I decided to do one last "feature-length" post. This one covers a method of doing one of the Accursed Ascension ("Abbot") raid puzzles, the tile puzzle. It is generally considered to be one of the hardest puzzles in DDO, requiring understanding of the puzzle to figure out a solution, playing skills and timing to execute it, and coordination between party members, since the one who can see the puzzle is separated from the one who executes it. In quite a few Abbot raid runs, it is common for most of the group members to complete the other puzzles first and then wait (usually in asteroids) for the tile specialists to complete their puzzle.
However, this need not be the case. Learning how to do the puzzle, and do it quickly, is just like any other game mechanic, in that it simply takes some practice to become familiar. With the advent of Epic Destinies, it is easier than ever for interested players to simply beat down the initial Abbot fight and scout out which dimension door leads to the tile puzzle, in order to learn how the puzzle works and practice solving it and executing it. This is doable even without the methods described in this post, but can be done with any method which is prevalent on each server (different servers have their own conventions).
The goal of the thread is to describe a method for running the Abbot tiles puzzle quickly with a high chance of success. It has been used in Over Raided guild runs, and was also used in Over Raided shortman runs such as here, where we often completed both sides of the tiles puzzle in around 40-60 seconds, enabling us to do all the puzzles in a single round using just two people (and without deaths). Because of methods such as the one described here, in Over Raided Abbot runs, the majority of the time it is the asteroids puzzle that is the last to complete, rather than tiles. It is considerably more advanced than more typical methods, in terms of game knowledge (both in understanding how the puzzle works and in the specifics of the tiles board) and player execution ability (being able to time multiple actions quickly), so it's not recommended for everyone, especially not for people who haven't learned one of the other methods of completing the tiles puzzle yet. But being able to do it successfully means both running tiles quickly and having a very low chance of failure, which is why it was developed.
I'm not the first person to discuss the Abbot puzzles, of course. There's an introductory visual guide for the entire raid here, for example, and a more detailed one here. I recommend looking at them if you want more information about the raid. There are also a variety of Youtube videos of people doing the tiles puzzle as well. So my overview will be a bit cursory.
The tiles puzzle consists of platforms which phase in and out. When they are phased out, they don't support players, so players will thus fall to their deaths. Contrary to what I've mentioned before, the timing of the tiles is actually 8.5 seconds visible, 3.5 seconds fading to red, and 4 seconds phased out (as opposed to the 9/3/4 that I said previously, although this doesn't make a difference in the phased timing and only in the amount of warning the player gets). Some tiles will never phase out and are marked in blue. The objective is to reach the other side of the room, by stepping on the platforms only when they are phased in.
Complicating this, the puzzle is split into two halves. The tiles are normally invisible, but each side has a pair of goggles which allows the player on that side to see the tiles on the other half. Thus players must guide each other across by communicating when to do each action, as there is a barrier in the middle preventing them from going to the other half directly. Since the tiles are invisible, the person who is on tiles will look to himself as if he were floating in mid-air. The tiles board is 4 tiles wide (which I'll call "lanes") and 12.5 tiles long (which I'll call "rows"), the .5 because the last tile is only about half as long as all the other tiles -- but stepping on it when it is phased out will lead to a death just like with all the other tiles.
An example board can be see here:
http://www.youtube.com/watch?v=qTmPvFBKicQ (from the initial part of the puzzle facing the end)
http://www.youtube.com/watch?v=WYxWO2leC8g (from the middle part of the puzzle facing the end)
http://www.youtube.com/watch?v=BtRs7FxXQiY (from the end part of the puzzle facing the beginning)
Note that though this shows the tiles board on the same side, normally you only see the tiles on the other side; the videos were set up this way specifically for the purposes of this thread with the help of guildies.
For brevity, I will call the person running across the tiles the "runner", and the person with the goggles the "guide". There are two primary ways that the runner will cross the tiles: either by facing the guide, or by facing the end. For brevity, I'll call facing the guide the Khyber method while facing the end will be called the Orien method, even though the methods aren't necessarily server-specific and I believe both are used on all servers; it just depends on the preferences of each Abbot-capable group (particularly, how the people doing tiles prefer to do it). There are advantages and drawbacks to each method.
The runner faces the guide. As the guide moves toward the end, the runner moves sideways along with him, stopping when he stops as well. This ensures that the runner stays near the middle of each tile, since he is visually using the guide as a marker. The guide can also visually give commands for switching lanes by jumping or attacking. The advantage of this method is that the communication lag is relatively small, since the game will update each player's movements to other players very quickly, compared with the other communication methods described below. Thus, the response time between when the guide gives a command to when the runner executes it is relatively small. The disadvantage of this is that the runner will move more slowly across the board, since sideways movement is slower than forward movement, and he cannot use forward-traveling abilities such as the favored soul Leap of Faith or the monk Abundant Step (which I'll call "wings") to move across the board more quickly. He also cannot easily see if he's lined up in terms of which lane he's in, although I believe there usually aren't enough lane changes to make this an issue.
The runner faces the end. He can line himself up in terms of which lane he's in by looking at the columns on the initial side and at the end side (with draw distance set to maximum). To line himself up in terms of which row he's in, he uses the guide's dot on the minimap and aligns himself with the dot. To know when to move forward and when to stop, he can use the minimap or more commonly, via commands given over voice. The advantage of this method is that the solution can be executed more quickly; not only does the runner move faster running forward compared with running sideways, but he can also use wings to speed things up. It also allows the guide to disregard some missing tiles, since the runner can fly over them with wings, allowing for solutions with fewer commands and hence, simpler solutions. The disadvantage is that there can be substantial lag between when the guide gives a command and when the runner notices it, in terms of both the minimap dot and in giving commands over voice. Thus, the runner needs to have fairly good reflexes. In guild runs, to counter this, we will occasionally test out the voice lag by having the guide saying "jump" and then seeing how long it takes for the runner to jump, so that the guide can anticipate this lag in timing when to give out his commands. This has the advantage of also accounting for the runner's reaction time.
To begin, the following is an animated gif of the tiles board in the above videos, with the start on the right and the end on the left:
Here's a video of me running around the board, showing that the animated gif is indeed synced up with the board:
For the tiles that phase in and out, they are present for 12 seconds and phased out for 4 seconds, so they cycle every 16 seconds. The state updates once per second (although the visual fading to red is updated smoothly). Out of the 16 seconds, there are 11 different possible seconds where a tile may appear, so for 5 of those 16 seconds, no tiles will appear and start its 16-second cycle during that second. Those are spread out fairly evenly; the timing for when a tile can appear looks like:
1 2 3 - 5 6 - 8 9 - 11 12 - 14 15 - (16-second cycle, no tiles will start its cycle on the 4th, 7th, 10th, 13th, or 16th seconds)
Some of the tiles will be permanently present; I've seen anywhere from 0 to 10 on one side, so I'm guessing there's about a 10% chance for a given tile to be permanently present, but that's just a guess. Also, note that usually, the runner can run at a speed of about 1 tile per second, depending on his run speed modifiers. Nowadays, just having 30% striding boots or quiver is sufficient for this. In most of the gifs below, the top border flashes on/off every second, to give you an idea of the timing when looking for a solution.
To analyze the tiles board and compare the efficacy of different solution methods, I'm going to make the assumption that a runner's chance of success is relative to how often he moves onto a tile, i.e. that failures generally happen when moving from one tile to another, rather than from staying on one tile for too long. (so the runner is on each tile for short enough a time that it disappearing isn't a big concern.) A more sophisticated way would be to look at how long both tiles are present, i.e. when one appears to when the other disappears, since that's the margin of error for when the runner has to get from one to another, but that makes the probabilities a lot more complicated. Another way would be to say that each command (move forward, stop, switch lanes, etc.) has a chance for failure, so the less commands the better. Anyway, so using the Khyber method, a solution for the board might be:
A couple of other solution examples from the runner's view might be:
Again, note that normally, the runner (my character in the video) does not see this; the tiles are normally invisible to the runner, and the video was made under special conditions. All the actions shown -- running to a tile, stopping, switching lanes, etc. -- have to be communicated from the guide to the runner, at the right time, and the runner has to ensure that he stays lined up on the intended tile throughout, with help from the guide.
At a minimum, the Khyber method will involve the runner going across 13 tiles (counting the last half-tile as a full tile, since he can fall to his death if it's phased out just like the other tiles). Although communication lag is less using this method, there are a lot of tile crossings, each potentially causing a solution failure.
Now if the runner were facing the end, as with the Orien method, he could not only run across, but he can also use wings. A wing on its own is a distance of about 1.66 tiles. However, the runner can jump forward, wing in midair, and continue moving forward as he comes down, in which case he will travel just about exactly 3 tiles. This may take a little bit of practice, but it's actually not that hard to learn, and can be practiced anywhere (you don't need to be in tiles to do this). In fact, Over Raided members who do tiles are all expected to be able to do this consistently. For added safety, the guide will usually align with the target tile before telling the runner to wing, to help the runner line up and end up on the middle of the tile when he lands. Using wings, a solution may look like this:
In the animated gif above, a "thin" rather than square runner icon indicates that he is in the air, and thus can skip over tiles that are phased out without falling to his death.
Another solution from the runner's perspective may look like this:
This requires more from the runner, since he has to learn how to wing a set distance, and have fairly quick reflexes in gauging his position based on the minimap for a safe landing. But it simplifies the solution considerably; the runner will only need to land on 4 tiles, usually the 3rd, 6th, 9th, and 12th, but sometimes the 2nd, 5th, 8th, and 11th instead. So it makes things easier for the guide in terms of picking out a good solution, at the expense of requiring more from the runner. Although the animated gif doesn't show it, the runner can always switch lanes, but in practice we only do that occasionally since usually a single-lane solution can be found.
Of course, the runner doesn't have to just stand there while waiting for wings to recharge. He can run across safe tiles in between winging, so maybe something like this:
And yeah, in the previous board the runner ran to the 2nd tile anyway. So this really just opens up the solution space even more, allowing for potentially better (easier and less prone to failure) solutions, but once again, requiring more from the runner, since it's not just a matter of winging the right distance, it's a matter of winging the right distance as well as running in between.
The addition of the Cannith Boots of Propulsion, which gives a one-time wing every 10 minutes, allows for the runner to start off with a "double wing". From the start, as soon as the runner jumps forward he uses his regular wing, then as soon as it completes he uses his boots wing while still in midair. The initial jump is the only time where this can be used, because the platform is higher up than the tiles, allowing for just enough air time to squeeze two wings in. This is much more difficult than just doing one wing at a time, with potentially disastrous consequences if the runner mistimes something or server-client lag causes one of his actions to not "take", such as it not registering his jump prior to his first wing (in which case he simply wings directly off the platform and potentially to his death). So it requires a lot more from the runner to do this.
However, this can simplify the solution space even more. By using a double-wing, the runner can reach the 5th row. Effectively, rows 1-4 don't matter at all; the first third of the tiles board can be ignored. So the solution may look like this:
Another solution from the runner's perspective may look like this:
Again, the runner can also walk when safe to speed it up, such as this:
At this point, the runner will only be on 3 tiles, out of the 13 rows of tiles, so over 3/4 of the board can be eliminated from consideration by the guide. The guide only really has to look at the timing difference between the 5th, 8th, and 11th rows, and select the lane that give the largest margin for error. As with before, this opens up the solution space even more, allowing for more possible solutions, and potentially better ones.
Now the above can be done by any character with wings and 30% striding. However, if the runner has barbarian or ranger sprint boost as well, and possibly with monk levels for even more speed (although I'm not sure if monk levels are necessary, but can't say conclusively since I haven't tested it without monk levels), then he can actually reach the first half of the 6th tile with a double wing. Also, from one tile to another, his wing will travel about 3.5 tiles. What this means is that the runner can start off with a double wing to get to the 6th tile, walk forward a bit on it, maybe about half a tile (he shouldn't walk forward enough to reach the 7th tile though since then he might fall off), then wing and he'll land on the first half of the 10th tile. From there, he can again walk forward about half a tile, staying on the 10th tile, and then wing directly to the end. Normally, a wing from the 10th tile to the end is not advisable since the end platform is higher than the tiles board, so the runner loses some distance on this, and might end up bumping against the lip and falling to his death if the end half-tile is phased out, but with sprint boost it's highly doable. A solution may look like this:
What this means is that the runner only has to land on 2 tiles out of the entire board. The guide only has to look at the 6th row and the 10th row, and pick out a lane that has the best timing (ideally, the 10th row phases out about 2-3 seconds after the 6th row, if neither are permanently phased in). Once again, this requires even more from the runner; he has to have the right build for this (sprint boost and possibly also monk levels), be able to time a double-wing, as well as be able to judge how far forward to move on each tile (remember, he can't see the tiles normally, so it's done blind) to move to its forward half without falling off, in order to reach the next tile or the end successfully. When Nix and I did our 2-man hard Abbot, both of us were monks with a ranger splash, in order to be able to do this.
There is one final piece to doing Abbot tiles, and it is the namesake of this thread. Everything so far has assumed that the runner will try to stay near the center of a given tile, so that he doesn't overstep its boundaries and fall to his death. But the runner will in fact be supported as long as he is anywhere on the tile, including its edges. If a runner should land directly on the crack in between two tiles, he will be safe as long as either of the tiles are phased in. So he will fall to his death only if both tiles happen to be phased out. So if the runner can stay on the crack, then the guide's job is even easier: rather than looking at the 6th and 10th rows and choosing which lane for the runner to go on, he just needs to look at the crack between the tiles in those rows, and ensure that it doesn't have a "double blank" or where both tiles are phased out, when the runner is on it. As long as either of the tiles are phased in, then it is a safe path.
Now, it is difficult for me to overstate the difficulty of doing this. This is not recommended for beginners. In our 2-man Abbot runs, this was the single part of the raid that required the most practice, until we could do it relatively consistently (and quickly!), and we already had plenty of Abbot tiles experience under our belts before attempting this. We had to practice not only timing the double wing correctly, but also being able to gauge on the fly how much forward we should move on each tile, using nothing but our surroundings and a sense of how far we've really traveled to guide us. In other words, estimating how far we were based on seeing how big the end platform was on our screens. However, when done successfully, the results were spectacular: we could get both people across within about 40-60 seconds, enough time to also do ice and then enter asteroids before the puzzle doors went down (which was about 90 seconds after they spawned).
There is a further benefit: For a given pair of tiles, if their phases are 4 or more seconds apart, then there will always be a tile present on that crack, which means that (with some limitations) the runner can cross it safely at any given time. Thus, the guide doesn't even need to issue commands and time each movement for the runner, other than to simply specify which of the 3 cracks to use. The guide can then go off and do other things (namely, get into position for his own run, or start running to the end) while the runner runs along the crack.
In this particular board, note that for the bottom crack, there is always one tile present in the 6th and 10th rows. Thus, the guide just needs to tell the runner to use that crack, and the runner can then be left to his own devices, and be assured of success. This is shown below, where the runner starts a run every 9.6 seconds regardless of the timing; as you can see, he is always safe:
Even if the guide doesn't say which crack to use (i.e. the runner does this blindly), if he chooses the top crack, then there is a 1-second interval where both tiles are phased out for the 6th and 10th rows, so he has about a 88% chance of making it across, without any guidance. If he chooses the middle row, then there's a 1-second interval for the 6th row and a 4-second interval for the 10th row (both tiles are perfectly in sync), so he has about a 70% chance of making it across. So for this particular board, he has about an 86% chance of making it across if he chooses which crack at random, without any help from the guide, in other words doing it blind.
Let's assume for now that the timings for the tiles are independent of each other and that there is a 10% chance of a given tile being static and permanently phased in (although I don't have definitive proof of either of these). So right off the bat, there is a 19% chance that at least one of the tiles will be permanently phased in and that the row will always be safe for that crack. Of the remaining 81% of the time, the runner will be safe as long as either one (or both) of the tiles are present. A facile analysis, since a tile is phased in 3/4 of the time, would be that the runner is safe 1 - (1/4)*(1/4) = 15/16 = 93.75% of the time. Taking into account that there are actually only 11 different tile timings (as opposed to 16) would yield about 93.6% of the time, or that he fails about 6.4% of the time. Thus for a given landing, the runner's probability of failure is 81% * 6.4% = 5.2%, or he'll succeed 94.8% of the time. Since the runner only lands twice, then his overall probability of success for the entire crossing is basically 90% of the time. And this is blind, without any guidance from the guide, and it's likely comparable or better than the success rate of regular "guided" methods, while taking only a fraction of the time. In fact, based on those assumptions, the probability of making it across blind are:
Obviously, when running completely blind would already succeed 9 times out of 10, having a guide specify which crack to use, and possibly when to start if needed, means the crack method will virtually always successfully complete if executed correctly. For the guide, the solution search space has changed from finding one contiguous path through all 13 rows of tiles, to simply looking for (and avoiding) any double blanks in the 6th and 10th rows. Below is a video of all this in action, with the runner wearing the goggles so you can see the 6th and 10th row jumps, as well as being on the crack:
Method Tiles Success Desc
Khyber 13 3.6% Walking straight across
Orien3 4 36.1% Winging onto 3rd, 6th, 9th, and 12th or 2nd, 5th, 8th, and 11th rows
Orien5 3 46.6% Double winging onto 5th, 8th, and 11th rows
Orien6 2 60.1% Double winging onto 6th and 10th rows
O6Crack 2 89.9% Double winging onto 6th and 10th rows, crack method
As I mentioned above, this particular crack always has a tile present in the 6th and 10th rows, so I can do this completely blind. Here are three videos of that:
The videos also show what it looks like, in terms of gauging how much to walk forward on the tiles based on your surroundings.
Original Elite Tiles
One final note on this. In Update 13, Turbine inadvertently changed the timing of the tiles, so that (I assume) they reverted to their original timings which differed by difficulty, rather than the current version where the tiles puzzle is effectively always on normal difficulty regardless of the raid difficulty. The timing of the tiles on elite was 7 seconds on and 7 seconds off, so it was phased out 50% of the time rather than the current 25% of the time. A video of the tiles setup is here:
An animated gif of the map is below:
It's understandable that people complained about how difficult this was. Certainly it seems impossible using the Khyber method, without any jumps (although there's a solution that predominantly uses the top half of the board, but with multiple tile crossings which have only 1 second between when one tile phases in and the other phases out). However, it's readily doable using any of the Orien methods. Examples are below:
Orien method, 3-tile wings:
Orien method, 5-tile double wing:
Orien method, 6-tile double wing:
Orien method, 6-tile double wing, crack method:
As the runner's abilities progressively increase and more can be expected of him, the solution methods become faster and less convoluted, with a greater chance of success. For the original elite version of tiles, blind running across all 13 tiles would have a success rate of 0.012% (less than 1 in 8000), but blind running on a crack with two landings would have a success rate of 56.25%. By comparison, blind 5-tiling (landing on 5th, 8th, and 11th rows) on the current tiles has a 46.6% chance of success, which is lower. Of course, with a guide saying which crack to use and when to do each wing if needed, the success rate would have been very high. Although it would probably have been slightly more difficult to find a solution, a guided crack run would still have had a very good success rate under these more challenging conditions, showing the effectiveness of this solution technique.