All Categories

Hydra fix: the right way

Search
Nov 8, 2024, 22:4411/08/24
11/03/24
3

Hydra fix: the right way

Exponential decay of points.

i

An abstract example

It's like a decreasing multiplier. E.g. you get less points with every next 100 turns or every next 100K damage you deal (that could be done like a reversed Scarab shield so that you actually deal less damage).

The stronger the team, the bigger the penalty.

Newbie players would be barely affected. A Player A dealing 10 times more damage than Player B could then be competed by 2-3 Players B (and not 10, like in the current or former implementations).

Adjustable, predictable.

This approach would automatically equalize any champion or mechanic - existing or added in the future. It wouldn't break any existing team in Hydra, just make them less OP. It wouldn't affect anything in locations other than Hydra.

Views
30
Comments
13
Comments
Nov 8, 2024, 23:3911/08/24
06/25/20
7024

They more-or-less already have this. It's essentially what the decreasing turn-counter on both the duration of devour and the turns before the next devour are for, on top of the increasing HP on devouring heads once the turn-counter drops to two.

The problem is they simply can't balance their game around the absurd mechanics they themselves introduce - and they likely can't nerf Trunda too much because they risk alienating their whales.

Nov 9, 2024, 12:0811/09/24
Nov 9, 2024, 12:08(edited)
12/16/21
2071

redditbot reply above ^. a combination of ignorant and ironic :D

ShortOnSkillzModerator
Nov 9, 2024, 13:1211/09/24
06/24/19
3583
MooredRat

redditbot reply above ^. a combination of ignorant and ironic :D

Cant believe you'd call Kram a bot.  Rude.

ShortOnSkillzModerator
Nov 9, 2024, 13:1311/09/24
06/24/19
3583

My only concern with something like this is you'd end up with really homogeneous scores and at the same time disincentivize people trying to improve their times.

Nov 9, 2024, 13:2911/09/24
12/16/21
2071
ShortOnSkillz

My only concern with something like this is you'd end up with really homogeneous scores and at the same time disincentivize people trying to improve their times.

I still say it woulda been better to have an absolute cap on damage, (like say 250M x difficulty modifier) and in the event of a tie for Clash, the win goes to whichever clan hit it in fewer total turns

Nov 9, 2024, 15:2511/09/24
06/25/20
7024

I stand by my suggestion of capping turncount at 500 instead of 1500.

Nov 9, 2024, 15:4811/09/24
12/18/19
389

We need 6 turn limit, you buff + debuff + give 1-2 smacks, every one of your char gets to go once. easy and classy

harleQuinnModerator
Nov 9, 2024, 19:3011/09/24
02/24/19
8091
kramaswamy.kr

I stand by my suggestion of capping turncount at 500 instead of 1500.

Less turns, reduced scores for max rewards sounds like a very time saving idea.

ShortOnSkillzModerator
Nov 9, 2024, 21:5711/09/24
06/24/19
3583
kramaswamy.kr

I stand by my suggestion of capping turncount at 500 instead of 1500.

Only if they move the scores for PR chest down as well.

Nov 9, 2024, 22:3611/09/24
06/25/20
7024
ShortOnSkillz

Only if they move the scores for PR chest down as well.

I mean, yeah that's a given lol.

ShortOnSkillzModerator
Nov 9, 2024, 23:2411/09/24
06/24/19
3583
sharkium

We need 6 turn limit, you buff + debuff + give 1-2 smacks, every one of your char gets to go once. easy and classy

1000x multipliers on scores.  Easy 

Nov 27, 2024, 19:4811/27/24
01/20/20
1

All I know is at Nov 9 2024 my team knocked off heads at 2.5M and did 80-90 rounds for 7.5M now Nov27 2024 that same team gets 1.5M and 50 round and noticable less damage per hit 

Nov 27, 2024, 21:5211/27/24
12/16/21
2071
MacAries

All I know is at Nov 9 2024 my team knocked off heads at 2.5M and did 80-90 rounds for 7.5M now Nov27 2024 that same team gets 1.5M and 50 round and noticable less damage per hit 

It's also a different rotation; the heads' affinity and starting line-up changes.

A nuker that smacked the wrath and suffering heads last week might get weak hits this week because it's a different affinity.