Talk:Sea Serpent and Haspid

Haspids are broken as hell in large numbers because when they take casualties, they get buffed. The formula looks complicated but it's not, just takes some time to crunch. I've did 2 cases as frankly it's depressing to see how much stronger Haspids become when they lose numbers.

2 starting Haspids: when 1 Haspid dies (half of the starting number) the remaining Haspid gets sqrt ((2x 300 + 300 / 1x 300 + 300) - 1) x 100% = sqrt ((900 / 600) - 1) x 100% = sqrt 0,5 x 100% = 0,7 x 100% = +70% Attack. So 49 because the game rounds 49,503 down. This is 19 more than Archangels have Defense. Gives Haspids 95% damage bonus against the defensively strongest town troop. Compare to other crazy strong unit, Ancient Behemoth. 49 Attack vs 19 Defense is 30 points of Attack more than Behemoths' Defense so 30x5=150% damage bonus. 75-137,5, average damage 106,25. Behemoths deal 52-87 to Haspids, average 69,5 as their Defense is 20/5 = 4. Something's not quite right, eh?

1000 starting Haspids: 500 Haspids die. According to formula they get sqrt ((1001 x 300 / 501 x 300) - 1) x 100%, so let's say sqrt(2-1) x 100%, so 1x 100%... +100% Attack. So 58 Attack. Let this sink in: 58 Attack. 28 more than Archangels' Defense so 140% damage bonus. Or 39 Attack over Defense so +195% damage bonus vs Ancient Behemoths. This is pretty much TRIPLE their normal damage that wouldn't be low anyway as Behemoths are squishy. 90-165, average 127,5. Remember that Behemoths' damage to Haspids is 52-87, avg 69,5. The horror... The horror!

Looks like Haspids are the new alpha predator in Heroes 3 ecosystem. --Fafner (talk) 17:41, 9 December 2020 (UTC)

Cool, but the formula increases the Damage parameter, not the Attack. And there is "+" in the denominator, not "*", so yes, Haspids are pretty broken in theory. But in practice you usually don't want to loose such powerful creatures, and that means that you will not receive that great damage bonus.

When compared with Ancient Behemoths, do not forget that Behemoths have superior scaling against enemy heroes, especially in late game, thanks to their ability, which lets them steadily deal heavy damage.

Yet in 1v1 battle, ceteris paribus, they are trully the strongest lvl 7 creature in the game. -- Mortarial (talk) 21:23, 9 December 2020 (UTC)

That's right ^. --FirePaladin2 (talk) 21:26, 9 December 2020 (UTC)

I stand corrected. Still, the crunching was basically correct as I've just shortened current number of Haspids x 300 + 1 x 300 to (current number of Haspids + 1) x 300. Let's see what happens when we apply the results to Damage.

2 starting Haspids, 1 dies, 1 remaining gets +70% Damage, 59,5-93,5, avg 76,5. Still crazy. 89,25-140,25, avg 114,75 vs Ancient Behemoth thanks to Haspid's Attack advantage over Ancient Behemoth's defense.

1000 starting Haspids, 500 die, 500 remaining get +100% Damage, 60-110, avg 85. I believe this is the scene in which Unreal announcer yells "Holy SHIT!": half of the stack deals as much damage as full stack. As I like hurting Ancient Behemoths, it would be 90-165 Damage, avg 127,5 because 10 Attack over Defense so 50% extra damage. It's painful to look at. Gotta get my hands on HotA and do some live testing, my work laptop can't run it for some reason.

Now let's see what happens to Archangels with their grand 30 Defense: 1 Haspid (2 in stack initially) does 1,7x its normal damage that gets -2,5% penalty because 29 Att < 30 Def. 56,525-88,825, avg 72,675, close to 75 that Archangels will do because 50 base x 150% thanks to 30 Att > 20 Def. And Haspids have 6/5 HP but deal almost the same damage. Crazy, this.

500 Haspids of 1000 in stack: Damage gets simply doubled so 60-110, avg 87,5, 5% penalty = 57-104,5 avg 80,75. More HP, more damage. "Not broken! Is feature!"

Yes, Behemoths scale well, but they are always double-edged sword - they have bad Defense so they also take heavy damage and are general damage magnet because nobody sane will ignore them, they're too dangerous to any army, and as they take casualties their damage goes down, while Haspids losing numbers doesn't really diminish their output as Revenge partially makes up for losses. These calculations are all based on "clean lab" conditions - no att/def bonus, no morale, no luck, no spells, no terrain advantages, etc. to get some general idea which unit would win against which. Oh, and this was just Revenge crunching, the real kicker is Poison as if it procs turn 1, the opponent is just screwed.

In the future will try to double-check details of stuff I don't know by heart. --Fafner (talk) 10:14, 10 December 2020 (UTC)