Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[Serenitis]

The AI has a 'thing' about salamanders. It really doesn't like them anywhere near the ships its controlling, and will go to great lengths to (and do a decent job of) avoiding them.
However, salamanders need room to work.
If the firing ship is too close to the target there's a chance that the missiles will just hit whatever is in front of them as they leave the launchers, before they start thier maneuvering.
And also the firing ship being close to the target can encroach on the 'threat level' of the missiles enough that the AI won't turn enough to expose anything, while still blocking the missiles.
The AI ship needs enough space to feel 'safe' enough to give all its attention to the missiles.

Shrike is primarily a skirmisher full of energy weapons which have limited range. Space is something it can't afford to give its targets, it just doesn't have the reach.
So in order to use it's mostly short-range weapons, it will almost always be too close to effectively take advantage of the quirk with salamanders,
A player can do it. Fire missiles, wait for target to turn, then burn in and engage.
An AI can't do that (yet).

[Grievous69]

This may be a teeny tiny bit off topic but I'm curious. Do people put Salamanders on warships? I've done so before because I thought "they're unlimited so it can't hurt", but as time passed by, I realised my ships do way better with just a few extra caps or hullmods. Salamanders are really costly on OP if you take a look what AI gets out of them. Now I'm not saying they're bad as missiles, they certainly have a place, but I can't think of a combat ship that would do equally well as with other missile options.

Actually Doom has a decent benefit from Salamanders due to mines, but that's such a unique thing that doesn't appear anywhere else. Like would you put them on a Dominator, Aurora, Onslaught, Odyssey, Eagle, Apogee, new Fury, or even Enforcer for some reason? I don't think the game would break if the OP cost got reduced to 4/8 honestly. Maybe I'm just using them wrong who knows.

[Igncom1]

I find them to be all right as a way of addition additional EMP to a ship that would otherwise not have access to that damage type. They can mess up space stations real good.

But yeah they are expensive. If you need to min-max it's often good to just not bother with missiles altogether. But otherwise they can be nice to slot onto lowtech ships that might otherwise just be stuffed full of missiles they don't need or the AI forgets to use half the time (like torpedoes.)

[Megas]

I use Shrike (P) like with any other warship.  Shrike (P) is practically a half Medusa (thanks hybrid mount), and I use it as such.  I use it as a cheap brawler, and Shrike (P) does that decently.  Shrike is piloted by AI, because my ship is the starter Apogee.  I only use Shrike (P) early in the game, before I get cruisers and capitals.  Shrike (P) is part of the trio of destroyers that pirates overuse, and my early-game fleet is composed primarily of Enforcers, Mules, and Shrike (P)s, and I occasionally buy a pristine Shrike (P) or two from the Black Market.  By the time I can build Shrikes and the weapons it needs, it is obsolete and I never touch it.  Yes, I would like Sabot pod, but I generally do not get them until after Shrikes become obsolete.

I do not trust Harpoons without Missile Specialization.  They are too slow, too unreliable, and not enough ammo.  If I use Harpoons, it is because that is all I have available on hand.

This may be a teeny tiny bit off topic but I'm curious. Do people put Salamanders on warships? I've done so before because I thought "they're unlimited so it can't hurt", but as time passed by, I realised my ships do way better with just a few extra caps or hullmods. Salamanders are really costly on OP if you take a look what AI gets out of them. Now I'm not saying they're bad as missiles, they certainly have a place, but I can't think of a combat ship that would do equally well as with other missile options.

Actually Doom has a decent benefit from Salamanders due to mines, but that's such a unique thing that doesn't appear anywhere else. Like would you put them on a Dominator, Aurora, Onslaught, Odyssey, Eagle, Apogee, new Fury, or even Enforcer for some reason? I don't think the game would break if the OP cost got reduced to 4/8 honestly. Maybe I'm just using them wrong who knows.

I use them early game because enemy fleets are not too big and the ships are small and fairly weak, meaning they will score hits.  Early-game is more favorable to dueling.

Later in the game, fights are generally too big for Salamanders to be effective.  Doom is an exception because it and the enemies that fight it peel away from the melee and get into a duel.  I have seen Salamanders herd the AI into mine kills, so I like Salamanders on Doom.

As for ships...
Enforcer is used early in the game, and I use whatever I loot.  Salamanders are useful early.  I prefer Annihilators on Enforcers, but I use what I can find.

I use Salamanders on other early game ships like Mules.

I do not use Salamanders on Dominator or Onslaught.  Their dissipation is atrocious and if ballistics they can support alone is not enough firepower (it is not against Radiants), then I need missiles (Annihilators and Expanded Missile Racks) for more damage.

I use Salamanders on Tempest.  Also on Apogee before I find Locusts.

I do not use Salamanders on Aurora or Odyssey, the poster children of naked hull warships.  Missile mounts are Sabots or empty.

Conquest needs Expanded Missile Racks for more large missiles than any medium missile, and Conquest does not have a lot of OP to spare.

I might put Salamanders on Falcon and Eagle if I have OP to spare.  Otherwise, missile mounts are empty.

In general, if I do not have enough OP, missile mounts are the first to get sacked for OP.

I would not mind Salamanders being worth less.

[Thaago]

I usually prefer more aggressive missiles than salamanders, but I'll put them on ships whose job it is to be away from the battle line harassing more valuable enemies (the good old distraction Wolf with unstable injector and a search and destroy order is a good candidate) and those with the fast missile racks system. FMR + Salamanders = a lot more Salamanders, to the point where even with good PD a few will get through. The AI also tends to use it correctly, spamming all the charges at once - a Condor with Salamanders isn't as good as  Drover but it has value.

[intrinsic_parity]

What did you mean when you wrote "especially for results that only difference by 1-2"?

My last comment was mostly in reference to the results about how frequently things retreated/won. Particularly for the frequency of the wolf retreating, the numbers were like 0,1,2,4/24 runs and I was just saying that those numbers could easily be due to variance rather than the performance of missile. Also, regarding the frequency that the shrike won against the prometheus (if one missile won 18 times and the other won 14 for instance), that's probably not enough of a difference to conclusively say that one missile out-performed the other. I'm not saying you were making those claims, just pointing out that there are some conclusions that are tempting to draw from the data that we probably shouldn't.

There's definitely useful information in the data though, and it's very interesting to look at it and theorize about the game.

I'm also wondering if the 1x shrike with annihilator pod got shut down by PD, but two shrikes with pods got through and were able to win easily because of that. 1x annihilator sucked super hard and 2x was amazing. There could be some variance involved there as well.

See Thaago's comment above.    With numerical superiority, ships gets more opportunities to strike.

Also worth mentioning: the Prometheus has 2 Railguns and 1 Light Needler. So kinetic damage and a higher range than Shrike. So a nightmare for AI Shrike, and probably an unpleasant experience for a human player.

Also in 1v1 situation, if target is stronger Annihilators are wasted against enemy shield.

I agree that 2v1's generally allow for maneuvers and situations that don't arise in 1v1's and present opportunities to fast ships like shrikes to deal more damage, but I was thinking there might be some interaction beyond that because the annihilators seemed to improve far beyond any other missile in the 2v1 vs the 1v1. I was wondering if there was some 'saturation' point for the PD where 1 annihilator pod was getting shut down by PD before reaching shields, but 2 annihilators pods were able to get through and gain value, creating a very binary performance result. Annihilators hitting shields build hard flux which is always good shield pressure since it costs no flux to fire, while annihilators getting shot down by PD only builds a small amount of soft flux due to PD weapons which is much less significant and often just disappears into dissipation. I don't even know if the prometheus had PD though, so I could be completely wrong.

Actually Doom has a decent benefit from Salamanders due to mines, but that's such a unique thing that doesn't appear anywhere else.

It's funny, I would almost argue that doom mines perform the same role as salamanders (making salamanders unnecessary) in a lot of scenarios/loadouts. I tend to pop doom mines behind stuff to prevent them from retreating and force their shields to the back while I land shots in the front. Reapers work REALLLY well for this. I suppose you could try to use salamanders and pop doom mines in the front to deal damage, but it feels like that would be less reliable than other options.

You could also try to make the aurora use that strategy with salamanders to distract shields and conventional weapons to deal damage. I wonder how well that would work. Probably well in duel, but poorly in a fleet since the salamanders will often not be able to get to enemy engines in a big group.

[Megas]

It's funny, I would almost argue that doom mines perform the same role as salamanders (making salamanders unnecessary) in a lot of scenarios/loadouts. I tend to pop doom mines behind stuff to prevent them from retreating and force their shields to the back while I land shots in the front. Reapers work REALLLY well for this. I suppose you could try to use salamanders and pop doom mines in the front to deal damage, but it feels like that would be less reliable than other options.

Salamanders make it easier to pin enemies with mines, and if a Salamander connects and knocks out an engine, the enemy is pretty much dead when the mines go off.

Mines do not respawn fast enough when I want to spam them.  They are a bit like Hydra from Diablo 2.  It is hard to pin enemies with five mines alone.

You could also try to make the aurora use that strategy with salamanders to distract shields and conventional weapons to deal damage. I wonder how well that would work. Probably well in duel, but poorly in a fleet since the salamanders will often not be able to get to enemy engines in a big group.

Not practical except in one-on-one duels in the sim.  (I used 0.8a era Odyssey when I experimented with that idea.)  It works, but if the enemy has 360 shields, your ship needs to run far enough away to fool enemy AI to drop shields altogether before your ship makes another attack run.  Basically, a slower variation of Hyperion cheese usable by fast big playerships.  Player is unlikely able to do this in a real fleet battle.

That was how three lance 0.8 era Odyssey could kill Paragon in the sim.

[pairedeciseaux]

What did you mean when you wrote "especially for results that only difference by 1-2"?

My last comment was mostly in reference to the results about how frequently things retreated/won. Particularly for the frequency of the wolf retreating, the numbers were like 0,1,2,4/24 runs and I was just saying that those numbers could easily be due to variance rather than the performance of missile.

Thanks for elaborating, I think I see your point. Yet I'm not sure: do you mean "variance" as in statistics, or "variance" as some kind of external force? 

(I'm not a native speaker, so sometimes I'm confused by casual / formal language and by technical / non technical language)

Please note those retreats were ignored in the Wolf test performance assessment. Maybe I should have refrained from putting the "outcome" graphic in the Wolf case because it can mislead people. Hopefully you got (or will get) the important bits that came after.

Also, regarding the frequency that the shrike won against the prometheus (if one missile won 18 times and the other won 14 for instance), that's probably not enough of a difference to conclusively say that one missile out-performed the other. I'm not saying you were making those claims, just pointing out that there are some conclusions that are tempting to draw from the data that we probably shouldn't.

I agree one shouldn't draw too much conclusions from such synthetic tests. The performance assessment is valid in the test conditions, aims to be objective, yet contains subjective elements.

If you care about the results for this test, I encourage you to have a closer look at them: outcome graphic, ranking, time distribution graphic. The ranking is based on the battle outcome first, and then on the battle duration of won battles (as seen in the time distribution graphic).

The differences are huge from rank (1) to rank (2), and from rank (2) to rank (3). For rank (5) there is not enough data, and Shrike was largely operating outside of its PPT... which could be considered as invalid test if we were not doing a comparison.

From best to worst, the test results are:
(1) Sabot: 24 win; median is 136.5
(2) Reaper: 18 win; median is 219
(3) Harpoon: 8 win; median is 228
(4) Pilum: 7 win; median is 255
(5) No missile: (7; 344), Salamander (7; 430) and Annihilator (2; 395)

So rank (1) is head and shoulder ahead, rank (2) is the last acceptable performance, rank (3) is pretty bad already. I suspect if anything would change by running 10 times more repetitions, it would possibly be rank (3) vs rank (4) ... but it would not really matter in the end because the performance might stay bad anyway.

If you and/or others need explanations about the results and/or wants to see the raw data, feel free to ask!    For instance I guess the distribution graphics can be quite confusing to people that have never seen such a thing. And admittedly I haven't explained much about the process nor the results in general.

[intrinsic_parity]

Thanks for elaborating, I think I see your point. Yet I'm not sure: do you mean "variance" as in statistics, or "variance" as some kind of external force? 

(I'm not a native speaker, so sometimes I'm confused by casual / formal language and by technical / non technical language)

Variance in the statistical sense. Although I should clarify: I was thinking that the performance metrics used are based on the missile in use as well as some other random variables describing the AI behavior/battle so then there might be some difficulty in distinguishing between the effect of the missile and the effect of the random AI behavior which might also cause the mean results for two sets of data to be different. I said 'due to variance' before, when I meant 'due to the underlying random variables which cause variance in the performance metric', sorry if that was confusing. At least that's how I was thinking about it.

Of course there are big differences in performance that are clearly a result of the choice of missile, but for missiles that perform somewhat similarly, it's not as clear (to me) how much of the difference in performance is due to the missile choice as opposed to randomness in the AI behavior itself. I could be wrong about the significance of that underlying randomness though.

[pairedeciseaux]

Ahh, thanks for clarifying. It does make sense, and my thinking on this is the same as yours. 

It is true that a lot of randomness/noise is embedded in the data. The way I dealt with that fact in this test series is:

• run enough repetitions
• consider sum of won battle duration as the prime metric for part 1, 2 and 4 of the series
• consider battle outcome as the prime metric for part 3 of the series
• do not use arithmetic mean (average) of won battle duration
• use median of won battle duration as the secondary metric
• when both sum and median felt too close for two datasets, consider performance to be the same
• always perform a visual check of the distributions
• avoid magical/arbitrary interpretation, manipulation and aggregation of data

All of this translate into having high variance in input raw data (inevitable when missiles are involved) and yet relatively robust output metrics.

If still in doubt about this:

• have people review the process/methodology to spot flaws
• have people run the same tests, see if the same results are obtained
• always compare test results to real in-game experience

For reference:

https://en.wikipedia.org/wiki/Median
https://en.wikipedia.org/wiki/Median#/media/File:Visualisation_mode_median_mean.svg
https://en.wikipedia.org/wiki/Arithmetic_mean#/media/File:Comparison_mean_median_mode.svg