Analysing the Thargoid Simulation

[Ian Doncaster]

As suggested by Randomgazz, here's Invasions not in their last week (as final week invasions have a very different activity balance available). That makes the correlation stronger - again, surprisingly good given the weaknesses of using EDDN FSDJumps as a proxy measure - and it'd be even more solid without the two 1400ish outliers.

 

[Disemboweled Ego]

I tried bruteforcing a couple progression pips in the Jementi alert basically solo because it's near our home system and this is what it took based on the journal logs:

Spoiler: My actions during the first progression pip

Missions completed:

• Mission_TW_PassengerEvacuation_Alert_name: 120,094,798 (49)
• Mission_TW_Collect_Alert_name: 49,973,180 (215)
• Mission_TW_CollectWing_Alert_name: 19,756,409 (8)
• Mission_TW_Rescue_Alert_name: 5,076,993 (53)
• Mission_TW_Massacre_Scout_Plural_name: 4,977,919 (12)
• Mission_TW_Massacre_Cyclops_Singular_name: 3,000,000 (1)
• Mission_TW_Massacre_Cyclops_Plural_name: 2,861,504 (1)

Total profit: 205,740,803
Total missions: 339
+Inf: 1435

Total cargo delivered: 19,828 (of which 1806 were Passengers/Refugees and 157 were Occupied Escape Pods)

I took random rewards initially with rep/low tier materials I wasn't maxed on and mostly G5 mats for the passenger missions. There's ~30-50 scout kills and 1 cyclops kill and some completed AX ship repair events in the system too, but that shouldn't affect the data much. Most of the kills for the scout massacres were done in a CZ in Palanti after stacking 5x wing scout massacres and a few low kill count ones. I kept an eye on traffic and it was 80-90% me with less than 70 ships over the period and nothing that looked like repeat traffic doing missions.

Based on this I tried testing a theory that it's just ~250m profit per pip, but after an additional 253,202,696 profit from 33 passenger missions there wa no second pip. I kept doing passenger missions picking money rewards until 500m profit. Thinking it could be slightly higher due to my estimation being off or much higher if the progression per mission is based on the base credit reward even if I take the mat/rep reward - that amount could be calculated from the journals, but it'd take more effort to compute.

After 500m profit from passenger missions I switched to picking inf rewards along with picking up occupied escape pods during the passenger runs for more inf+/missions and finally observed another progress pip after this much total effort:

Spoiler: My actions during the second progression pip

Missions completed:

• Mission_TW_PassengerEvacuation_Alert_name: 685,597,030 (112)
• Mission_TW_Massacre_Scout_Plural_name: 12,543,010 (14)
• Mission_TW_Rescue_Alert_name: 6,030,747 (76)

Total profit: 704,170,787
Total missions: 202
+Inf: 638

Total cargo delivered: 6,951 (Passengers 6741, 210 Occupied Escape Pods)

After this I did ~300 inf from various missions for the ruling faction to test if only that matters and because inf gained for the ruling faction was the only thing close in both datasets, but there's no 3rd pip yet and I probably won't push it much more.

Not sure what to make of this overall:

• Is the first pip actually 0-2 and the 2nd 2-3 so the first one takes roughly twice the effort to move?
• Does inf/missions completed matter or bulk/value delivered?
  • Was it skewed towards passengers by the military economy mostly requesting low-value (but difficult to source) cargo?
  • Does actually picking inf for mission rewards do anything?

Doing bulk cargo to counter alerts might be feasible it if you can park a few carriers next to an large orbital station pre-emptively (since you can't carrier jump into alert systems currently). Preferably pre-loaded with various in-demand goods, but even loading them up later would still be more effective in most cases.

 

[Ian Doncaster]

HIP 23716 completes its repairs and goes directly into Alert state, so it looks like there's no need for an intervening Safe week.

 

[yttrbio]

New week, new exciting notes.

4 more invisible control systems for Leigong, bringing the total to 21:

Hyades Sector MY-I b9-0
Hyades Sector OT-I b9-1
Arietis Sector LM-W c1-14
Arietis Sector NX-U c2-12


The watch systems I referenced here have behaved according to pattern. So a 4-week cooldown on alerts seems likely, and seeing as how HIP 23716 went into alert immediately after its repair, it's likely to occur immediately after every 4-week repair period. This doesn't necessarily mean it's better to wait for invasions to progress far enough to force a 4-week repair period, because it still seems quite possible that there is a certain number of alerts each week, and forcing repeat alerts on nearby systems will slow the growth range of the maelstrom.

The 10 ly range continues to hold

My prediction for this weeks alerts was 67, and the actual number was 65 (+ an unknown number of Leigong alerts). The way I did this prediction is by fitting a line as if the number of new alerts is inversely proportional to the number of new control systems in the week. The line fits quite well so far, counting only the weeks when there were 8 maelstroms. But remember: Correlation is not causation (there may be some other factor that correlates with both of these), and the January 12 number may be low, since there may be invisible Leigong alerts. Next week's prediction by this method would be 76, though that may go down if we find any invisible alerts.

 

[Ian Doncaster]

Next week's prediction by this method would be 76, though that may go down if we find any invisible alerts.

And up if invasion systems in their last week (or, less likely, uninhabited alerts) are successfully defended? Or am I misunderstanding when "new control systems" gets counted for this theory?

Interesting correlation. Another notable thing about the points so far is that - ignoring the X-axis entirely - they're showing an oscillation which so far is decreasing in magnitude. 76 would be again reversing direction but would also go beyond the 5 Jan measure. Whether it does or doesn't do that might be helpful for figuring out what the actual cause is.

 

[yttrbio]

Technically, yes. Defended alerts/invasions drops new control systems so would increase alerts, if this is a direct relationship. This would make some sense as a design for maintaining a somewhat consistent expansion rate regardless of player action. Still, this is only a correlation, not necessarily a causal relationship.

 

[Ian Doncaster]

Okay, Invasion duration seems to just be complicated. Muruidooges is 5 weeks left, 4 ports. Vogulu is 4 weeks left, 5 ports. Both "fresh" invasions.

 

[Ned Flandalorian]

Tracked activity required to move a previously populated system by 1 pip:
System: Humarala (controlled, depopulated)
Maelstrom: Raijin (Pegasi Sector IH-U b3-3)
Distance from Maelstrom: 17.83 LYs
Time controlled: ??? (possibly 2 weeks)
Other thargoid controlled systems within 10LYs: Pegasi Sector NN-S b4-10 (3.02LYs), Pegasi Sector NN-S b4-6 (5.19LYs), HIP 112219 (6.56LYs), Pegasi Sector NN-S b4-9 (8.24LYs), HIP 113785 (9.74LYs)

204 Scouts
15 Cyclops
18 Basilisks
16 Medusas
Total combat bonds: 1,208,320,000

 

[Ralph Vargr]

Tracked activity required to move a previously populated system by 1 pip:
System: Humarala (controlled, depopulated)
Maelstrom: Raijin (Pegasi Sector IH-U b3-3)
Distance from Maelstrom: 17.83 LYs
Time controlled: ??? (possibly 2 weeks)
Other thargoid controlled systems within 10LYs: Pegasi Sector NN-S b4-10 (3.02LYs), Pegasi Sector NN-S b4-6 (5.19LYs), HIP 112219 (6.56LYs), Pegasi Sector NN-S b4-9 (8.24LYs), HIP 113785 (9.74LYs)

204 Scouts
15 Cyclops
18 Basilisks
16 Medusas
Total combat bonds: 1,208,320,000

Wow. That's a lot of kills and time.

 

[yttrbio]

On the more positive side of the equation, our test at retaking a controlled uninhabited system was successful. Hyades Sector EQ-O b6-3 (31.07 ly away from Taranis, 1 nearby control system) will return to uncontrolled, assuming the alert cooldown is 4 weeks. It will be realerted Jan. 26 if it follows the pattern. Thanks to all the commanders who participated, and Canonn Lab 69, Operation: Reform Aegis, and The Anti-Xeno Initiative for adding it to their goals for the week. Some numbers:

The initial effort was just me, and at 511 scouts, it had 4 pips. The estimate, then, is about 128 scouts per pip.

At a checkpoint 34 pips in, Inara had 3036 scout kills and 33 interceptor kills recorded. That has the potential to be an undercount, but there's no good reason for anyone to go to the system unless pointed there by this project, and I assume the folks connected this way are more likely to use Inara, so it's likely not as much of an undercount as usual. Using the 128 number from above, 3036 scout kills accounts for 23.7 pips, so if the 33 interceptors were the entirety of the remainder, they count for something like 40 scouts each. Obviously, there's likely some variation for different types, so this is only an average, and any unrecorded data brings this number down, so I'm seeing 40 as a maximum number for something like a cyclops.

Overall, this is something like the equivalent of 6400 scouts to kick the Thargoids out of an unpopulated system positioned this way. Not a trivial amount, but hardly insurmountable for a small group of players.

 

[Ian Doncaster]

At a 40 scouts per interceptor exchange rate, that would imply Humarala needed about 17 times as much effort as EQ-O for each pip.

It's closer in, but if there was a general 17x difference over that sort of distance range it would have been very noticeable elsewhere. That does suggest that inhabited and uninhabited are likely to be on separate scales.

Next week, Paitra at ~25 LY from Cocijo will be the furthest out control system, which might make it slightly easier.

 

[Ned Flandalorian]

I would be very surprised if all interceptor types are treated equally. Testing multiple pips in one uninhabited controlled system could probably reveal that info.

 

[Ned Flandalorian]

Actually it seems I forgot to post here the experiment I did last week (moving an uninhabited controlled system by one pip, similar to Hyades Sector EQ-O b6-3.
System: Pegasi Sector MN-S b4-0 (controlled, unpopulated)
Maelstrom: Raijin (Pegasi Sector IH-U b3-3)
Distance from Maelstrom: 20.76 LYs
59 Scouts
1 Cyclops
6 Basilisk
2 Medusa
Total Bonds: 236,720,000

I wish now that I'd gone for another pip with a different batch of interceptors, since it would only take a few collections to be able to solve for X/Y/Z weights. But it was a last minute job late in the cycle, I didn't expect the bar to move at all / as easily as it did.

 

[yttrbio]

This week on observations:

Leigong Invisible Map Report:

Newly Controlled:
Ceti Sector ZU-Y b3

Invisible Alerts:
Arietis Sector KM-W d1-81
Ceti Sector BQ-Y b4
HIP 10168
HIP 10972

The prediction method I mentioned above did not hold up this week. After adjusting for invisible alerts and control system changes, the prediction was 80, and the actual number of alerts was 94, so go ahead and chuck that in the bin.

29 e Orionis did not get an alert this week after its 4-week recovery period. This was true for a number of such systems, but the others were never alerted in the first place (they were invaded automatically because they were 15-20 ly out from the maelstrom). This suggests that not every once-alerted system is going to be re-alerted at the first opportunity, indicating one of 2 things:

1. There is a random element to alert choice, or

2. The method used to prioritize alert choice put 29 e Orionis further down the list due to new alert options becoming available (which seems plausible, as it was alerted week 1, before any other maelstroms were putting out alerts)

Finally, Huile is alerted again, only 3 weeks after its last alert, so the 4-week cooldown on alerts does not hold up

All newly-alerted systems continue to be within 10.02ly of a control system the previous week. I'm very comfortable calling 10 ly the range at this point, with some allowance for rounding issues pushing it a tiny bit past.

 

[Ian Doncaster]

After adjusting for invisible alerts and control system changes, the prediction was 80, and the actual number of alerts was 94, so go ahead and chuck that in the bin.

That's concerningly high, too - if it rises again next week then that would be very bad and imply some level of acceleration to their generation.

29 e Orionis did not get an alert this week after its 4-week recovery period. This was true for a number of such systems, but the others were never alerted in the first place (they were invaded automatically because they were 15-20 ly out from the maelstrom). This suggests that not every once-alerted system is going to be re-alerted at the first opportunity, indicating one of 2 things:

1. There is a random element to alert choice, or

2. The method used to prioritize alert choice put 29 e Orionis further down the list due to new alert options becoming available (which seems plausible, as it was alerted week 1, before any other maelstroms were putting out alerts)

Finally, Huile is alerted again, only 3 weeks after its last alert, so the 4-week cooldown on alerts does not hold up

The simplest theory based on this is probably
1) Process all recoveries to set systems back to safe
2) Determine the budget for new alerts
3) While budget remains, pick a random control system, and then pick a random safe system within 10 LY to become an Alert

Systems close to the Maelstrom - and therefore within a local region of lots of control systems and few safe systems - are very likely to get re-Alerted fairly quickly, but no guarantees, while outliers would probably be safe for some time once secured as there are thousands of alternatives.

The tricky thing would be determining that theory to be false if it was: I suppose if an outlier system was repeatedly re-attacked it might be evidence against it. It would also tend to lead to entirely non-directional expansion from the Maelstrom if so - which did vaguely seem to be true earlier; I wonder if that still holds when considering systems >20 LY from each Maelstrom only?

 

[yttrbio]

The simplest theory based on this is probably
1) Process all recoveries to set systems back to safe
2) Determine the budget for new alerts
3) While budget remains, pick a random control system, and then pick a random safe system within 10 LY to become an Alert

I think there's more to it than that. Recall that 6 of the 8 maelstroms had 0 alerts in week 2, and it wasn't for lack of targets. Instead they had to wait until a new set of control systems became available. Obviously the natural trend would be to have fewer alerts on week 2, but the fact that most of them were 0 stretches the "fully random" hypothesis. I might be able to cook up a test for it, but that sort of distribution is not consistent with randomly-picked control targets without some other factors at play.

 

[Ian Doncaster]

True - though the distributions seem to have been becoming more symmetrical over time. There certainly looked to be an alternating high-low pattern of Alerts to start with, but other than maybe Thor and Raijin that seems to have stopped now, and Taranis never seemed to have it at all.

Oya (heavily opposed) and Cocijo (never successfully opposed so far) which arrived at the same time have much the same alternating pattern of Alerts up to last week, and then both break alternation in about the same way this week. Admittedly on a per-Maelstrom basis the difference between heavy opposition and light opposition is maybe 2 systems a week, so that maybe wouldn't be expected to make much difference. ... on the other side, Indra seemed to have the pattern weaker and break out of it earlier than its co-arriver Leigong ... while Hadad has behaved quite differently to Raijin and Thor in the last couple of weeks.

I agree that there's likely something going on here besides entirely random assignment - some sort of "pick a Maelstrom" step, perhaps? - but I really can't come up with a convincing idea for how it might work that fits the data

 

[Ned Flandalorian]

Failed attempt at measuring progress in an uninhabited alert system (last week).
System: HR 8536 (alert, unpopulated)
Maelstrom: Raijin (Pegasi Sector IH-U b3-3)
Distance from Maelstrom: 27.73 LYs
68 Scouts
21 Cyclops
11 Basilisk
5 Medusa
Total Bonds: 637,440,000

Result: NO PROGRESS OBSERVED

This was the most frustrating experience I have had during my time playing ED, and not because of the lack of a result. Finding the thargoids is far more difficult than actually fighting them and because the system is uninhabited, it is the only defensive action available. I can say without hesitation that this mechanic is surely broken because trying to do more than 50 times the amount of kills that I put in within the space of one week, even with many more players helping, simply isn't feasible due to the extremely low numbers of targets.

 

[Kryxix]

Thargoid War Update: January 3309​

Vox Galactica reporter Jade Sanderlyn has published a summary of the current situation in the conflict against the Thargoids.

“Many of us who witnessed those terrifying events in HIP 22460 felt that it was both an ending and a beginning – that history had turned a page. Five months later, the Second Thargoid War has escalated to a point which may yet close the book altogether.”

“The arrival of the eight Maelstroms resulted in 44 populated systems being completely overwhelmed by the Thargoids. Dozens of neighbouring regions were also attacked by their fleets, leaving starports and settlements ablaze. Billions were killed and billions more fled from their homes as panicked refugees, creating the greatest humanitarian crisis of all time.”

“Despite the lack of coordination previously provided by Aegis and Salvation, anti-xeno efforts have been herculean. Allied, Federal and Imperial warships rapidly responded to form a vanguard against the invaders, and thousands of independent pilots have valiantly fought the Thargoid swarms across multiple battlefields.”

“Military experts believe that neutralising the Maelstroms will stem the endless waves of Thargoid vessels. Unfortunately, their caustic clouds and repulsing energy waves prevent our ships from exploring whatever resides within. They also generate the same disruptive field found in HIP 22460, which prevents any Guardian-based weapons from functioning within the system.”

“But there is hope. A small but steady flow of unclassified relics, those strange artefacts recovered from Thargoid surface sites after the Proteus Wave fired, have made their way to Professor Ishmael Palin and Ram Tah. A leading theory is that they share esoteric properties with several of the Thargoids’ recent technological advances, and may hold the key to developing countermeasures. Many other scientists are focusing on xenological research, striving to locate a weakness or develop an advantage.”

“Until such a breakthrough occurs, we are reliant upon our spacefaring warriors to contain the most destructive adversary the human race has ever faced. This is Jade Sanderlyn for Vox Galactica. Stay safe out there.”

 

[Ned Flandalorian]

Since alerts seem to be tied to specific control systems (within ~10lys) do you think it might be worth testing if eliminating that "anchor" has any effect?
Something like this:

• Look for an uninhabited alert (system A) with one control system (also uninhabited) within 10LYs (system B)
• Let A rollover to controlled
• Test how much effort is needed for one pip of progress in A and at the same time organise a group to completely liberate B from thargoid control
• After the weekly tick, make sure A no longer has any control systems within 10LYs and test again how much effort is needed for one pip of progress

Purpose would be to determine if distance to the maelstrom is the sole factor determining difficulty, or if nearby control systems also have influence.