From rolling ancient tunnel maps to completing the grail

[Albatross]

I'm updating the guide due to recent discoveries. Would have done it sooner had I not been occupied with rerolling my map. I've found a good one some 30 hours in, but it turned out it's, at best, on par with my last one. Even with everything I now know about AT, it's still very hard to find that "perfect" map.

A short summary of updates will be in this post, so those who read the whole thing don't have to re-read everything. So... buckle up, we're going in with a quote!

The exact mechanics appear to be unknown and I've come to terms that I'll not be the one who discovers the inner workings behind density distribution. This might have been uncovered at some point, and lost in the void that is the Internet, and may yet again be revealed by code diggers.

And so it came to pass that the code digger, @Fruit, brought to light those exact inner workings.

Several people already figured out that the number of upgraded spawns have something to do with the path you take. Thanks to Fruit we now know more about how this works.

Here's the summary of guide updates:

This is what happens when you go down the drain:

• the area around the entrance (red) will never spawn anything to prevent stair traps, in AT at least... and it's about the size of the segment
• as soon as you jump in all segments adjacent to the one you are in will be populated with monsters and some of those monsters will be upgraded to uniques (80%) or 1-4 champion groups (20%)
• when you cross the border of another segment, all the segments which are adjacent to it will be populated

There are two important variables which control the spawn upgrade process: segments_spawned and total_segments. As shown in the image segments_spawned will increment for each segment which is adjacent to the entry point. After each increment the upgrade check function is called for each group of monsters which were spawned in the segment:

• segments_spawned * 100 / total_segments

The result is compared to a random roll 0-99 and if it's less than 20, a group of monsters will be upgraded to uniques or champions. It's still unknown in which order this plays out, but let's take a guess and say segment #4 would get upgraded first.

segments_spawned will equal 2 (because the initial segment is also counted), while total_segments is 11 for this map.
This gives segment #4 18,18% change to upgrade any of it's spawns. Let's assume segment #5 is next. The following would apply: 3 * 100 / 11 = 27,27%, etc. This will continue until the minimum upgraded monster count condition is satisfied, which is 4. Afterwards, a simple 6% chance is applied to each consecutive spawn until the maximum count reaches 8. The maximum count does not include the mandatory chest spawn, this is why some maps are able to exceed the cap and give 9 packs. It's also very important to note that each champion counts individually. So not only are their drops worse, but they'll eat up the maximum possible upgraded packs and there's nothing we can do about it.

With all this in mind, I drew the following conclusions:

• the starting room can't spawn anything
• a single room can spawn up to N packs, but anything above 5 is exceptionally unlikely (I've never seen it happen)
• the chest room may spawns next to nothing, barring the default pack (unless it's bugged and fails to spawn it)
• champions spawn instead of uniques 20% of the time, confirmed both by my findings and the Amazon Basin Wiki
  
• a map can spawn no less than 3 packs (chest spawn included)
• a map can spawn no more than 9 packs
• maps which have their chest room adjacent to the entrance will never be able to spawn 9 packs because the upgraded monster counter is "wasted" on the mandatory spawn
• it's still not necessary to finish your run with the chest room, just avoid spawning it for as long as possible to have the best chance of reaching 9 packs

I'm still fairly certain that 10 segment maps are superior because they have a greater chance to "burst spawn" up to 4 packs early in the run, hitting another 2-3 before the chest room shouldn't be uncommon. Having several packs in one room does wonders for efficiency and you'll want to be able to reach the maximum number with as least segments visited as possible.

You don't necessarily want the chest room to be at the end of the run, but you certainly don't want it to spawn early. Try to find a path which will spawn at least 4-5 rooms before the chest room is in range. During focused runs, you can count the number of packs you kill and see if you reach the maximum or maximum -1 your map can spawn by clearing only 2-4 segments, then save and exit immediately. This way no time is wasted teleporting through empty rooms just to reach the chest pack.

Here's a quick summary of what to look for:

• best Rogue camp layout, 1-2 TPs from act II WP to entrance
• 10 rooms
• the chest room is not adjacent to the entry point room
• the layout allows not spawning the chest room early in the run
• 4 minimum spawns
• 9 spawns are possible
• full clear density is >6
• almost no double backing
• 1-2 shrines, both on route
• the first boss is only 2-3 TPs from entrance
• the chest room is not the final room you visit

So be careful when activating, but you should still examine those unlabelled because they might roll into monster shrines which produce either a unique or a single champion if there's a normal monster nearby. Although I cannot attest to their drop odds, the converted monsters should be fully capable of dropping TC87 items. However, you should also take into account that successful upgrades from shrines increment the total packs count and can actually lower your overall density, unless you use them after you've activated all of the segments you'd normally run. This could make a "10 pack" run possible.

After tens of thousands of runs, I've come to the conclusion that 105 FCR becomes better as your gear progresses. While there are diminishing returns on magic find, you'll always cast each spell 0.04s faster with 105 FCR, compared to 63 FCR. This might seem a minuscule amount, but considering that at least 50 spells are cast during a run the total saved time would amount to 2 seconds.

 

[Fruit]

Very glad I was able to contribute a bit to this awesome guide!

 

[art_vandelay]

@Albatross So if every individual champion counts towards the bosspack limit then what if during your 20 (or whatever) test runs where you count bosspacks you get an unusually high amount of champions and you trash the map because it has 'too few' bosspacks?

In fact any amount of champion spawns greater than one could ruin the count we are most interested in: # of regulars upgraded to champion/unique

This probably means that the only way of finding the true amount of bosspacks possible on a map is to only count runs without champion spawns. I.e. finding a map becomes this:

• Look for a good layout (as described in your update above)
• Count bosspacks on all runs without champion spawns.
• Also count number of champion + unique packs (i.e. champions not individually) on maps where champions spawned but the maximum number of bosspacks is not reached (for that criteria we count champions individually though)
  
• Keep if those numbers are high enough on average

This would be an absolute pain since the chance for a champion pack is still quite high.

 

[Bukama]

Thank's for the great guide!

I've read this guide at the weekend and instantly searched for a new map. Found one with 3 ports from the waypoint and a nice circle with chest in the middle (but in the last set beeing filled with mobs).

But there's one thing I don't understand: Why rogue camp? I port back to act 4 where the way from spawn to waypoint is one telekinese cast?

 

[Albatross]

@art_vandelay I've gone through the same thought process and came to the following conclusion. It's still somewhat painful, but not as bad.

You count champions individually and divide the number by 2.53 to get a unique equivalent, then add the number to the uniques total. See my latest spreadsheet for an example. I've also experimented with minions counts in this sheet, but this varies wildly so I've just used 3.5 per unique as approximation. Unless my math is way off (which is very possible ) this should be the most accurate method of determining density and counting killed uniques/champions in the way we are interested in.

There's one more thing to mention about unique/champ count, it's possible for a map to exceed the limit, depending on the order of upgrades. Best theoretical scenario: the map upgrades 7 groups to uniques and the last attempt yields 4 champions, with the chest spawn you'd hit 12 total, laughing at "MonUMax"'s value of 8. As long as the total number is less than 8, the map will attempt to upgrade a spawn, but wont cap the champion count. Anyway, the highest I've seen was 11.

I also have a feeling there's some interaction between monster shrines and the count, but not in a way which would benefit runners. I'm still far from drawing any real conclusions.

@Bukama thanks!

You go back through the AT entrance to the waypoint and jump to act 4? That might work on some maps, but I doubt it's faster than save & exiting in AT, then talking to Warriv in the new game to go to act 1 where you are a few steps and a TK from the waypoint.

Edit: here's an example, the linked video starts a few seconds before I finish the run and S&E. Then start a new game, talk to Warriv and port to Lost City from the Rogue Encampment WP.

 

[Bukama]

Wow - I've would never came up with the idea to talk to warriv to go back to camp. Thanks!

 

[Albatross]

It's an old practice so I've taken it for granted. I'll include it in the guide with another update.

 

[ioupain]

It's an old practice so I've taken it for granted. I'll include it in the guide with another update.

It is known...

 

[Kefir-Tribe1]

Thank you @Albatross for a great guide! And thanks to @Fruit and others who contributed. I have a few questions:

After each increment the upgrade check function is called for each group of monsters which were spawned in the segment:

• segments_spawned * 100 / total_segments

The result is compared to a random roll 0-99 and if it's less than 20, a group of monsters will be upgraded to uniques or champions.

I don't quite understand the explanation of the algorithm here. Is it the result of the formula segments_spawned * 100 / total_segments compared to a random roll of 0-99? If so what does "if it's less than 20" mean? A comparison would return a boolean true or false.

Are you saying the upgrade check is run on all groups of monsters in one segment before it moves on to the next segment, or does it perform a first check in one segment, jump to the next, then jump back if there are no more segments, etc.?

When the player moves to a new segment, does segments_spawned keep being incremented, or is it reset and then incremented as the upgrade check works on segments adjacent to the characters new position?

Once you've documented a 100 runs, calculate how many packs each room spawned and add this information to the grid. This will help determine your run route which will hopefully resemble something like the following image.

In white, the percentage of pack occurrence, while yellow stands for pack % with the chest spawn excluded:

If I understood it correctly, the percentages in the map indicate "out of all boss packs spawned during test runs, x% spawned in this room". Why did you choose this statistic? Could an alternative be to calculate average number of boss packs spawned per run in each room?

When collecting statistics on a map, do you follow exactly the same path in all test runs? Do you think it is worth collecting statistics on several different paths if the layout allows, or is it better to just traverse the rooms using the most efficient path and reroll the map if the results aren't good enough?

Isn't there a risk that skipping "unproductive rooms" could tilt the upgrade check algorithm to roll less boss packs in the good rooms?

the chest room is not the final room you visit

What's the motivation to not visit the chest room last?

Hold down the "highlight items" key while teleporting to avoid getting stuck

Is this to avoid name locking things with teleport?

Finally, what player setting do you recommend running on?

 

[Albatross]

@Kefir-Tribe thanks! Let's see if I can answer all of these...

I don't quite understand the explanation of the algorithm here...

I should probably reword that bit, it was written in a hurry. This is what I believe happens, but can't guarantee anything since I've never gone through the code and am only interpreting Fruit's findings.

You go down the drain and all adjacent segments are spawned. Next, the regular monsters are randomly positioned throughout the segment. Checks are performed on tiles, it's possible we can to some extent influence this distribution by taking different paths. Upgrade checks go next, segment by segment. This is where the segments_spawned * 100 / total_segments formula applies. The 20 in the less than 20 part is an example value if only 2 segments were spawned on a map with 10 segments total. The game rolls a random number between 0-99 and if this number is lower than segments_spawned * 100 / total_segments a pack will get upgraded.

The segments_spawned never goes down and the upgrades should go segment by segment, no skipping.

Could an alternative be to calculate average number of boss packs spawned per run in each room?

Sure, a different approach can work, as long as you know which rooms are valuable and should be in your path. Knowing what I know now about AT, I'd probably just stick to "heat maps". It will tell you more about where you should move.

Yes, I test run the map a bit and try to find a nice route. Test 50 runs, then try another 50 counter-clockwise or slightly different path, if the map shows promise in the first test. The risk is rather low once several rooms are spawned, besides your aim is to "force" spawns on the route you take by avoiding to trigger too many rooms. If there's some spawn happy room, but it's really out of the way you'll probably be less efficient by clearing it than skipping it.

What's the motivation to not visit the chest room last?

If you don't normally end your run with the chest room, there's a chance the map will spawn max or close to max by the time you reach it. You can save & exit immediately after that and ignore the segments you'd normally check after the chest room. In the long run, this will make your runs more efficient.

Is this to avoid name locking things with teleport?

Finally, what player setting do you recommend running on?

Yes, it's annoying and slows you down.

P1 for shiny stuff, P+ for exp if you like AT. There are better places for a blizzard sorc to level up, if that is you goal.

I don't have access to my game at the moment, so here's a really bad drawing instead. I think it's easier to explain all of this with an image so here's one type of map I consider "ideal". From entrance to chest room, you won't trigger any rooms out of your route and by the time you reach the chest there a very high chance everything that could have spawned already has in those 7 segments. Of course, if you encounter and kill 6-7 uniques when you reach the chest, just S&E and skip the remaining 4.

View attachment 52219

An even better map wouldn't even have the segment where the arrow ends. It would leave you with 6 rooms to spawn stuff, which should be enough for max and likewise, only 6 segments to check. A map which seems to spawn very little in that same room would also be great.

 

[Luhkoh]

This is fascinating. I understood the spawn mechanics stuff (i mean mostly anyways ) but your conclusion with that last drawn map is something I would not have come up with and it makes tons of sense.

 

[Peytron]

First of all, thanks for the guide, lots of useful information here!

Second, this is slightly off-topic, but I have to ask how everyone is getting these haul pictures?

 

[Albatross]

This is fascinating. I understood the spawn mechanics stuff (i mean mostly anyways ) but your conclusion with that last drawn map is something I would not have come up with and it makes tons of sense.

It comes naturally when you do >40k AT runs There might be better layouts though, it's not like I've checked this mathematically. Really hard to roll one though and even if you do it can still have mediocre density and won't be a record breaker.

First of all, thanks for the guide, lots of useful information here!

Second, this is slightly off-topic, but I have to ask how everyone is getting these haul pictures?

Some people self MP and haul the stuff in with another character. I just "photoshop" 2-3 images (usuallywith paint ) together. You should be careful though, even unique items disappear after laying on the ground for 30 minutes (I think).

 

[Peytron]

It comes naturally when you do >40k AT runs There might be better layouts though, it's not like I've checked this mathematically. Really hard to roll one though and even if you do it can still have mediocre density and won't be a record breaker.



Some people self MP and haul the stuff in with another character. I just "photoshop" 2-3 images (usuallywith paint ) together. You should be careful though, even unique items disappear after laying on the ground for 30 minutes (I think).

Nice. Now I see two Warrivs! You should crop Waldo in here some time too .

 

[Kefir-Tribe1]

Thank you for your reply @Albatross

 

[Peytron]

@albtatross do you think this is a decent map for AT?

I'm not an expert AT runner and I don't spend much time rolling maps. The red X marks where uniques/champions usually spawn and the gold X is the sparkly chest. The ugly green line is my path through the map. It is pretty consistent except for the southern most pack which is somewhat rare to spawn. I could speed my runs up a bit by just ignoring that last stretch I suppose, but do you think it is worthwhile to try to "reroll"?

 

[Albatross]

@peytron depends on your goals. If you plan to spend hours in AT, then yes definitely.

You have that unfourtunate type of map which will spawns the chest unique as soon as you go down the hatch. Normally, you'd want at least one room between the starting segment and the chest. It also looks like you have to visit a few too many rooms to reach the kill quota. Ideally, you should be skipping at least 4 segments on an 11 room map.

 

[art_vandelay]

@peytron You should try some different routes and see which boss pack spawns persist, for example if you take this yellow route:
View attachment 52438
After you reached the end of the yellow route you could check the remaining route (green) to see whether this new route spawned anything in the southeast. If not then you would be good to use the new route and you get like 5 seconds of free timesave. (no idea what your times are right now, but that should help a lot)

I routed the first arrow directly to the chest because as Albratross said the chest pack spawns immediately after entering the AT anyway so why delay the inevitable

What's your 'Lost City + Rogue Camp' like? If the 'Lost City' is bad for you then it doesn't matter anyway and I wouldn't waste time optimizing this map if I were you.

 

[testyfish1]

Related question: How preferable is Blizz to FO? I feel like FO does obviously less damage, but at high gear levels, how much of a time difference is it? Anyone have experience with both in AT?

 

[Albatross]

Besides the low damage issue, you can't really hit stuff reliably with FO in AT because there are many... well, tunnels High end gear won't help much because you want MF, not +skills.

I remember someone testing Frost Nova. The area of effect was nice, but like FO it just didn't do enough damage.