridge panel in Treehouse, without solving it. Entering solve mode in front of most panels pulls you into position right in front of them. We use this to get consistent position and camera angles before doing precise snipes. You then walk into the top left corner of the bridge, make sure your 3D sensitivity is at the default, and then line up your reticle with a specific part of the mountain. From there, you can double click and immediately be clicked into the panel.
Not only was this the first through-walls snipe with real consequences in a category that people actually ran, it was one of the first snipes with a "perfect lineup". Lining up for distant snipes is pretty difficult, and often involves a lot of clicking around in solve mode. But around this time, people (IHNN in particular) were discovering new methods of lining up that put the burden on your camera angle and position rather than the solve cursor, and reduced the start of the snipe to a double click. Needless to say, the combination of new glitches and strats made this a pretty significant era in The Witness speedrunning.
I awoke at 5am, and immediately looked at Discord on my phone, as is my habit to this day. In the time since I'd gone to sleep, a new through-walls snipe had been discovered, and ADGOD had beaten the All Discarded Panels world record by nearly five minutes. My reaction to this was understandably surprised:
I went back to sleep soon after, but not before congratulating him, and declaring my excitement at the idea of trying out the run myself.
In the morning, I went to the oral surgeon, and had my wisdom teeth extracted.
It was a pretty quick and simple procedure, although I was disappointed when I asked them if I could keep the extracted teeth and they told me no. But it wasn't long before I was back home again, slightly loopy from the sedation, resting in my bed. I fell asleep and woke up a couple of times, until one point when I, in a haze, remembered the cool new snipe that had been discovered the previous night. So I got out of bed, went to my computer, and practiced the trick a few times.
Within two hours of waking up from surgery, I'd achieved a 12:39, breaking ADGOD's record by nearly a minute.
ADGOD/IHNN's cool snipe got me thinking, though. It was becoming clearer that there was true power in these through-walls snipes. I had an idea I wanted to try out, and I spent some time toying around with different panels and different positions. A couple of hours later, I posted two videos in the speedrunning chat, preceded by the inscription "oh my god oh my god oh my god":
These videos show me sniping two panels referred to as the Catwalk door, and the Town Shortcut door. Both videos make use of a game mod made by darkid that allows you to zoom your view in much more than the game normally allows, as well as allowing you to lock your camera angle to the start point of a puzzle. While I'd used a mod to help find these tricks, they were performable (with much more difficulty) in an unmodded game. And they, together, formed what I dubbed "UTM Skip" (although my friends call it the "no-wizzies snipe", after my newfound lack of wisdom teeth).
There are two main objectives in The Witness, as evidenced by the two Steam achievements: Endgame, and Challenge. Endgame is done by getting 7 lasers, proceeding through mountain, and entering the "Wonkavator" in the pillar room. Challenge, on the other hand, is done by getting all 11 lasers, going through the secret area "under the mountain" (or UTM as we call it), and completing a randomized, 14-puzzle timed challenge located in a cave under Town. Two of the most popular mainboard categories, 7 Lasers and All Lasers, map directly onto these objectives.
For a long time, there was something interesting about the way "All Lasers" was defined, though. There was no mention of lasers in the rules at all. Instead, the category was defined as so:
Beat "The Challenge" (Secret Area) from a new game.
Over the years, the question had been raised again and again: "what if we find a way to skip into Challenge without getting 11 lasers?" And every time, the answer was "we'll worry about that when it happens."
Perhaps you can see where I'm going with this.
The other end of the Challenge cave contains a door which, when the panel on it is solved, opens onto a catwalk suspended over the Theatre. This catwalk leads to a secret part of the tunnels, from which there are a few shortcuts out to the main island. One of them exits into Town. Using the two snipes I'd discovered, you could open both of these doors without activating any lasers. It was now possible to start a new game, complete Tutorial, make a couple of detours, and walk right into the postgame.
This was it. This was the holy grail of Witness speedrunning skips.
There are two important things to discuss about UTM Skip. First was that these snipes were hard. Catwalk Snipe was (and still is, I believe) the farthest away useful snipe in the game, at 222 metres. The title was previously held by Mountaintop Redirect Snipe at 208 metres. The redirect panel is also a straight line (sort of...), compared to the Catwalk door, which is a 4x4 panel whose optimal solution has 16 segments and 6 turns in it. Town Shortcut snipe is about half the distance (which is still high when it comes to snipes), but the panel itself is a 7x7 grid, which is massive. The solution I use has 43 segments and 11 turns. It would take a ludicrous amount of precision to solve it after managing to click into it.
This discovery had officially broken open the floodgates, though. Only a couple of hours later, Talchyr found an easier snipe that could stand in for Town Shortcut Snipe:
While standing at the top of the Keep Tower, you can snipe the UTM Desert Shortcut door.3 This door connects the final room of Desert to the secret tunnels I mentioned earlier. This panel is a simple, straight line, which means it is very easy to solve once you manage to click into it. Catwalk Snipe was still required, since the Catwalk door stood between Challenge and the tunnels, but this did a lot to make UTM Skip seem more viable.
A caveat of this simpler snipe, though, is that it is overall slower than the route with Town Shortcut Snipe (assuming you could do Town Shortcut Snipe quickly, which, need I remind you, no human had yet managed to do without mods). While Town Shortcut Snipe can be done with zero lasers, Desert Shortcut Snipe requires one: Desert. The UTM Desert Shortcut door sits in the final room of Desert, behind an elevator bearing the Desert laser on top of it. When you solve the laser panel, the elevator rises to the surface, and the laser activates. It is impossible to walk around this elevator; you have to activate the laser if you want to get past it to the shortcut door. Not only that, but the snipe itself is done from Keep Tower, which means you have to do enough of Keep to end up at the top of the tower, even if you don't then actually activate the laser.
Still, doing one and a half lasers to get to Challenge was undeniably faster than doing eleven lasers and going through Mountain to get to Challenge. UTM Skip was real, and it was effective. A category extension -- foolishly named Challenge% -- was created, where the only rule was that you had to beat Challenge from a new file. For a little while, Cynthia_Clementine and I did 1 Laser runs, trading the WR back and forth.
One day, after she'd beaten my time again, I was out taking a walk, the autumn breeze chilling my bones, and I thought to myself, "Why am I not doing zero lasers?" I had sort of written off Town Shortcut Snipe as too difficult to do in real time, but there's no reason that needed to stop me. I have a great love of snipes -- seeing FearfulFerret do Town snipes in his 2016 GDQ run is what got me interested in running this game to begin with. So I sat down. I practiced the snipes. I tried some runs. And eventually, I got it:
This is a video in which someone (me) starts a new file, and then completes The Witness's final challenge in under eight minutes. It's something I'm really proud of. Completing the Challenge in zero lasers was something people had theorised about since the release of the game -- and it was finally, provably possible. It's strange that such a run didn't belong in a mainboard category; instead relegated to the category extensions.
That brings me to the other important thing to discuss about UTM Skip, which is that it was controversial.
Here's what we know about sniping-through-walls:
- It is possible sometimes to solve puzzles through solid collision.
- You must be standing very far away from the nearest solid collision in order for it to work.
- You must be playing on Low or Medium geometry detail for it to work.
- It has been shown to work on both PC and Mac, with Intel, AMD, and NVIDIA graphics cards. It is a glitch with the game, not with the environment the game was being run in, which is significant, because platform differences are a legitimate reason to ban a glitch.
With those axioms in mind, we need to ask ourselves: how will this affect speedruns?
Even though no reasonable person would expect to be allowed to only get one laser in a category called "All Lasers", the debate about its legality and what needed to be done to the leaderboards sprung up immediately. Additionally, now that sniping through walls had proven to be useful in a real, competitive category, it shone a light on its other pre-existing uses, and the moderators responded by banning through-walls snipes across the board until the community could decide what to do about them. Crucially, this involved declaring that the top two 7 Lasers Low% runs may no longer be considered legitimate because they used a through-walls snipe to save seven seconds; a decision which several people balked at.
The discussion about this went on for over a month, in part because everyone had different ideas about what to do about this, and many people were joining into the conversation without being informed about what the trick actually was. Despite this being a high-profile case within the community, the moderators did not have time to supervise the conversation around it, so I took it upon myself to make sure that we did not stall and end up with sniping-through-walls permanently banned. I wrote a doc describing the glitch and its current known uses. I also created a poll outlining what I saw as the potential solutions to this issue:
- Ban sniping-through-walls in all cases.
- Allow it in all cases.
- Allow it in most cases, but ban it in All Lasers.
- Allow it, but redefine All Lasers to require you to activate 11 Laser Box.
- Allow it, but redefine All Lasers to require you to get 11 lasers and then go through Mountain.
Options 3-5 may seem similar, but they represent a difference in ideology when it comes to defining speedrun categories. Speedrunning is, in arguably all cases, arbitrary. People are able to compare their times against one another because someone came up with rules for when the timer starts, when it ends, and what is allowed to happen in between. When something is found that the community does not want to be allowed in a certain category, there are two main ways of handling it: banning it, or redefining the category.
Banning something is pretty easy to understand, but let me explain what I mean about redefinition. Consider this potential rules definition for All Lasers:
Beat "The Challenge" (Secret Area) from a new game. Catwalk Snipe is banned.
This does effectively return All Lasers to the state it was in before October 29th,4 because Catwalk Snipe is required for UTM Skip to work. But what if a different way of getting into Challenge without following the normal route is discovered? Should the rules be updated again, adding another clause banning something? At that point, the category gains more of an "arbitrary" smell to it, even though it's always been arbitrary.
Consider, now, an alternative definition:
From a new game, activate 11 lasers, go through the Mountain, and complete the Challenge.
Just like the previous definition, the player is prohibited from using UTM Skip, and the category remains the same as it was before its discovery. But this time, it's resistant to any other tricks found in the future. This way, the category explains what it really wants you to do, instead of being vague about it, and banning things to make sure you properly infer the correct route.
Option 5 was the most popular answer, which I'm pleased about because it's the one I personally thought made the most sense. The moderators implemented the change on the leaderboard -- but we weren't out of the woods yet. Within the time the community had taken to discuss UTM Skip, one more significant snipe had been discovered.
Keep Shadows Shortcut from Mountaintop
Nothing else in this post is going to be able to top UTM Skip in terms of coolness and impact. But we do still have more to talk about. On November 6th, I finally did what I'd been trying to do months earlier, and I found a way to lower the Keep solve count in 7 Lasers Low%.
In the back of Keep, there is a shortcut door between Green Pressure Plates and Blue Pressure Plates that takes you into Shadows. This door is at the bottom of a staircase, placing it at a lower elevation than the rest of Keep, which made us think for a long time that there was nowhere you could snipe it from. However, that was before we knew that you could solve puzzles through walls sometimes, and it turned out that you could stand on the mountaintop and open that door through the wall of Keep tower.
If you recall from earlier, the reason that Keep Back Laser would always require more solves than Keep Front Laser is because Front Laser only required Hedges 4 + Yellow Pressure Plates, whereas Back Laser required Yellow Pressure Plates, Green Pressure Plates, and Blue Pressure Plates. Blue Pressure Plates was the only real requirement for Back Laser, but there was no way to get to it without solving the other two.
Unless you could snipe Keep Shadows Shortcut.
The shortcut panel, just like the Tower Shortcut panel, is a non-counting panel. So, if you snipe that panel, you can walk into Keep from the side, solve Blue Pressure Plates, and then solve Back Laser. This gives us a total of 2 solves, compared to the 3 that the old strategy used. It is amusing that I did, in the end, find a way to make Back Laser useful in Low% (although sniping the laser itself was still completely unnecessary and infeasible).
Low% categories are interesting. Unlike most categories, lowering the solve count is more important than getting a faster time. The new strategy requires you to go up and down the mountain in the middle of the run, as well as perform a precise snipe through a solid wall. It is undoubtedly slower than the old strategy, but it results in a run where you end with a solve count of 110 instead of 111, which makes it better. Which means we, as a community, had to figure out what to do with the runs that did 111 solves... if Keep Shadows Shortcut Snipe was decided to be legal.
The conversation about 7 Lasers Low% overlapped somewhat with the one about All Lasers. It's a less popular category, so fewer people were talking about it, but I made sure to outline the impact of sniping through walls on 7 Lasers Low% in my doc. The poll only explicitly mentions All Lasers, but that's because All Lasers is the only category in which the glitch is truly controversial. Options 2-5 would all allow the glitch's current uses in 7 Lasers Low% and All Discarded Panels. Not everyone who voted on the poll understood this to be the case, including the moderators, and as a result, the glitch was not legalised in 7 Lasers Low%. The majority of community members seemed to agree that anything fair that lowers the minimum solve count should be allowed in a Low% category, but there was a minor contingent that believed that the trick made the run too difficult to be fun, and that it was more important to prioritise the enjoyment of the run.
The discussion continued, of course, for so long that another panel save was found before a decision was made. A glitch called "invisible boat", which had already been known about for almost a year, was found to reduce the 7 Lasers Low% solve count by 2. This made 108 the true known minimum count. Amusingly enough, using the invisible boat in this run is much harder than the Keep Shadows Shortcut Snipe, and it's much more dangerous too, because while the snipe can be attempted repeatedly until you get it, if you make a mistake during invisible boat (which, by the way, happens near the end of the route), your run is dead. I completed a 108 run, but didn't submit it because of the ongoing debate over both of these tricks.
This was not resolved until August of 2022, believe it or not. 7 Lasers Low% was split into three categories: 117 (the minimum before Latch Skip was found), 111, and 108. Finally, I could submit my run where I achieve Endgame with the true known minimum solve count.
And then Undalevein beat me. And then I beat him again.
And then Boat Storage happened and threw everything out the window but that's a topic for another time.
The excitement over sniping-through-walls finally died down a bit after this. It's still a very useful tool in our arsenal when it comes to hunting for skips, and we can't assume that we've already found everything that could be useful.
Undalevein, of Eclipse Skip fame, found quite a few through-walls snipes following the initial craze. To my knowledge, none of them have seen use in any category, although a couple of them (Tutorial Discard from Mountaintop, and Town Rooftop Discard from Keep Tower) would be usable in All Discarded Panels Low% to reduce the solve count by 3. Since a lower count is more important than a faster time, that world record is pretty much free for the taking -- if you consider learning a run with four very difficult required snipes in it to be "free".5
Desert Shortcut Snipe, the one that Talchyr found for 1 Laser UTM Skip, has also found new life in the 99.8% and 100% routes. When Eclipse Skip first happened, the 100% route changed so that we go into Theatre early, which includes doing the two Windmill puzzles on the way in. This means we don't have to go up to Windmill later, and it gave me the idea of deferring Desert to after Quarry Laser, and then using the shortcut into the tunnels to reach Town. This allows us to get Desert Elevator EP right away, and also allows us to skip either waiting in the elevator or walking out of Desert ourselves. Originally, I thought this would only save time in 100% because we do Theatre early, but darkid did some timing and it turned out to save a nice chunk of time in 99.8% as well. This makes Desert Shortcut Snipe the biggest through-walls timesave in a mainboard category.
Where will things go from here? Sniping through walls is a powerful glitch, and it's unlikely that this is the end of the road. If you looked at the doc with the list of snipes, you may have noticed one out-of-the-ordinary through-walls snipe, found by Rever, entitled "UTM entrance from floor 3":
There's something different about this snipe. Whereas the rest of the tricks I've discussed in this post have involved sniping a panel that's very far away from you, this video shows Rever walking right up to a wall and opening a door that's right on the other side of it. This is a completely different trick than the others. In this case, it seems like the wall right there really doesn't have any 2D collision, and that raises a new question: are there any other walls like this? What new skips could be waiting for us, ready to pop out any day?
To me, sniping through walls is an example of what makes The Witness so great as a speed game. It's a very well-written game -- the dev team sure took their time working on it! -- and for a long time it seemed relatively glitch-free. And even though we've been finding more and more glitches in recent years (a movement very much spurred on by Rever), the new tricks have felt fair, almost like they were intended game mechanics. They add to the run, instead of making them a horrible, miserable, reset-laden slog.
Or at least, that's what I think. I have been told that my point-of-view when it comes to sniping is a bit warped. :)
Laser panels (as well as the Tutorial Gate and the Record Player) are unique in that they generally cannot be sniped. The panels have a built-in distance gate preventing you from solving them if you aren't close enough. There's actually a category extension2 where you mod the game to remove the distance gates and then try to do 7 Lasers; it saves at least 3 minutes compares to regular 7 Lasers. Anyway, both Keep lasers, as well as Shadows Laser and Quarry Laser are exempt from this, likely because they are full fledged puzzles, rather than the straight line panels that the rest of the lasers are. ↩
A category extension is a category that is not important enough or popular enough to be on the main leaderboards. In The Witness, category extensions can be requested by anyone, and category extension runs do not require verification to appear on the leaderboards. ↩
I feel the need to point out that this exact snipe is one of the things I'd attempted while looking for UTM Skip, and it hadn't worked for me. GG to Talchyr for finding a workable lineup! ↩
Which is today, by the way. Happy two year anniversary of the No-Wizzies snipe! ↩
There's actually more that can be saved in this category. Invisible boat, which I mentioned briefly before, would also lower the solve count by 2, so you'd also have to learn that if you want to get the current theoretical minimum. There's also... one more snipe Undalevein discovered that would save a panel, and it's not even mentioned in the snipes doc because it's so ludicrous. It involves going out of bounds on the boat and sniping one of the Theatre exit doors through walls. Doing this at the same time as doing invisible boat almost doesn't seem human viable. Almost. ↩