|
|
Post by Mr. Beast on Jun 26, 2020 16:21:28 GMT
Both cavalry expert and surprise attack look like a good skill for Plumer. Which one should I add to him?
|
|
|
|
Post by Torvesta on Jun 26, 2020 17:21:46 GMT
Surprise
|
|
|
|
Post by Mr. Beast on Jun 26, 2020 19:25:41 GMT
|
|
|
|
Post by Boss Tweed on Jun 26, 2020 19:41:55 GMT
Plummer with Jungle Fighting is already quite fast. Therefore, using Surprise Attack could take that advantage of that high movement more damage.
|
|
|
|
Post by SolidLight on Jun 26, 2020 19:46:17 GMT
It's pretty complicated but I just ran some numbers and yeah Surprise Attack is better. They're about equal if you don't move. If you do move it's better. |
|
|
|
Post by Mr. Beast on Jun 26, 2020 19:48:41 GMT
It's pretty complicated but I just ran some numbers and yeah Surprise Attack is better. They're about equal if you don't move. If you do move it's better. Wait so it triggers even if you don't move? |
|
tokra
First Lieutenant
Posts: 32
|
Post by tokra on Jun 26, 2020 19:58:43 GMT
It's pretty complicated but I just ran some numbers and yeah Surprise Attack is better. They're about equal if you don't move. If you do move it's better. Wait so it triggers even if you don't move? Yes, you will allways have at least 8% if you don't move and it goes up to 20 % if IRC. |
|
|
|
Post by Mr. Beast on Jun 26, 2020 20:07:59 GMT
Wait so it triggers even if you don't move? Yes, you will allways have at least 8% if you don't move and it goes up to 20 % if IRC. So putting cavalry expert on Jackson was a waste I guess. |
|
|
|
Post by Nobunaga Oda on Jun 27, 2020 7:10:51 GMT
It's pretty complicated but I just ran some numbers and yeah Surprise Attack is better. They're about equal if you don't move. If you do move it's better. So it's like it's counterpart from back in 1804? That skill was quite valuable, with an atk boost given for a stationary gen and an increased boost for moving gens. It this has the same property, our priority list of cav gen skills may need to be reviewed. |
|
|
|
Post by Harry Lillis "Bing" Crosby Jr on Jun 27, 2020 8:40:51 GMT
It's pretty complicated but I just ran some numbers and yeah Surprise Attack is better. They're about equal if you don't move. If you do move it's better. So it's like it's counterpart from back in 1804? That skill was quite valuable, with an atk boost given for a stationary gen and an increased boost for moving gens. It this has the same property, our priority list of cav gen skills may need to be reviewed. It has the same property. Hence surprise attack is one of the essential skills for cav that can take precedence over expert. |
|
|
|
Post by ambitiousace on Jun 27, 2020 11:36:53 GMT
Until now I guess we have yet to clarify combat effectiveness or stable output tho but imo Suprise is better,much more benefits in conquests when you use tactics like March or Forced March paired with Tactic Master
|
|
|
|
Post by scipioafricanus on Nov 27, 2020 11:58:02 GMT
Until now I guess we have yet to clarify combat effectiveness or stable output tho but imo Suprise is better,much more benefits in conquests when you use tactics like March or Forced March paired with Tactic Master Basically its a multiplier of ALL other fount of damage. I tested it with artillery expert and its exactly +25% (for art) damage. Its a final multiplier so its totally great, in my opinion a must have, at least for Artillery. Actually the topic is interesting and i’m not sure wich one is better for cavalry...but what if we use them both? Cavalrly expert should multiply even the surprise attack bonus.... |
|
|
|
Post by Erich von Manstein on Nov 27, 2020 13:08:28 GMT
I heard that "combat effectiveness" means increasing the minimum damage. If this is the case then it would work on artillery the best as the fluctuation range of artillery is quite large.
|
|
|
|
Post by jonblend on Dec 4, 2020 23:18:33 GMT
Here we can see the effect of surprise. I tested it for no movement and 3 hexes moved and the results align nicely with the suggested values from the damage formula link
at lv5 surprise hexes moved (average) damage multiplier 0 1.08 1 1.12 2 1.14 3 1.16 4 1.18 5 and more 1.20
Now for the ...Expert skills. There are some test results in the picture above. I also did some exemplary testing for ships and infantry. The results for artillery are in this post link
So the question is, can we use these results to better understand the effect of these "increase stable output" skills? First let's take a look at the "final step" of the damage formula:
"The damage will be an integer randomly selected between [maxDamage * attackFloatingCoefficient] and [maxDamage]"
And now take a look at the atkFloatingC that I calculated by deviding the recorded min damage by max damage. This value stays the same (within the statistical fluctuations) but is significantly different when an Expert skill is involved.
Moreover, I think the altered atkFloatingCoefficient is the basic atkFloatingCoefficient + the value from the skill description!
[regarding the exactness of the calculated value for atkFloatingCoefficient: since I cannot guarantee that I saw the highest and lowest possible damage roll within the sample size of 20-25 we must assume that the "real value" is slightly smaller (lower min/higher max results in smaller number)
Frigate (0.7) with Gunsight 0.85 New inf (0.8) with InfExpert 0.9 Armored Car (0.75) with CavExpert 0.9 Howitzer (0.6) with ArtExpert 0.85
|
|
|
|
Post by jonblend on Dec 8, 2020 17:19:36 GMT
Assuming this is indeed the correct deduction, how will it influence the average damage? Here we need more assumptions  Let's say there is a symmetric probability distribution (gaussian) and all the rounddowns within the damage formula have only negligible effects on this symmetry. Then average damage [av] is (max damage [max] + min damage)/2 Or with the atkFloatingCoefficient [I'll use f for it]: av = 0.5 * (max + max * f) = 0.5 * max * (1+f) And now with an ...Expert skill, where e is the value taken from the skill description av_e = 0.5 * (max + max * (f+e)) = 0.5 * max * (1+f+e) Now the direct impact of the skill on average damage should be x = av_e / av = ... = 1 + e/(1+f) Let's compare the theory to the test results Frigate (0.7) with Gunsight -> 1.0882 vs 1.0528 New inf (0.8) with InfExpert -> 1.0556 vs 1.0543 Armored Car (0.75) with CavExpert -> 1.0857 vs 1.0644 Howitzer (0.6) with ArtExpert -> 1.1563 vs 1.1540
What can I say?
The artillery results are imo pretty solid. There I also have 2 scenarios and the biggest sample size. With cavalry, there was a lot of interference with morale up and down because I was testing surprise with the same setup. The inf multiplier seems very accurate. Navy and Cav might also suffer from the fact that much higher numbers were involved (already high base damage and several other multipliers): It is harder to randomly get a representative amount of results with a given sample size. Thanks to the low unit in city and weapon vs armor coefficient, the distribution for Artillery is quite narrow, making for smaller error margins. Similarly for Infantry.
Anyways, that was quite the dive into statistics while trying to reverse engineer the game mechanics. If someone comes up with a better explanation or notices any mistakes let me know.
|
|
