alpha 4 Target Fast Speed: How it works and how to fix it

[Evil4Zerggin]

How it works

The chance to be hit is multiplied by approximately 164.2% - 4.15% * speed in knots, but not greater than 100% and not less than 10%. (For the modifier displayed in-game, subtract 100%.) I derived this by watching AI enemies until their Target Fast Speed modifier stopped increasing, then recorded their modifier and top speed. It's possible that the AI runs their ships at cruise speed rather than top speed, but even if so this doesn't change the broad conclusions. I did not record any very slow ships (< 16 knots) but 

What is going wrong

Many people have complained that fast ships are too difficult to hit. It's not hard to see why---the cap is hit at about 37.5 knots, which is attainable by many ships in late game, and makes them 10 times as hard to hit as a ship of 15 knots. Furthermore, the curve has a strange shape. Here is a plot of the hit difficulty (i.e. multiplier to the average number of shells needed to score a hit, which is inversely proportional to the chance to be hit above) versus speed.

• Going from 16 to 19 knots only makes you 1.15x harder to hit
• But, going from 34.5 to 37.5 knots makes you twice as hard to hit.
• But, going from 37.5 knots to 40.5 knots (or even to 49 knots) doesn't make you any harder to hit at all.

This is the problem with using negative modifiers to a value whose reciprocal is important; at some point, you get an ultra-sharp rise that quickly either reaches a cap or infinity.

How to fix it

If we are going to stick with a product-of-factors model of accuracy: rather than using a decreasing multiplier to accuracy, use an increasing divider. For example: 2 / (1 + speed / reference_speed)

If you want a stronger effect of speed, you could square this, effectively applying the multiplier twice. For example: (2 / (1 + speed / reference_speed))^2

For example, if you set reference_speed = 15 knots:

The square version resembles the current curve up to about 28 knots, but doesn't have the sharp takeoff after that that we currently have.

Edited February 9, 2020 by Evil4Zerggin

[Shaftoe]

42.5 knots BCs with 9' armor and 16' guns shouldn't be a thing at all. It's ridiculous. 

[Reaper Jack]

Agreed, also a 7 knot difference from 30.5 to 37.5 knots should not be eight or ten times as effective as the difference from 20-27 knots, that's just poor consistency. I wouldn't be too bothered by poor hit chances on destroyers attaining these speeds, but precious few cruisers and no capital ships ever managed more than 36 knots. With capitals never exceeding 32 until the Iowas and only then making 33 knots. And even then the vast majority of battleships top out at 30 knots with battlecruisers topping out at 31.5 (Scharnhorsts and Dunkerques.) On paper the only capital ship designs to exceed these boundaries before the Iowas were the Project 1047 BC's and the O-Class raiders. Both of which sacrificed massive amounts of armour to hit 34-35 knots. 

Edited February 9, 2020 by Reaper Jack

[NoZaku]

16 minutes ago, Reaper Jack said:

Agreed, also a 7 knot difference from 30.5 to 37.5 knots should not be eight or ten times as effective as the difference from 20-27 knots, that's just poor consistency. I wouldn't be too bothered by poor hit chances on destroyers attaining these speeds, but precious few cruisers and no capital ships ever managed more than 36 knots. With capitals never exceeding 32 until the Iowas and only then making 33 knots. And even then the vast majority of battleships top out at 30 knots with battlecruisers topping out at 31.5 (Scharnhorsts and Dunkerques.) On paper the only capital ship designs to exceed these boundaries before the Iowas were the Project 1047 BC's and the O-Class raiders. Both of which sacrificed massive amounts of armour to hit 34-35 knots. 

There were quite a few designs from the 20's and 30's that were intended to achieve ~35kn (even comparatively reasonable ones)

Still, I reckon short of nuclear power or the like 35-36kn should be about the cap for large vessels. 

[akd]

The way to fix it is to modify accuracy using relative speed (i.e. rate of range change and rate of bearing change) not absolute speed.  There is no difference in the gunnery problem (ignoring own ship speed effects like vibration, spray, etc.) presented by two ships sailing on parallel courses at 20 knots and two ships sailing on parallel courses at 30 knots.

[Reaper Jack]

18 minutes ago, NoZaku said:

There were quite a few designs from the 20's and 30's that were intended to achieve ~35kn (even comparatively reasonable ones)

Still, I reckon short of nuclear power or the like 35-36kn should be about the cap for large vessels. 

Intended to achieve is just that though, many designs had fundamental flaws after production that there was no or at least very little method of seeing in the design itself. The Yamatos in particular come to mind here, as well as over-engineered German ships post Deutschlands. The KGV's as well with quad turrets that had issues for several years, same issue for the French. Some designs would exceed their designed speeds (Scharnhorst and Gneisenau both did by half a knot.) while others fell short. (Bismarck almost never made her 30 knot design speed, one of the reasons the Tirpitz was given larger engine power.) While other navies fudged the numbers so their ships appeared faster. (Soviet Navy and Tashkent, doing speed trials with low fuel, no guns installed and no ammo carried, on the calmest sea in existence.) 

Another note on the Bismarck in particular; from Battleship Bismarck, A Survivor's Story.  'As the ship's stability and manoeuvrability were being tested, a flaw in her design was discovered. When attempting to steer the ship solely through altering propeller revolutions, the crew learned that Bismarck could be kept on course only with great difficulty. Even with the outboard screws running at full power in opposite directions, they generated only a slight turning ability' 

Edited February 9, 2020 by Reaper Jack

[arkhangelsk]

1 hour ago, Evil4Zerggin said:

How it works

The chance to be hit is multiplied by approximately 164.2% - 4.15% * speed in knots, but not greater than 100% and not less than 10%. (For the modifier displayed in-game, subtract 100%.) I derived this by watching AI enemies until their Target Fast Speed modifier stopped increasing, then recorded their modifier and top speed. It's possible that the AI runs their ships at cruise speed rather than top speed, but even if so this doesn't change the broad conclusions. I did not record any very slow ships (< 16 knots) but 

What is going wrong

Many people have complained that fast ships are too difficult to hit. It's not hard to see why---the cap is hit at about 37.5 knots, which is attainable by many ships in late game, and makes them 10 times as hard to hit as a ship of 15 knots. Furthermore, the curve has a strange shape. Here is a plot of the hit difficulty (i.e. multiplier to the average number of shells needed to score a hit, which is inversely proportional to the chance to be hit above) versus speed.

• Going from 16 to 19 knots only makes you 1.15x harder to hit
• But, going from 34.5 to 37.5 knots makes you twice as hard to hit.
• But, going from 37.5 knots to 40.5 knots (or even to 49 knots) doesn't make you any harder to hit at all.

This is the problem with using negative modifiers to a value whose reciprocal is important; at some point, you get an ultra-sharp rise that quickly either reaches a cap or infinity.

How to fix it

If we are going to stick with a product-of-factors model of accuracy: rather than using a decreasing multiplier to accuracy, use an increasing divider. For example: 2 / (1 + speed / reference_speed)

If you want a stronger effect of speed, you could square this, effectively applying the multiplier twice. For example: (2 / (1 + speed / reference_speed))^2

For example, if you set reference_speed = 15 knots:

The square version resembles the current curve up to about 28 knots, but doesn't have the sharp takeoff after that that we currently have.

The modifier is actually linear (I've used actual ship speeds instead of maximum speed), but yeah at the ends the effect won't be very linear at all.

43 minutes ago, akd said:

There is no difference in the gunnery problem (ignoring own ship speed effects like vibration, spray, etc.) presented by two ships sailing on parallel courses at 20 knots and two ships sailing on parallel courses at 30 knots.

Actually to be fair, this is something that's true on paper but less in reality, because this is only true IF the ships' courses are exactly known. In reality, of course, you are guessing, either by looking at the scatterplot of ranges and bearings over time being thrown out by your towers or guessing the course by reading angle on bow and similar visual cues (if the target turns, you'll have to start by doing that because your scatterplot takes time to accumulate new data). And any error in the course reading translates to an error in the intercept position which does grow in proportion to the speed - if for example you are thirty degrees off, at a flight time of 1 minute at 18 knots the error is sin30*600=300 yards. At 36 knots, that becomes 600 yards.

Edited February 9, 2020 by arkhangelsk
Removed one photo to reduce height

[RedParadize]

55 minutes ago, akd said:

The way to fix it is to modify accuracy using relative speed (i.e. rate of range change and rate of bearing change) not absolute speed.  There is no difference in the gunnery problem (ignoring own ship speed effects like vibration, spray, etc.) presented by two ships sailing on parallel courses at 20 knots and two ships sailing on parallel courses at 30 knots.

I originally liked your post, but after reflection I changed my mind. Its not that I do not like the basic idea, problem is that ship in this game almost always run parallel. Even when going in opposite direction, ship are circling around a central point both range and bearing remain constant. So most case scenario there would be no penalty if just Δ range and Δ bearing is used. Something else need to be added on top of that.

After a quick reflection, what could work is something like: "How much do you have to shot ahead of target to hit it, vs targeting accuracy and shell travel time" In other term, more you have to shoot ahead, bigger is the effect of range and bearing error in the targeting solution. This compound with target size obviously. But ignoring that part for a moment, it mean that it can be reduced to shell travel time vs target speed. Note that bigger shell are slower...

What do you think?

[Evil4Zerggin]

2 hours ago, Shaftoe said:

42.5 knots BCs with 9' armor and 16' guns shouldn't be a thing at all. It's ridiculous. 

Agreed, but this isn't an issue with the Target Fast Speed modifier per se, more to do with the later hulls having absurd power-to-weight ratios for their powerplants.

Also the power curve is based on purely quadratic drag, i.e. required HP grows as the cube of speed, and doesn't take into account wavemaking resistance, but this type of curve (with proper coefficients) might be good enough given that we don't have fine control over hull form etc.

1 hour ago, arkhangelsk said:

The modifier is actually linear (I've used actual ship speeds instead of maximum speed), but yeah at the ends the effect won't be very linear at all.

The modifier is linear, but it's the reciprocal that we care about, and a linear modifier creates a very non-linear reciprocal, which is what I plotted. A similar thing happens with armor weight modifiers and reload time modifiers, though not as severely since in those cases the modifiers don't reach as large as -90%--imagine getting x10 armor thickness per weight or x10 rate of fire!

Edited February 9, 2020 by Evil4Zerggin

[arkhangelsk]

But that would have to be entered in terms of becoming a formula using percent. Perhaps this polynomial function would suffice?

The destroyers might get murdered by this, though. As it was, it wasn't very difficult to sink them using modern weapons in Modern Battleship vs Destroyers. All 12 of them went down in one hit by 12 inchers (Hero Weapon). The only thing is the torpedoes.

[Evil4Zerggin]

1 hour ago, arkhangelsk said:

But that would have to be entered in terms of becoming a formula using percent.

You can turn any multiplier to hit chance into a percentage modifier just by multiplying by 100 to get percent, and then subtracting 100. (This works for accuracy since they stack like multipliers, rather than the percentage modifiers being added together like most shipyard stats.)

1 hour ago, arkhangelsk said:

The destroyers might get murdered by this, though. As it was, it wasn't very difficult to sink them using modern weapons in Modern Battleship vs Destroyers. All 12 of them went down in one hit by 12 inchers (Hero Weapon). The only thing is the torpedoes.

That has more to do with the changes to Target Ship Size modifier, where small ships became up to ~4x as easy to hit based on that multiplier alone in Alpha 4 while large ships remained roughly at the same chance. After all, Target High Speed does not distinguish between a 38-knot BC and a 38-knot DD. With respect to Target Ship Size, Alpha 3's displacement effect with Alpha 4's Target Signature effect might be the best combination of what's existed so far, though of course we can always consider completely new values.

Edited February 9, 2020 by Evil4Zerggin

[arkhangelsk]

1 hour ago, Evil4Zerggin said:

That has more to do with the changes to Target Ship Size modifier, where small ships became up to ~4x as easy to hit based on that multiplier alone in Alpha 4 while large ships remained roughly at the same chance. After all, Target High Speed does not distinguish between a 38-knot BC and a 38-knot DD. With respect to Target Ship Size, Alpha 3's displacement effect with Alpha 4's Target Signature effect might be the best combination of what's existed so far, though of course we can always consider completely new values.

I have yet to experience a UA:D, be it Alpha 2, 3 or 4 where destroyers are actually "hard to hit" by modern fire control systems and once hit they always go down very quickly. The only torpedo boats that were ever hard to hit were those in Battleships vs Torpedo Boats, and that's because the guns are so crap on that battleship it can't even hit the armored cruisers - that's where all the challenge came from and the one time I appreciated the value of little guns. Otherwise, my battleships are almost consistently naked of them except when i need something to balance out the ship.

[RedParadize]

@arkhangelsk Destroyer were pretty hard to hit in alpha 3... well except against 18"er. A dd could get in the 2km range up to mid 20s quite easily. One of the main problem is that AI wait too long before launching torp and stick around even after. Now its bit of another story, Destroyer die as they enter range more or less. This obviously is a hidden nerf of some kind.

[Cptbarney]

5 hours ago, RedParadize said:

@arkhangelsk Destroyer were pretty hard to hit in alpha 3... well except against 18"er. A dd could get in the 2km range up to mid 20s quite easily. One of the main problem is that AI wait too long before launching torp and stick around even after. Now its bit of another story, Destroyer die as they enter range more or less. This obviously is a hidden nerf of some kind.

We need moar clarification lol, seems like lots of people are confused as to what has been changed entirely or even added.

[RedParadize]

@Cptbarney Here of some of the picture I took form alpha 3:

Me Killing a BC with only guns:


Enemy AI torpedoing me at 600m

[Cptbarney]

@RedParadize talking about patch notes, as it seems they have changed a load of values but because we dont know what they have changed we cant give them accurate enough feedback unless its well solely based on playing the game and collating data that way (similar to how weegee posts stat changes so we have a ‘trail’ to compare current stats with previous and also gameplay to help provide decent evidence and help to illustrate any further feedback).

[RedParadize]

@Cptbarney In the case of Destroyer I am convinced that they added a direct "to hit" bonus against them. For that reason I think it should be left out of the of the current discussion about speed vs hit chance.

[Cptbarney]

1 minute ago, RedParadize said:

@Cptbarney In the case of Destroyer I am convinced that they added a direct "to hit" bonus against them. For that reason I think it should be left out of the of the current discussion about speed vs hit chance.

Yeah dd’s are fine. But battlecruisers going 45knots is mental lol, nevermind trying to hit the bloody things when they start dodging as well to make mister torpedoes life that much tedious.

Makes me wonder if they are doing extremes then balance the mechanics towards a mid point, unless they have other things they want in the game.

[RedParadize]

@CptbarneyHahaha,I do not think DD's are not fine at all. Try 1 BB vs 6 DDs or more, from both sides. In Alpha 4 they get annihilated at max gun range, often in few salvo. I do manage to launch torpedo, but in agressive mode only. AI never does that and almost always loose all dds before launching a single torp.

Damn, I am derailing the tread again.

[PainGod]

Yeah, that kind of hit modifier is just ludicrous. Failures to hit for targets going fast should mostly be a result of evasive maneuvering, though a moderate, nearly linear decrease in hit probability from speed alone should also be present.

The point where hit probability becomes noticeably inversely exponential to speed is way, way higher than tens or even hundreds of knots.