|
| Thread: Breakdown of the best upgraded Level 7 Original Heroes 3 | |
|
fanofheroes
Famous Hero
|
posted August 07, 2020 05:31 PM |
|
|
Breakdown of the best upgraded Level 7 Original Heroes 3
Hi everyone,
So I had a lot of free time this week and I thought I’d do a little experiment. I saw a video on YouTube made by Harrowk where he created battle simulations to show the best level units in each class. I saw it and it didn’t feel quite right for me so I changed the parameters of the battle to more accurately show the creatures against each other, and see if that would better fit the AI and Fight values. I made a summary using excel of the battles at the bottom.
[img] [/img]
Method
a) AI: The AI is self explanatory. I set up 2 hero’s armies against each other, and hit the ‘auto combat’ button.
b) Magic: Well the game isn’t called ‘Heroes of Might 3’. Magic is an element of the game just as much and when you have dragon’s having spell immunity, Phoenixes having fire immunity, Titans having mind immunity, and Ghost Dragon’s that are undead, that must be for a reason to benefit them in battle. You could even see this play out in the AI’s strategy. Titans would try to blind immediately to take advantage of the range attack, first strike units would cast Frenzy, Armageddon (if they were immune). That being said, I used the Original Gog HOMM3 Complete so that both hero’s would cast spells equally.
c) In Harrowk’s videos he had a single stack of x creatures vs an equal number of other creatures. I didn’t think this set up actually depicts the strengths of the units. It negates the Archangels from resurrection, Chaos Hydras from all-around-attacks, and any breath attacks from Dragons and Phoenixes. These are all qualities that are apart of them, and the computer used these abilities in its strategy when attacking. Also imagine a real battle, each unit would technically be able to work both as a group and individually. So I made 3 equal sized stacks.
d) Army size: Initially I made the battles 1000 vs 1000, but I started to notice that the AI was only using frenzy, blind and curse repeatedly. The truth is, when the armies are so large, using Armageddon, Chain Lightning or any other direct damage spell was useless. It just wasn’t worth the AI to loose mana for. Once the mana was out, then it would just be a strict battle of brute strength and Behemoths ruined every other monster. So I decided to keep the battle 45 vs 45. I approximated 8 months* of level 7 growth, and figured there would be a dwelling here or there too.
e) Primary skills: I figured by around 8 months(updated: not weeks, sorry for the typo) a hero would have primary skills acquired from leveling up (excluding artifacts, enchanted libraries, and any other primary skill booster on the map). To keep it simple, I gave each hero’s primary skills 10. I suppose if I used the 1000 vs 1000 army numbers, I could give primary skills proportional to that accumulation of creatures, but there was no way for me to accurately estimate the hero’s primary skills after getting that many units. It would probably cap prematurely because there wouldn’t be much leveling up at that point.
f) Terrain: all battles were fought in the sand to avoid giving one unit a speed advantage beyond their natural terrain.
g) Update: Secondary skills. I forgot to mention that each hero had pathfinding and logistics as their SS so that it would make each hero as equal as possible.
The results:
Each column I recorded how many units died during the battle. This gave me a ratio of how many units of x must die off to kill units y. Then I divided them to give a real number ratio. So if you look at Archangel (AA) vs Archdevil (DEV) you can compare the cells A22 and A23 vs B22 and B23. Interpreting them, you can say that if it were 360 AA vs 360 DEV, the angels would only loose 30 units. I guess if you split them into 3 groups of 120 each, you may come to that same conclusion. Then the ratio equivalency would me that in large scale battles, 12 archdevils would equal 1 archangel. Now we all know that if you set that up in a simulation, the one angel would loose badly. This is under the premise of large number battles and each angel having the advantage to resurrect each other.
I also made a chart on the upper right, starting at AB1 showing the numbers against each other. the way to read it is by starting on the x-axis first and then the y-axis. Let's say you look at cell AC10 for the titan and go up that column, it would mean the following:
AC8 = you would need 1.55 Titans to defeat 1 Archangel
AC7 = you would need 0.55 Titans to defeat 1 Archdevil
AC6 = you would need 0.31 Titans to defeat 1 Ancient Behemoth
....and so on
[img] [/img]
Final ranking:
1 Archangel
2 Black Dragon
3 Gold Dragon
4 Titan
5 Ancient Behemoth
6 Archdevil
7 Chaos Hydra
8 Phoenix
9 Ghost Dragon
Just slightly different from the AI list in the game, but also notice that the Behemoth has the closest point differential to the Archangel (1.07 to kill 1 angel), but is still ranked 5th (Instead of 2nd). This is because of the special abilities for Titans, Gold and Black Dragons. The Titans would blind the behemoths until one creature stack. and no matter what, the behemoth stacks would be one short of equaling the Titan stacks, giving the titans a free unit to shoot. The Dragons would blind the Behemoths and then use their breath attacks to the best of their advantage.
Final thoughts:
Each battle had a different strategy even though they were the same units. Sometimes they would frenzy frist, blind or Armageddon. Gang up on one unit right away, or 2 units (i noticed the units never attacked one-on-one. There would always be at least one unit exposed or free.
Some battles the AI would demonstrate some poor strategy. 3 times in a row I had Chaos Hydras beating Archangels. In these battles, the Archangels would reduce the CH to one stack, and then if that final stack was surrounded by the AA, the angels wouldn't attack but just block. so it was 3 stacks of 15 AA and one stack 15 for the CH (45 vs 15). The CH would just repeatedly attack without retaliation until the angels were reduced to 1 or 2 each, and then
claim victory. If they were so worried about the all around attack, the angels would split up and then return one-by-one until the CH was dead. So those battles I reset. Actually it was as if the AI learned from me: I manually played one battle as the AA going against the CH. When I got down to one stack of CH, I just attacked the CH, endured one hit on all units, and then the other 2 stacks i had attacked without retaliation. When I reset the battles, it was as if the AI saw what I did and then incorporated that strategy. I repeated the simulations 24x and the AI never repeated it's poor strategy again!
I'll do something similar for the non-upgraded units. Those poor ghost dragons, I hope they do better against the non-upgraded units!
|
|
Phoenix4ever
Legendary Hero
Heroes is love, Heroes is life
|
|
posted August 07, 2020 05:45 PM |
|
|
Interesting analysis, although I'm not sure how much these comparisons are worth. There are heroes, skills, morale, luck, spells, artifacts and strategy, that are usually also part of the equation.
Yeah Ghost Dragons suck, but at least Vampire Lords, Power Liches and Dread Knights are great.
|
|
bloodsucker
Legendary Hero
|
|
posted August 07, 2020 05:59 PM |
|
|
Yeah! One on one Ghost Dragons suck. But if you have a Necropolis, a Fortress and a Dungeon their numbers can beat Phoenix growth...
And they have that awful quality of being stronger the stronger the enemy is, that is common to Behemoths, cows and sea-dogs. Halve the HP of Azure Dragons is far from pointless.
As many other units (like Archdevils, for instance), they can be extremely useful against AI but a waste of money against a skillful human opponent.
____________
|
|
NimoStar
Responsible
Legendary Hero
Modding the Unmoddable
|
|
posted August 07, 2020 06:00 PM |
|
Edited by NimoStar at 18:03, 07 Aug 2020.
|
Equally size stacks is not a good gauge of creature strength, like 15 goblins vs 15 behemoths.
A better way of seeing how good a creature is is its resource efficiency and weekly growth.
So 4 weeks of behemoths vs 4 weeks of goblins,
or 20000 gold of behemoths vs 20000 gold of goblins.
Phoenix advantage is that they recruit more than other Lvl7 creatures, for example.
Also you tampered with the results of the Archangels vs Chaos Hydra... if you set up a method, it should be used always, good or bad result.
____________
Never changing = never improving
|
|
FirePaladin
Legendary Hero
DoR Modder
|
|
posted August 07, 2020 06:47 PM |
|
|
Ghost Dragons are support units, if you split them in stacks for the Aging effect. Or also siege units and such, shooter blockers, etc., since they aren't that valuable anyway.
____________
"Goblins use pistols because a shotgun recoil would kick them off the ground flying."
The Reckoning.... it's drawing near....
|
|
fanofheroes
Famous Hero
|
|
posted August 07, 2020 07:15 PM |
|
|
NimoStar said:
Equally size stacks is not a good gauge of creature strength, like 15 goblins vs 15 behemoths.
A better way of seeing how good a creature is is its resource efficiency and weekly growth.
So 4 weeks of behemoths vs 4 weeks of goblins,
or 20000 gold of behemoths vs 20000 gold of goblins.
Phoenix advantage is that they recruit more than other Lvl7 creatures, for example.
Also you tampered with the results of the Archangels vs Chaos Hydra... if you set up a method, it should be used always, good or bad result.
I wanted to compare monsters of the same level. So there's no crossover. But based on this algorithm, if I want to see Level 1 vs level 7, then I would select the number of lvl 1 units accumulated in 8 months (i made a typo and said 8 weeks in the main thread, it should be months, not weeks). I think that's the key point, after 8 months, with one town, all the dwellings are built, and earning 4000 per day, you'll be able to afford every number of monsters generated every week, making the cost/unit nulified.
In my opinion, cost/unit is a different variable in head-to-head matches The AI doesn't think about cost/unit when roaming monsters grow on the map 
Phoenixes growing 2x the rest of the units is a lot. Maybe because they are 7/8 on the ranking list. I'll do another trial run with them doubled because it never said the Phoenix special ability was to grow twice as fast in the unit description, it was just put in for the conflux town.
bloodsucker said:
Yeah! One on one Ghost Dragons suck. But if you have a Necropolis, a Fortress and a Dungeon their numbers can beat Phoenix growth...
And they have that awful quality of being stronger the stronger the enemy is, that is common to Behemoths, cows and sea-dogs. Halve the HP of Azure Dragons is far from pointless.
As many other units (like Archdevils, for instance), they can be extremely useful against AI but a waste of money against a skillful human opponent.
Vampires are my favourite unit so I know the feeling of making a well balanced army supporting each other to their strengths
Phoenix4ever said:
Interesting analysis, although I'm not sure how much these comparisons are worth. There are heroes, skills, morale, luck, spells, artifacts and strategy, that are usually also part of the equation.
Yeah Ghost Dragons suck, but at least Vampire Lords, Power Liches and Dread Knights are great.
Ofcourse player strategy, artifacts, hero levels, spells, army support plays a huge role. My inspiration was to figure out how the designers came up with the AI and fight values, and if they were dependent on how they do head-to-head. I don't think the AI/Fight value takes into effect if the unit is support or not.
|
|
Phoenix4ever
Legendary Hero
Heroes is love, Heroes is life
|
|
posted August 07, 2020 07:26 PM |
|
|
bloodsucker said:
Yeah! One on one Ghost Dragons suck. But if you have a Necropolis, a Fortress and a Dungeon their numbers can beat Phoenix growth...
Another reason why I dislike original Town Portal, but that's a bit off topic I suppose.
|
|
heymlich
Famous Hero
|
|
posted August 07, 2020 10:15 PM |
|
|
|
I once did a similar test, but with higher hero stats (20A/20D). Under these conditions Ancient Behemoth beats everything except Titans.
|
|
fanofheroes
Famous Hero
|
|
posted August 07, 2020 11:10 PM |
|
|
heymlich said:
I once did a similar test, but with higher hero stats (20A/20D). Under these conditions Ancient Behemoth beats everything except Titans.
I noticed too, that the relationships are not linear. as the stats go up, the units that favour might have the advantage. Unless if spell power went up more than the A/D rate. When I did the 1000 vs 1000, Armageddon only did 510 damage at SP10. So the most it took off was 2 or 1 unit/stack. That basically made damage spells insignificant for all units. I guess I could've calculated that if Armageddon killed ~14% of my 15 unit stacks, I could've calculated what SP I would need to kill off 14% of a 1000 stack. So to kill archangels proportionally, I would need an Armageddon that does 34,000 points of damage. Who knows what spell power that would be 
I tried to keep it consistent among all units, and I estimated that to have a proportional Might (A/D and number of units) and Magic (SP/K) balance, primary skills of 10 would be a good place to start. If I had spell power and knowledge up to 1000 (hypothetical), then the hero's would cast spells almost every round at maximum levels beyond normal. The higher the magic, the more it favours magic-immune monsters.
|
| |
|
|
