The Tyranny of the Accuracy Equation

[Cordane]

I started off trying to understand why Evasion (not even "effective Evasion") is hard-capped at 90%. Mostly this impacts Corvettes, as only the most specialized of late-game Destroyers can reach the current cap otherwise, but a Corvette (base 60% Evasion) can reach the cap with nothing more than basic tech improvements to computers and thrusters (Sapient Computer +25%, Impulse Thrusters flat +15). It didn't seem right to me that Corvettes could only improve their Evasion by half while all the other classes could improve by leaps and bounds (base 35% Evasion for Destroyers, 10% for Cruisers, 5% for Battleships and Titans, and 2% for Colossi).

But why was this important? For one, I was trying to justify including Destroyers (and to a lesser extent Cruisers) in otherwise mono BB fleets (with or without Titan support). I was seeing reports where Battleships, using only L-slot weapons, were nearly or just as effective as Destroyers using S-slots in staving off Corvette attacks - certainly well enough while still being useful against other capital ships. But wait - I thought L-slots were intended to be horrible at attacking high Evasion ships. They are if you look at bonuses as only being multipliers against a base rating; even a 100% bonus against a base Tracking rating of 5% should only make the effective Tracking 10%, right?

That's not always the case, though - some bonuses simply increase the rating by their value. For example, sensors offer a flat bonus to Tracking, unweighted by either ship class or weapon slot size. Tachyon Sensors offer +20 to Tracking - that's a 40% increase against the base Tracking for most S-slot weapons, but a 400% increase for most L-slots. Should these types of bonuses be percent multipliers instead (e.g., +40% Tracking is +20 for S-slot Mass Driver, +12 for M, +2 for L) or at least weighted flat bonuses (e.g., +20 for S-slots is +10 for M, +5 for L)?

(More importantly, am I just misreading the charts, in Stellaris or on the Wiki?)

Combining my initial concern about capped Evasion with the latter one about bonuses in general, I would look to change the rules around "effective Accuracy" as follows:

• Tracking and Evasion can each generate a rating greater than 100% before the "effective Accuracy" equation starts (for now, Accuracy still caps at 100%)
• Evasion minus Tracking is still "effective Evasion", with a minimum value of 0%, but no cap on max value
• Accuracy minus "effective Evasion" is still "effective Accuracy", with a maximum value of 100% and a minimum value of 0%

By combining very high Evasion with reduced ability for L-slot weapons to build Tracking, Corvettes should be much harder for typical Battleships to take down, forcing fleets to build Destroyers to follow the now better-enforced rock-paper-scissors. Of course, there also needs to be efforts made to make single Destroyers more viable against pairs of Corvettes. (Question: do Cruisers have similar issues with taking down pairs of Destroyers, or Battleships against pairs of Cruisers?)

One area where there could be issues with these proposed changes is the Accuracy stat. Right now, Accuracy still caps at 100% because - in its basic form - you can't be more than 100% on target, and the normal aspect of accuracy in shooting at a moving target is already covered by Tracking, so there's nothing else that improved Accuracy can counter to justify it being above 100% to start. This is especially important for L-slot and larger weapons that can't improve Tracking well, because of their low base value, but could still get great improvement to effective Accuracy with bonuses against their high base Accuracy.

Perhaps diminishing returns for overflow Accuracy (2:1 or 3:1)? If a weapon has overflow to its Accuracy stat, which would lead to situations where the "Accuracy minus effective Evasion" equation comes up greater than 100%, could there be a small proportional damage bonus applied, representing precision firing at vulnerable spots, rather than just "center mass"? Would that favor high-Accuracy/low-Tracking (L/X-slot) builds too much? Could overflow Tracking (beyond what is necessary to counter Evasion) have a similar diminishing returns bonus (e.g., "extended lock-on")?

 

[Kinkness]

Seems legit, though some of it went a little over my head

 

[Diettinger]

Out of curiosity, is tracking defined as
i) the ability to detect a target, perhaps with the resolution to determine evasive maneuvers very quick, or
ii) the ability to follow targets with a weapons bearing.
I think part of the problem with this dimension is that there appears to be a blending of the two distinctly different attributes.

 

[Cordane]

Out of curiosity, is tracking defined as
i) the ability to detect a target, perhaps with the resolution to determine evasive maneuvers very quick, or
ii) the ability to follow targets with a weapons bearing.
I think part of the problem with this dimension is that there appears to be a blending of the two distinctly different attributes.

Just looking at basic data, you would think that Tracking is based mostly on the ability of the mount to match the weapon's bearing to the target's. There may be a portion that is trying to consider the range to the target, i.e., information on a target that an S-slot would shoot at would deal with less sensor lag than one an L-slot would shoot at.

Just playing with numbers, but if you look at, for example, an S-slot Mass Driver's 50% Tracking and simply say that the 4:2:1 ratio of the slot sizes has the L-slot turret at 12.5% for a same range target. Then double the range (50 vs. 100 for max range on each), and the L-slot's Tracking would be at like 6.25% (compared to 5% normally). The M-slot version might be better viewed as 16.67% Tracking (twice the size, +50% range), or barely more than half of the current 30%.

If you were to do something like that, Tracking would have two components: speed/precision of turret movement and sensor timeliness. (Projectile speed might also be a factor, similar in nature to sensor timeliness, but not improved by the same technologies - might be easier for this discussion to just focus on the first two.) Improvements in turret movement would proportionally improve each slot, but sensor technologies (especially Tachyon or Precognitive) might basically eliminate the range difference portion of the Tracking limitation.

I had previously asked what kind of RW distances combat might be taking place at, specifically for understanding the ratio of the two Tracking components. Battles typically taking place at light-second or greater ranges would have a much larger component of Tracking being range-related - base Tracking scores would look about the same, but switching to a Tachyon or Precognitive Sensor would eliminate a majority of the difference between S- and L-slots. Conversely, a battle at Star Wars "knife fight" ranges would have Tracking be almost entirely about turret speed, and a sensor upgrade might mean virtually nothing.

 

[James_K]

I think the problem here is that accuracy and evasion shouldn't be additive.

I'd suggest a system where chance to hit = Accuracy / (Accuracy + Evasion). This lets Accuracy and Evasion operate as unbounded values (any positive number would be valid), but has diminishing returns built-in. However, I'm not a programmer, so I don't know if this approach would be complicated enough to adversely affect game performance.

 

[roman566]

I don't understand the point of this exercise. Why boost something in combat only to change multiplayer meta? Because it will affect only multiplayer. The AI will keep its mixed fleets, the player will not bother with destroyers as they die too often in such battles, nothing will change.

More importantly, if the tracking gets nerfed what makes you think destroyers will be used as screening elements? For all we know, this new meta will bring mono cruiser fleets back, or better - mono destroyers.

 

[Shermanator]

Seriously why won't Paradox just nerf Corvette evasion? (by a meaningful amount) I get that it certainly wouldn't solve all the balance issues but it would certainly go a long way to making corvettes more balanced against destroyers and cruisers.

 

[Belhedler]

Nothing new. It has been designed to counter the well known naked corvette strategy that players used to abuse.

 

[Slynx]

if evasion will be uncapped - that'll mean that you'll use all the bonuses on corvettes computer, afterburners, thrusters, etc (in addition to admiral and other things). which will make them even better then right now.
with a nerf to evasion of the L slot I think that may lead to a situation where BB will be obsolete. cuz everyone will be building full evasion corvettes which is almost impossible to hit with big guns. on the other hand corvettes can use G weapons to deal massive damage to ships without PD

iirc corvettes > destroyers if they have same or even more +hp techs. so people will be discouraged to build destrs during the early game either (there is not enough bonuses to tracking anyway yet)

which may lead to a full corvette meta again. not that I'm opposing it, but I though people didn't liked it.

 

[SteelCrow]

But why was this important? For one, I was trying to justify including Destroyers (and to a lesser extent Cruisers) in otherwise mono BB fleets (with or without Titan support). I was seeing reports where Battleships, using only L-slot weapons, were nearly or just as effective as Destroyers using S-slots in staving off Corvette attacks - certainly well enough while still being useful against other capital ships. But wait - I thought L-slots were intended to be horrible at attacking high Evasion ships. They are if you look at bonuses as only being multipliers against a base rating; even a 100% bonus against a base Tracking rating of 5% should only make the effective Tracking 10%, right?

We can mostly solve this by mixing in Battleships that use Broadside and/or Hangar sections. Battleships still need some L-slot weapons to mitigate their slow speed. Cruisers can use Picket computer, so I'd say those are better than Destroyers against Corvettes. Corvettes disengage easily, so the more M- or L- slot weapons that can hit them the better.
Battleships struggle most at Point-Defense. This is when Destroyers are most valuable.

I think the problem here is that accuracy and evasion shouldn't be additive.

I'd suggest a system where chance to hit = Accuracy / (Accuracy + Evasion). This lets Accuracy and Evasion operate as unbounded values (any positive number would be valid), but has diminishing returns built-in. However, I'm not a programmer, so I don't know if this approach would be complicated enough to adversely affect game performance.

Armor used to function somewhat like this. Might work better, but I like how Tracking makes evasive ships vulnerable to short range weapons, so Evasion and Tracking should compare additively. Thus:

chance to hit = accuracy * (100 / (100 + max(0, evasion - tracking)))

where "max" means take the larger of (this, or this) and "100" establishes the scale for Evasion and Tracking.
What could be confusing is tooltips trying to explain.

Now I did not test it.

But the description and the equations on the wiki do both sound wrong (in any case they do not match):

Ie. I do not think that the equation (on the wiki)

chance to hit = max(0, accuracy - max(0, evasion - tracking))

is the equation used in the game.

As max(0,evasion-tracking) would mean that we always have a value greater or equal to zero and a higher evasion value would help.

The term

max(0,min(0.9,evasion-tracking))

makes more sense than max(0,evasion-tracking).
Also the part
"max (0, accuracy - A), where A = max(0,evasion -tracking)"
looks not right either.
It probably should be:
max(0,min(1,accuracy - max(0,min(0,9,evasion-tracking))))

BTW the last equation is actually how I think the game computes the chance to hit (if not PDX might consider to change it, so that
the last equation is used).

ps. The difference is where it is enforced that some value is within certain limits (before or after substraction of another value).

Depends on your perspective. You aren't wrong; the game might actually enforce limits at this equation to put everything in one place, but it doesn't have to. The programmers could define Accuracy and Evasion's limits elsewhere and substitute them in with a variable like "total_evasion" or "final_accuracy".
From what I've experienced with making mods, programmers do frequently split stuff like this up. I think it's to make editing easier.

From a UI persepctive, though, we already know Evasion caps at 90% (and we assume Accuracy at 100%), so we can just simplify the equation.

 

[Friedrich von Deutschland]

if evasion will be uncapped - that'll mean that you'll use all the bonuses on corvettes computer, afterburners, thrusters, etc (in addition to admiral and other things). which will make them even better then right now.
with a nerf to evasion of the L slot I think that may lead to a situation where BB will be obsolete. cuz everyone will be building full evasion corvettes which is almost impossible to hit with big guns. on the other hand corvettes can use G weapons to deal massive damage to ships without PD

iirc corvettes > destroyers if they have same or even more +hp techs. so people will be discouraged to build destrs during the early game either (there is not enough bonuses to tracking anyway yet)

which may lead to a full corvette meta again. not that I'm opposing it, but I though people didn't liked it.

Isn’t the obvious solution to this buffing the effectiveness of Destroyers? In fact, if evasion is allowed to be uncapped it would make buffing tracking weapons to compensate while nerfing corvette durability, making destroyers corvette killers. If an L weapon typically 1-shots a corvette and has a 1/20 chance of hitting they aren’t much less effective than the small weapons that take about as long to kill. On the other hand if the odds to hit are 1/100 instakills are suddenly balanced again.

 

[Arutar]

I started off trying to understand why Evasion (not even "effective Evasion") is hard-capped at 90%. Mostly this impacts Corvettes, as only the most specialized of late-game Destroyers can reach the current cap otherwise, but a Corvette (base 60% Evasion) can reach the cap with nothing more than basic tech improvements to computers and thrusters (Sapient Computer +25%, Impulse Thrusters flat +15). It didn't seem right to me that Corvettes could only improve their Evasion by half while all the other classes could improve by leaps and bounds (base 35% Evasion for Destroyers, 10% for Cruisers, 5% for Battleships and Titans, and 2% for Colossi).

The reason why evasion was capped was to balance corvettes, plain and simple.

Before the evasion nerf, there hardly was a reason to not go full corvettes if you reached 95%+ Evasion (tech + Events + Admirals), as they were way too powerful.

Even now massed corvettes are still optimal against most space monsters and most endgame Crisis (as they bassicially nullify all the massive one hit, one kill weapons).

So, I do not think corvettes need a buff now, if that is what your post suggests.

 

[Blurb]

The term
max(0,min(0.9,evasion-tracking))
makes more sense than max(0,evasion-tracking)

If evasion is hardcapped at 0.9, then max(0,min(0,9,evasion-tracking) is functionally the same as max(0, evasion-tracking).
Your alternative is only different if raw evasion can go above 0.9, which it currently cannot.
Given that corvettes are the only ships that can consistently hit this upper boundary, they would be significantly improved by this change.

Also the part
"max (0, accuracy - A), where A = max(0,evasion -tracking)"
looks not right either.
It probably should be:
max(0,min(1,accuracy - max(0,min(0,9,evasion-tracking))))

Again, your suggested alternative is only relevant if accuracy is not a bounded value - which it currently is.
I'm not sure about the impact of such a change, as it depends mostly on what "chance to hit" (as from the 'line' ship computer) means:
If it is a bonus to raw accuracy, then the difference is potentially huge, inversely if it is a bonus to effective accuracy then the difference is nil.

Nothing new. It has been designed to counter the well known naked corvette strategy that players used to abuse.

Naked corvettes was about tech progression first and the mechanics of combat second.
When Stellaris launched, higher-tech ship components were only marginally more effective but vastly more expensive.
Without tech progression being so backwards, it would just have been corvette spam as opposed to naked corvette spam.

 

[DeadEyeTucker]

The equations for hit work fine.

As for talking about tracking and the guns being able to move fast enough to keep bearing etc. You're getting into EVE territory here. Where turrets have an optimal range and tracking speed. Tracking speed was in rad/sec (how fast can the gun turn). The turrets could hit enemy ships as long as their orbital speed was less than or equal to the tracking speed of the gun. Orbital speeds in rad/sec is based on tangental velocity of the ship and range from the ship. You could be flying at 5km/s but if you were orbiting far away the large guns would still have no problem hitting you.

I don't think it would be good to implement this as it sounds like a lot of extra back end calculations that would slow the game down. But I would like to see Large slot and XL slot weapons have a good chance to hit corvettes at extreme range. When the corvettes are flying straight in from thousands of kilometers away (or whatever the end of L/XL slot weapons are) then my battleships shouldn't have a hard time of hitting them and should only be constrained by the accuracy (I am guessing the resolution of their sensors and target locks) of their systems and weapons. I think this would help counter corvette spam. Sure you could build a load of torpedo corvettes, but expect to take many losses at extreme range before they close in. This would force pure corvette fleets to have to tactically engage battleships instead of flying straight at them and brawling. Either they have to ambush the battleships as they jump in system or jump in behind them.

 

[Slynx]

Isn’t the obvious solution to this buffing the effectiveness of Destroyers? In fact, if evasion is allowed to be uncapped it would make buffing tracking weapons to compensate while nerfing corvette durability, making destroyers corvette killers. If an L weapon typically 1-shots a corvette and has a 1/20 chance of hitting they aren’t much less effective than the small weapons that take about as long to kill. On the other hand if the odds to hit are 1/100 instakills are suddenly balanced again.

no. it's not obvious.
unless you propose it there is no wayto know if there is anything besides uncapping evasion. else OP would've mentioned it.

if evasion - tracking < 0 then there will be 0 chance to hit.
1/20 chance to hit is 95% evasion. 1/100 is 99%evasion
also don't forget that some weapons have less then 100% accuracy. so the shot can miss on it's own. reducing further chance to hit.

 

[Friedrich von Deutschland]

The equations for hit work fine.

As for talking about tracking and the guns being able to move fast enough to keep bearing etc. You're getting into EVE territory here. Where turrets have an optimal range and tracking speed. Tracking speed was in rad/sec (how fast can the gun turn). The turrets could hit enemy ships as long as their orbital speed was less than or equal to the tracking speed of the gun. Orbital speeds in rad/sec is based on tangental velocity of the ship and range from the ship. You could be flying at 5km/s but if you were orbiting far away the large guns would still have no problem hitting you.

I don't think it would be good to implement this as it sounds like a lot of extra back end calculations that would slow the game down. But I would like to see Large slot and XL slot weapons have a good chance to hit corvettes at extreme range. When the corvettes are flying straight in from thousands of kilometers away (or whatever the end of L/XL slot weapons are) then my battleships shouldn't have a hard time of hitting them and should only be constrained by the accuracy (I am guessing the resolution of their sensors and target locks) of their systems and weapons. I think this would help counter corvette spam. Sure you could build a load of torpedo corvettes, but expect to take many losses at extreme range before they close in. This would force pure corvette fleets to have to tactically engage battleships instead of flying straight at them and brawling. Either they have to ambush the battleships as they jump in system or jump in behind them.

You forget that these are moving targets and only a brain dead moron would fail to take evasive actions when tachyon lances are being pointed in their direction. It’s hard to hit a target when they could be anywhere in a 500km radius by the time your projectile arrives. Tracking therefor is only relevant at closer ranges, since at long range only area of effect and active guidance can possibly neutralize evasion.

 

[DeadEyeTucker]

You forget that these are moving targets and only a brain dead moron would fail to take evasive actions when tachyon lances are being pointed in their direction. It’s hard to hit a target when they could be anywhere in a 500km radius by the time your projectile arrives. Tracking therefor is only relevant at closer ranges, since at long range only area of effect and active guidance can possibly neutralize evasion.

At 300k km it takes light about 1 second to traverse that. The angular velocity is going to be very VERY low at that range. It would be trivial to hit a target that far away with lasers/lances. The only variables affecting your accuracy at that range is real time data and sensor resolution. Which is reflected in the accuracy of the weapons I think. But evasion would be useless at that range.

 

[Ashram Darkstar]

At 300k km it takes light about 1 second to traverse that. The angular velocity is going to be very VERY low at that range. It would be trivial to hit a target that far away with lasers/lances. The only variables affecting your accuracy at that range is real time data and sensor resolution. Which is reflected in the accuracy of the weapons I think. But evasion would be useless at that range.

I agree with your line of thought, but since the speed of light is approximately 300,000 km/s it would be .001 seconds.

edit: my bad I read that as 300 km not 300k km.

 

[Talanic]

You forget that these are moving targets and only a brain dead moron would fail to take evasive actions when tachyon lances are being pointed in their direction. It’s hard to hit a target when they could be anywhere in a 500km radius by the time your projectile arrives. Tracking therefor is only relevant at closer ranges, since at long range only area of effect and active guidance can possibly neutralize evasion.

It would not matter. If tachyons exist as theoretically described (they probably don't), a tachyon lance would hit the target before it was fired. Possibly before it was aimed. Can't exactly dodge a beam going backward through time at greater than light speed.

Again, though, they probably don't exist in alignment with that theory. Other types DO, but I don't really understand them well enough to try to talk about them.

 

[DeadEyeTucker]

I agree with your line of thought, but since the speed of light is approximately 300,000 km/s it would be .001 seconds.

Then it would travel over 300k (that k means thousand) in about 1 second.