Analysing the Thargoid Simulation

[Ian Doncaster]

Another way to figure out if nearby control systems might have influence is to see what happens to difficulty at the same distance over time - on average, it should increase as the Thargoids fill in more control systems around it.

That doesn't seem to be the case - broken down by week, and ignoring whatever was happening in w/b 29 Dec, there's no clear trend in difficulty between weeks (and to the extent the differences between weekly trendlines might be meaningful, it's looking marginally easier in later weeks for the same distance)

 

[yttrbio]

Judging from the first galnet about difficulty, getting easier over time is consistent with what they said. I just wrapped up another experiment that strongly suggests against nearby control systems being relevant. The key results are:

1. A cyclops kill is worth 11-14.45 scout kills
2. If the difficulty is solely based on the distance to the maelstrom, the difficulty is not linear with either distance to the maelstrom or its reciprocal. It's even steeper.

There's good evidence that nearby control systems does NOT affect difficulty, as 2 systems were quite close but with very different nearby control system profiles and had quite similar difficulty levels.

Details here

 

[Imo]

We’ve been told that the Thargoid Simulation is fully autonomous and will make its own decisions on how to fight – and with hundreds of systems under attack, how else could it be? This thread is intended as a summary of the research into how it works for the benefit of military planners fighting against it, attempting to analyse it in the style of the Political Simulation.

At the moment there are more unknowns than anything else, of course. I’ll try to keep these first posts updated as we find out more.

State of the War​

We’re losing, though the bubble is very large so it will probably take several years to lose all of it.

Maelstroms are named after storm gods from different religions. Currently all are positioned in fringe bubble systems on the “south” side of the bubble, mostly below the Sol plane. Their numbering in the map does not reflect their original order of arrival (or indeed anything else obvious)

1. Leigong: HIP 8887
2. Indra: HIP 20567
3. Hadad: HIP 30377
4. Thor: Col 285 Sector IG-O c6-5
5. Cocijo: Col 285 Sector BA-P c6-18
6. Raijin: Pegasi Sector IH-U b3-3
7. Oya: Cephei Sector BV-Y b4
8. Taranis: Hyades Sector FB-N b7-6

More data on the evolution of Thargoid control over time will be added here later.

	Alert	Invasion	Controlled	Maelstrom	Recovery
Last Week	51	38	329	8	18
This Week	75	23	386	8	23

How long will it take for the Thargoids to consume the entire bubble? At the current rate of approximately 8 Alerts per Maelstrom per week, including uninhabited systems, it would likely take about 20 years even in the absence of opposition. At their current balance of inhabited versus uninhabited systems they might be possible to hold off indefinitely from populated space with a defence strength of 10-12 systems a week. Of course, a lot can change in 20 years!

In general it seems that systems further from the Maelstrom are easier to defend than closer systems, though both sorts still require a major effort. So far the distance range is still fairly narrow and data is limited so quantifying this is not practical yet. There does not appear to be a general overheads mechanism though there may be one based on the number of Maelstroms.

(Unknown: how are systems outside the main connected bubble attacked? It could take decades to reach even the nearer ones via spherical expansion)

(Unknown: what about nebulae systems which are already under some sort of old-style Thargoid threat from the NHSS spheres? At the moment the Thargoids there are maintaining their less aggressive behaviour)

The Other Sort of Bug​

• The main war progress panel, the individual Maelstrom summary panels, and the galaxy map all show slightly different numbers of systems in the various states. This is likely due to invisible states near Leigong.
• Despite what the documentation says, it appears impossible to speed up the Recovery progress slider by doing anything in the system.
• Systems being entirely depopulated takes away their CC generation in Powerplay but doesn't take it away on the per-system Powerplay displays
• The numbering of the Maelstroms changes as more arrive (Taranis was 1, then 3 and 5, is now 8). Hopefully this will remain stable now all 8 are here.
• As before, Research Limpets are very unreliable against Thargoids. Hopefully the Orthrus samples aren't needed for anything.
• Continuing reports of network instability, stuck progress counters, etc. in AX CZs and similar
• Inconsistencies in mission board display

The war against the Thargoids is currently not going well, with more systems being taken over by the Thargoids each week. The Thargoid Simulation is fully autonomous and it is difficult to predict its actions. The Maelstroms are currently positioned on the south side of the bubble, mostly below the Sol plane. At the current rate, it is estimated that it would take about 20 years for the Thargoids to consume the entire bubble, even in the absence of opposition. However, it is also noted that a lot can change in 20 years. Systems further from the Maelstrom are easier to defend than closer systems, but both require a major effort. It is also noted that there are still many unknowns about the Thargoids and their behavior, and there are reports of network instability and other technical issues in the game.

 

[Ian Doncaster]

Judging from the first galnet about difficulty, getting easier over time is consistent with what they said.

True, though since the 8th Maelstrom arrived I would say that distance rather than an overall drop in difficulty is what accounts for it.

I don't think any of the four clustered lines on that graph are meaningfully different; the blue one being a bit higher would make sense if the difficulty relationship isn't linear (and it would make sense for it not to be purely linear as that would put a definite cap on their expansion range)

 

[Ned Flandalorian]

Last week I tested killing a variety of interceptors to gain progress pips in HR 8536 (uninhabited thargoid controlled system 27.73 LYs from Raijin). These were the counts:

S = scouts, C = cyclops, B = basilisk, M = medusa
Pip 1: 35S + 1C + 6B
Pip 2: 2M
Pip 3: 28S + 4C + 3B + 1M
Pip 4: 3S + 1B + 2M
Pip 5: 41S + 1C + 1B + 1M
Pip 6: 124S
Pip 7: 39S + 3C + 1B + 1M
Pip 8: 2B + 2M
Pip 9: 5S + 1C + 1B + 2M
Pip 10: 3M
Pip 11 (INCOMPLETE, last kill was over an hour before downtime and was not enough to move the bar): 96S + 3C

Total counts for first 10 pips: 275 Scouts, 10 Cyclops, 15 Basilisks, 14 Medusas

No matter how much I fiddle with the multipliers, I can't get a "reasonable" result when using 11 or above for cyclops - the discrepancy amounts for each pip either end up too high or go negative. What does seem to work is using 8x/24x/40x for each interceptor variant (the same as their bond amounts) which works out to an average of 127.5 scouts killed for each of the first 10 pips (very close to the "raw" amount of scouts killed for pip 6).

 

[yttrbio]

With those numbers, if you were waiting after each set that got the pips, using the following numbers:
175 scouts per pip
12 scouts per cyclops
36 scouts per basilisk
60 scouts per medusa

It works out. The 1, 3, 4, and 5 pips overshoot, but that depends on how quickly you were killing things.

127.5 scouts per pip seems too low for something 27.73 out. I was getting 120-130 for 33.5 ly out, 130-140 for 29.77 ly out, and 230-292 for 21.47 ly out.

Now that we have ballpark figures, at least, it should be a lot easier to test specific values in the future (given the delay before the progress bar updates). I'll take another run at testing some of this next week, but for now, I don't think it makes sense to do any more testing until the next update drops, since it's supposed to change how bars progress.

 

[Ian Doncaster]

A couple of interesting points with the Alerts this week.
- the general oscillation in per-week numbers, especially with Thor/Raijin, seems to be continuing and has led to a similar number to two weeks ago.
- the proportion of Alerts affecting inhabited systems appears to be rising - a full third of this week's are.

 

[Rubbernuke]

A couple of interesting points with the Alerts this week.
- the general oscillation in per-week numbers, especially with Thor/Raijin, seems to be continuing and has led to a similar number to two weeks ago.
- the proportion of Alerts affecting inhabited systems appears to be rising - a full third of this week's are.

Check out the livestream today, they had some general information you'd might like re the simulation.

 

[Ian Doncaster]

Check out the livestream today, they had some general information you'd might like re the simulation.

I've seen a couple of summaries - the extended video format doesn't really work for me so I'm hoping they're comprehensive for the important bits

 

[Ned Flandalorian]

With those numbers, if you were waiting after each set that got the pips, using the following numbers:
175 scouts per pip
12 scouts per cyclops
36 scouts per basilisk
60 scouts per medusa

Those numbers do look appealing but for pip 9, I left the game closed down for hours before finally killing the five scouts... and it was the final three (with a 40 minute gap beforehand) that pushed it over the line, which doesn't match up with an overshoot of ~20. Another party might have come into the system and messed up the counting (which is a good argument for choosing systems on the fringes - kills from hyperdictions while passing through the system definitely contribute!).

 

[Ian Doncaster]

A couple more (negative) checks on what might affect Alert numbers - measuring the successful or strongly attempted (~>20%) systems per Maelstrom since all eight arrived, compared with the number of Alerts generated in the same time period. No correlation either way, so pushing back doesn't seem to either obviously weaken them or make them send more Alerts there next time - at least, pushing back on (mostly) Invasions doesn't.

 

[Ned Flandalorian]

I could not understand the suggestion that thargoid expansion was somehow related to opposition effort when Cocijo barely had any defensive action at all (up until the week beginning 19th) and is now the maelstrom with the most systems. Almost as if someone was attempting to push a pro-xeno agenda rather than looking at facts...

 

[Ian Doncaster]

One question I'm interested in is what the equilibrium point of the Thargoid expansion is. At the moment, human efforts win about 8 systems a week, and Thargoid expansion adds about 80 new ones. So it should reach some sort of equilibrium at the point where the decreasing difficulty with distance allows humans to win 80 systems a week as well - or perhaps only 40 if the approximately spherical expansion continues and we don't worry too much about the back hemispheres.

There's not really enough evidence at the moment to tell if the difficulty decrease is exponential (i.e. proportional to e^-nX where n is a constant and X is the distance) or power (i.e. proportional to X^-n) and until we're regularly fighting at 50 LY I wouldn't expect to be able to tell from Invasion data; Uninhabited Control data should be possible to analyse much sooner, though. I'd suspect "Power" is most likely - partly because it doesn't decay quite as quickly so gives the Maelstroms enough practical range to reach most of the bubble and still put up some sort of fight.

Extrapolation	Hold hemisphere	Equilibrium Date (hemisphere)	Hold full sphere	Equilibrium Date (sphere)
Linear (extremely unlikely!)	35 LY	6 months	40 LY	8 months
Exponential (n=-0.07)	50 LY	2 years	60 LY	4 years
Power (n=-1.66)	70 LY	6 years	110 LY	>10 years

These are all very rough estimates because - even allowing from them being long-range extrapolation from a very narrow distance window, so the uncertainty on 'n' is huge - once they get consistently past 50 LY they'll start overlapping, which will increase their ability to put Alerts in other directions. Additionally, once they get to 50 LY in size they'll start hitting an increasing number of systems we really don't want to lose, which will take away capacity from handling their more generic expansions into defending Meene et al over and over again. Conversely, as the difficulty decreases, smaller groups will be able to win fights on their own, which is likely to increase the effective capacity. I'm not attempting to model any of that!

At any rate, it looks like the answer remains "far enough away that something else will make the question obsolete first" ... but it's interesting that in their current configuration they aren't powerful enough to destroy the bubble in the face of opposition and could be held to an extremely costly stalemate after capturing about half of it.

 

[Ralph Vargr]

One question I'm interested in is what the equilibrium point of the Thargoid expansion is. At the moment, human efforts win about 8 systems a week, and Thargoid expansion adds about 80 new ones. So it should reach some sort of equilibrium at the point where the decreasing difficulty with distance allows humans to win 80 systems a week as well - or perhaps only 40 if the approximately spherical expansion continues and we don't worry too much about the back hemispheres.

There's not really enough evidence at the moment to tell if the difficulty decrease is exponential (i.e. proportional to e^-nX where n is a constant and X is the distance) or power (i.e. proportional to X^-n) and until we're regularly fighting at 50 LY I wouldn't expect to be able to tell from Invasion data; Uninhabited Control data should be possible to analyse much sooner, though. I'd suspect "Power" is most likely - partly because it doesn't decay quite as quickly so gives the Maelstroms enough practical range to reach most of the bubble and still put up some sort of fight.

Extrapolation	Hold hemisphere	Equilibrium Date (hemisphere)	Hold full sphere	Equilibrium Date (sphere)
Linear (extremely unlikely!)	35 LY	6 months	40 LY	8 months
Exponential (n=-0.07)	50 LY	2 years	60 LY	4 years
Power (n=-1.66)	70 LY	6 years	110 LY	>10 years

These are all very rough estimates because - even allowing from them being long-range extrapolation from a very narrow distance window, so the uncertainty on 'n' is huge - once they get consistently past 50 LY they'll start overlapping, which will increase their ability to put Alerts in other directions. Additionally, once they get to 50 LY in size they'll start hitting an increasing number of systems we really don't want to lose, which will take away capacity from handling their more generic expansions into defending Meene et al over and over again. Conversely, as the difficulty decreases, smaller groups will be able to win fights on their own, which is likely to increase the effective capacity. I'm not attempting to model any of that!

At any rate, it looks like the answer remains "far enough away that something else will make the question obsolete first" ... but it's interesting that in their current configuration they aren't powerful enough to destroy the bubble in the face of opposition and could be held to an extremely costly stalemate after capturing about half of it.

Very neat. Thank you.

 

[Banana's_Are_Good_For_You]

Anyone looking at the recovery elements since the changes?

 

[Ian Doncaster]

Anyone looking at the recovery elements since the changes?

The systems defended from last week all seem to be on 3-week timers rather than 4-week, so that bit seems to work.

 

[Banana's_Are_Good_For_You]

Hmm...
Well, that's a thing that looks like it's making your job harder, as it stands.
Are they that evil and gone full dystopia?

I just wanted to know if you were paying attention to it.
Have fun with it!
.

 

[Ian Doncaster]

Quick graph of how much effort a group capable of X%/week in its chosen system will waste by just fighting in that system until the Thargoid reinforcements are ground down compared with if it teamed up with 2-3 other groups of the same size to take on one system a week.

It's a nice move for making near-misses not an outright disaster, but failure to coordinate properly remains costly (sometimes incredibly costly), which seems a good balance for strategy.

I'll update the flowchart later

 

[Aleks Zuno]

Quick graph of how much effort a group capable of X%/week in its chosen system will waste by just fighting in that system until the Thargoid reinforcements are ground down compared with if it teamed up with 2-3 other groups of the same size to take on one system a week.

It's a nice move for making near-misses not an outright disaster, but failure to coordinate properly remains costly (sometimes incredibly costly), which seems a good balance for strategy.

I'll update the flowchart later

Great idea there and a correct conclusion, though actually I think your visualisation of it understates the wasted effort! In part, this is because it seems to define the amount wasted to be the amount drained divided by the total effort spent, as a proportion rather than as a total.

Moreover, I think a clearer view would be one which just considers the total effort spent to take one system, which then gives us two sets:

• Exclusive effort, which includes some wasted due to over-achieving in its last cycle after the system is already complete.
• Adaptive effort, which starts similarly but then stops being spent needlessly once 100% is reached.

They look like this:

Admittedly that curvature looks similar at a glance, but notice how that is now on a logarithmic scale; the reality on a linear scale is much, much more dire! My comments:

• Obviously 100 per cycle will cost a total of 100 for one system, the best one can do. Technically the Exclusive effort continues rising beyond 100, and now that I consider it, that does actually happen a lot due to Commanders not noticing when a system is complete.
• 99 is a good example of Exclusive versus Adaptive; 99 exclusively will be 198 for one system, whereas 67–99 adaptively will all be 133, the surplus moving elsewhere.
• To illustrate how dire 34 effort is, that total expenditure is 2278, and missing from that view is the fact that it takes 67 cycles to do it.

Obviously the amount wasted is just the above minus 100, though I think it well to leave it as a total. If it would help to have a visualisation which avoids a logarithmic scale, one could consider the efficiency to be the resulting 100 divided by the amount spent, which looks like this:

 

[Ian Doncaster]

Yes. Looking at the number of weeks needed also makes the exponential nature clear.

Rapidly increases once you fall below 50% capacity, and starts being noticeably inefficient below 67%.