Analysing the Thargoid Simulation
- Thread starter Ian Doncaster
- Start date
Excellent - and many thanks for going through all of that!
I did a check too and the rule of "Every alert system must have an unique control system within 10Ly" holds up for almost all off the dcoh.watch dataset (assuming my code works correctly).
The exceptions are:
All of the exceptions just had no control systems in range instead of them being used up for other systems.
To match up the alert/control systems in code I just sorted the alerts by lowest number of control systems within 10ly and sorted those control systems by distance picking the lowest distance unused control system. This might mean that stuff is also weighed by distance which might be helpful for predicting stuff.
The Leigong stuff is still weird unless there's missing data there, maybe it works on a similar basis to extended range bgs expansions?
The exceptions are:
- Some stuff, in the first(?) cycle in Leigong
- A system in Indra during the first(?) cycle (could also be a rounding error for 64 Tauri at 10.012785 ly?)
- A system in Leigong during a later cycle
2022-12-08
No attackers for HIP 20890
No attackers for Hyades Sector LX-U d2-95
No attackers for Arietis Sector NX-U c2-19
No attackers for Hyades Sector JN-K b8-1
No attackers for Arietis Sector XJ-R b4-2
2023-01-05
No attackers for Arietis Sector NX-U c2-12
No attackers for HIP 20890
No attackers for Hyades Sector LX-U d2-95
No attackers for Arietis Sector NX-U c2-19
No attackers for Hyades Sector JN-K b8-1
No attackers for Arietis Sector XJ-R b4-2
2023-01-05
No attackers for Arietis Sector NX-U c2-12
All of the exceptions just had no control systems in range instead of them being used up for other systems.
To match up the alert/control systems in code I just sorted the alerts by lowest number of control systems within 10ly and sorted those control systems by distance picking the lowest distance unused control system. This might mean that stuff is also weighed by distance which might be helpful for predicting stuff.
The Leigong stuff is still weird unless there's missing data there, maybe it works on a similar basis to extended range bgs expansions?
There are a few above 10 ly, but none above 10.02 when I was still checking every week. I wonder if the DCoH data is still missing the early invisible Leigong systems, and that's why you can't find attackers.
Potential Attackers for the systems listed:
HIP 20890: 64 Tauri
Hyades Sector LX-U d2-95: Hyades Sector KN-K b8-2, Hyades Sector KN-K b8-0, Arietis Sector NX-U c2-20
Arietis Sector NX-U c2-19: Arietis Sector NX-U c2-20
Hyades Sector JN-K b8-1: Hyades Sector KN-K b8-2
Arietis Sector XJ-R b4-2: Arietis Sector XJ-R b4-0
Arietis Sector NX-U c2-12: Arietis Sector KM-W d1-93
Potential Attackers for the systems listed:
HIP 20890: 64 Tauri
Hyades Sector LX-U d2-95: Hyades Sector KN-K b8-2, Hyades Sector KN-K b8-0, Arietis Sector NX-U c2-20
Arietis Sector NX-U c2-19: Arietis Sector NX-U c2-20
Hyades Sector JN-K b8-1: Hyades Sector KN-K b8-2
Arietis Sector XJ-R b4-2: Arietis Sector XJ-R b4-0
Arietis Sector NX-U c2-12: Arietis Sector KM-W d1-93
A couple of questions I have on this:
Is it possible to tell from the current data if a Control can place an Alert in its first week?
Would it be possible, since Thor is so constrained, to predict what its expansions next week will be, assuming 10LY and no reuse of a Control in adjacent weeks?
Is it possible to tell from the current data if a Control can place an Alert in its first week?
Would it be possible, since Thor is so constrained, to predict what its expansions next week will be, assuming 10LY and no reuse of a Control in adjacent weeks?
I looked at my code again this morning and if I don't allow ControlledNew to invade I get 2 more unexplained alerts (still a hit/miss of 935/8).
Trying to explain ControlledNew systems with the same rule gives a hit/miss of 1873/29 (including ~10 double counted systems).
It's a decent enough hit rate that I'm fairly confident of the rule and assume it's probably missing data due to invisible control systems or other weirdness.
With a few extra lines of code I can count how many times future invaded systems were skipped over while they were in range (with possibly errors in retaken systems). No obvious trend there pops out to me in position or FSS scan values of the systems.
If you mean a new control system has an additional cooldown of 1 tick before it can invade anything then that would invalidate a lot of alerts (but the code is getting messier as I add stuff).
Trying to explain ControlledNew systems with the same rule gives a hit/miss of 1873/29 (including ~10 double counted systems).
Code:
2023-03-02
2023-02-23
2023-02-16
2023-02-09
2023-02-02
2023-01-26
2023-01-19
No attackers for Vaipacnali (Thor/ControlledNew)
2023-01-12
No attackers for Garongxians (Leigong/AlertNew)
No attackers for Arietis Sector NX-U c2-12 (Leigong/ControlledNew)
2023-01-05
No attackers for Arietis Sector NX-U c2-12 (Leigong/AlertNew)
No attackers for Akbakara (Oya/ControlledNew)
No attackers for HIP 24329 (Oya/ControlledNew)
No attackers for HIP 8825 (Oya/ControlledNew)
No attackers for Tougeir (Oya/ControlledNew)
No attackers for Col 285 Sector SH-B b14-7 (Hadad/AlertNew)
2022-12-29
No attackers for Col 285 Sector WY-F b12-3 (Cocijo/ControlledNew)
2022-12-22
2022-12-15
No attackers for HIP 7819 (Leigong/ControlledNew)
No attackers for Arietis Sector AQ-P b5-0 (Leigong/ControlledNew)
No attackers for HIP 20890 (Indra/InvasionNew)
No attackers for Hyades Sector LX-U d2-95 (Leigong/ControlledNew)
No attackers for Arietis Sector NX-U c2-19 (Leigong/ControlledNew)
No attackers for Hyades Sector JN-K b8-1 (Leigong/ControlledNew)
No attackers for Arietis Sector XJ-R b4-2 (Leigong/ControlledNew)
2022-12-08
No attackers for 5 Mu Leporis (Taranis/ControlledNew)
No attackers for Matshiru (Taranis/ControlledNew)
No attackers for HIP 20890 (Indra/AlertNew)
No attackers for Hyades Sector LX-U d2-95 (Leigong/AlertNew)
No attackers for Arietis Sector NX-U c2-19 (Leigong/AlertNew)
No attackers for Hyades Sector JN-K b8-1 (Leigong/AlertNew)
No attackers for HIP 24329 (Oya/InvasionNew)
No attackers for Kanus (Oya/ControlledNew)
No attackers for Mapon (Cocijo/InvasionNew)
No attackers for Muchihiks (Oya/InvasionNew)
No attackers for Col 285 Sector WY-F b12-3 (Cocijo/InvasionNew)
No attackers for Arietis Sector XJ-R b4-2 (Leigong/AlertNew)
2022-12-01
2022-11-24
hit/miss 1873/29It's a decent enough hit rate that I'm fairly confident of the rule and assume it's probably missing data due to invisible control systems or other weirdness.
With a few extra lines of code I can count how many times future invaded systems were skipped over while they were in range (with possibly errors in retaken systems). No obvious trend there pops out to me in position or FSS scan values of the systems.
If you mean what's a ControlledNew system in the dcoh.watch data then that's only needed to explain 2 invasions so probably no.
If you mean a new control system has an additional cooldown of 1 tick before it can invade anything then that would invalidate a lot of alerts (but the code is getting messier as I add stuff).
I'll try but I think "non controlled system 10ly from any control system in Thor" will still be a long list.
If it turns out to be longer than 20 systems - after excluding the ones used this week - then that can be a prediction that Thor will use its full budget.
I got 25 total systems, I guess it's more sparse than I thought. I only ignored current invasion/alert systems and assumed all current Control/ControlNew systems would be able to invade next tick.
Code:
Col 285 Sector GF-G b11-5 [2.19, 6.55, 8.12, 9.69]
Col 285 Sector IA-G b11-0 [2.97, 3.78]
Col 285 Sector SH-C b13-1 [3.19, 5.87, 5.91]
Hyades Sector NM-M b7-1 [3.29, 9.22]
Col 285 Sector HL-O c6-22 [6.01, 8.85]
Hyades Sector SI-T c3-4 [6.35]
Col 285 Sector PW-D b12-0 [6.51, 8.83]
Col 285 Sector NQ-F b11-3 [7.0]
Col 285 Sector HF-G b11-4 [7.19, 7.64]
Col 285 Sector QW-D b12-2 [7.26]
HIP 20807 [8.0]
Col 285 Sector HF-G b11-3 [8.32, 9.08]
Hyades Sector JG-O b6-3 [8.42]
Aruntei [9.03]
Col 285 Sector QR-C b13-2 [9.05, 9.74]
Col 285 Sector KV-F b11-2 [9.05]
Unktety [9.16]
Col 285 Sector MQ-F b11-3 [9.32]
Col 285 Sector KV-F b11-3 [9.38]
Col 285 Sector SM-C b13-4 [9.4]
HIP 21548 [9.55]
Col 285 Sector FU-H b10-2 [9.76]
Hyades Sector NH-M b7-3 [9.8, 9.85]
Col 285 Sector HF-G b11-5 [9.83]
Col 285 Sector HP-H b10-3 [9.87]Even without taking cooldowns into account or assuming some control systems are "tied up" somehow in invasions or on cooldown there's only 15 attacking systems to cover these (I think only 2 of these spawned alerts into uncontrolled systems this tick):
Col 285 Sector IA-G b11-1
Col 285 Sector JG-O c6-18
Col 285 Sector SH-C b13-0
Col 285 Sector IA-G b11-2
Col 285 Sector CF-Q c5-5
Col 285 Sector KB-O c6-11
Hyades Sector MM-M b7-1
Col 285 Sector JG-O c6-15
Col 285 Sector EA-Q c5-16
HIP 19198
Col 285 Sector HF-G b11-2
Col 285 Sector RW-D b12-1
Col 285 Sector KB-O c6-10
Col 285 Sector EA-Q c5-12
Varlawoth
Col 285 Sector JG-O c6-18
Col 285 Sector SH-C b13-0
Col 285 Sector IA-G b11-2
Col 285 Sector CF-Q c5-5
Col 285 Sector KB-O c6-11
Hyades Sector MM-M b7-1
Col 285 Sector JG-O c6-15
Col 285 Sector EA-Q c5-16
HIP 19198
Col 285 Sector HF-G b11-2
Col 285 Sector RW-D b12-1
Col 285 Sector KB-O c6-10
Col 285 Sector EA-Q c5-12
Varlawoth
But wait, there's less! Some of these might be doubled up only being able to attack the same system(s). So I think it's only 14 unique attack opportunities unless even more can be disqualified somehow.
It still feels impossible to predict things because any error would cascade and ruin the rest of the prediction, so it's all or nothing. Which on the flipside could make theories easier to confirm.
I wonder if there's any systems clusters the bubble that are completely safe because they're cut off by a 10Ly gap from everything else.
Also I used actual distance instead of cube distance to calculate my stuff, I hope the thargoids do too.
Also I used actual distance instead of cube distance to calculate my stuff, I hope the thargoids do too.
With at least two inhabited, that might (just) allow a full use of the alert budget this time.
It wouldn't surprise me if there were some near the top or bottom, even further off-plane than Thor is. There's a clump of uninhabited systems barely outside Thor's initial 15 LY radius that it's having real trouble finding a way into.
If attacking control systems was feasible this might be a good excuse for it since taking out one of those 15 systems could protect multiple systems next tick. But you'd still need to predict the exact attack priorities to know how to shape it.
They almost certainly use actual distance. If the were using cube distance, the distribution of attack areas is incredibly under-represented in the corners.
In the models I've been doing I've been assuming that a control can place an alert the week after it becomes a control (i.e. every alert is placed by a system that was a control system the previous week). This assumption has generally held up, and it explains why the alerts show up the week AFTER the maelstroms arrive.
I'll see if I can figure out Thor's next week using the good old fashioned manual process. But Disemboweled Ego is right, all we'll really be able to do is eliminate certain combinations, a true prediction is going to be pretty close to impossible, and well and truly impossible once Thor joins its friends at having enough systems in range to hit 20 every week. We still don't have solid theories for the following questions:
1. When there are more than 20 points available, how does the maelstrom choose which control systems spawn alerts?
2. When a control system has multiple choices of alert, how does it pick between them?
And without being able to attack control systems, there's not a whole lot we can do with this data, anyway. One thing I want to investigate is if there's anything noteworthy in a system we KNOW is going to be an alert next week (I think we should be able to find at least one around Thor)
That's certainly true - I'm thinking more that if it's possible to get a roughly accurate prediction for this week in advance, it would be stronger for the theory than just another retrospective week of "well, it doesn't contradict it".
If we can at least get a candidate list for each of its (usable, non-exhausted) control systems of where it might place an Alert, we might be able to get some evidence towards your questions by which gets picked.
Agreed, this is a long way off being of practical use, but if we can get some information out of Thor while it remains at least somewhat constrained, it might be more useful for containing Maelstrom Nine.
We have enough past data to statistics+brute force it probably if we look at all the systems past invasions could've invaded, what they didn't and what they invaded some cycles after it was first in range.
Col 285 Sector RW-D b12-1 and Col 285 Sector EA-Q c5-12 alerted this tick, the rest did not. That leaves 13 potential attackers. But we don't really know how "smart" the system is: If it knows it will not hit 20 systems, does it prioritize populated systems? Will one control system block out another unnecessarily (i.e. if it had a choice but chose one that another control system had no choice about)?
You can draw up a mapping between attackers and potential alerts:
But it's immediately clear that the order in which you find targets is going to affect the outcome (though we can be pretty confident that Unktety is going to be hit)
Maximizing points used will get 18 points on 12 alerts. 12 is going to be the maximum number of alerts, while I think 11 is the minimum (with only one populated). (CORRECTED
The following systems will be alerted:
Unktety
Col 285 Sector SH-C b13-1
Two of:
Col 285 Sector SM-C b13-4
Col 285 Sector QR-C b13-2
Col 285 Sector HL-O c6-22
Aruntei
At least one of which will be Col 285 Sector QR-C b13-2 or Col 285 Sector HL-O c6-22
One or two of:
Col 285 Sector QW-D b12-2
Col 285 Sector PW-D b12-0
Two of:
Col 285 Sector KV-F b11-2
Col 285 Sector IA-G b11-0
Col 285 Sector KV-F b11-3
Hyades Sector NH-M b7-3
but not both Col 285 Sector KV-F b11-2 and Col 285 Sector KV-F b11-3
Four of: (CORRECTED)
Col 285 Sector GF-G b11-5
Col 285 Sector HF-G b11-3
HIP 21548
Col 285 Sector HF-G b11-4
Col 285 Sector HF-G b11-5
Col 285 Sector FU-H b10-2
Hyades Sector NM-M b7-1
Hyades Sector SI-T c3-4
Hyades Sector JG-O b6-3
With this matrix (only one entry allowed in each row or column)
Col 285 Sector SH-C b13-1
Two of:
Col 285 Sector SM-C b13-4
Col 285 Sector QR-C b13-2
Col 285 Sector HL-O c6-22
Aruntei
At least one of which will be Col 285 Sector QR-C b13-2 or Col 285 Sector HL-O c6-22
One or two of:
Col 285 Sector QW-D b12-2
Col 285 Sector PW-D b12-0
Two of:
Col 285 Sector KV-F b11-2
Col 285 Sector IA-G b11-0
Col 285 Sector KV-F b11-3
Hyades Sector NH-M b7-3
but not both Col 285 Sector KV-F b11-2 and Col 285 Sector KV-F b11-3
Four of: (CORRECTED)
Col 285 Sector GF-G b11-5
Col 285 Sector HF-G b11-3
HIP 21548
Col 285 Sector HF-G b11-4
Col 285 Sector HF-G b11-5
Col 285 Sector FU-H b10-2
Hyades Sector NM-M b7-1
Hyades Sector SI-T c3-4
Hyades Sector JG-O b6-3
With this matrix (only one entry allowed in each row or column)
Edit: I made a mistake. Corrected now.
Last edited:
Looking at some of your unexplained invasions, I believe these are the invasions that occurred when Maelstroms landed. When a maelstrom arrived, every populated system between 15 and 20 ly away had an instant invasion, even if it wasn't within 10 ly of a control system. And your unexplained Leigong alerts are likely the invisible systems. I do have a full set on my sheet (they're the systems highlighted in yellow) if you want them.
The attacker for Col 285 Sector SH-B b14-7 is Col 285 Sector WN-Z b14-0
With the number of times Oya has gone 5:0 or 4:4 (all of them, since it got to reliably use a full budget), plus the way "enclave" defended systems tend to be immediately re-attacked while some uninhabited systems <20LY away are still being left alone, I wouldn't be surprised if there was a general priority to hit inhabited systems.
which is also a strong case for not using cube-calculation, but distance. there were some system unattacked which would have been targets with cube calculations.
I'm fairly sure these will be able to attack again and the invasion system only looks at the current state at the end of the tick when making its next move.
I wasn't easily able to discern that yet from looking at what systems get passed over because re-invasions make it harder to count.