Menu
What's New
By:
Filters

Using AI to solve the Hunting Loot algorithm

The Great Adventurer

The Great Adventurer

Alpha
Joined
Jan 9, 2023
Posts
526
Tweak this code to help solve the algorithm:

Python: 
import random
import math
import time

weapon_efficiency = 0.85
looter_profession_level = 50
dpp_range = (0, 7.77)
weapon_min_damage = 50
weapon_max_damage = 100
critical_hit_ability = 0.11
weapon_critical_hit_ability = 0.02
buff_critical_hit_ability = 0.09
buff_additional_critical_damage = 3.0
creature_health = 100
creature_loot_probability = [0.4, 0.3, 0.2, 0.1]
armor_damage_cost = 10
healing_cost = 5
weapon_efficiency_loot_return = 0.07
looter_profession_loot_return = 0.07
minimum_loot_return = 0.86
wave_amplitude = 0.15
wave_period = 3600

current_time = time.time()
wave_value = wave_amplitude * (1 + math.sin(2 * math.pi * current_time / wave_period)) / 2
adjusted_loot_probability = [
    prob - wave_value * (len(creature_loot_probability) - idx - 1) / (len(creature_loot_probability) - 1)
    for idx, prob in enumerate(creature_loot_probability)]

damage_done = random.uniform(weapon_min_damage, weapon_max_damage) * weapon_efficiency
is_critical_hit = random.random() < (weapon_critical_hit_ability + buff_critical_hit_ability)

if is_critical_hit:
    damage_done *= (1 + buff_additional_critical_damage)

overkill_cost = max(0, damage_done - creature_health)
total_cost_to_kill = overkill_cost + armor_damage_cost + healing_cost
adjusted_dpp = (damage_done - overkill_cost) / total_cost_to_kill
total_loot_return_percentage = (
    minimum_loot_return +
    weapon_efficiency_loot_return * weapon_efficiency +
    looter_profession_loot_return * looter_profession_level / 100
)
loot_value = total_cost_to_kill * total_loot_return_percentage
loot_probability_distribution = [prob * adjusted_dpp for prob in adjusted_loot_probability]

random_number = random.random()
loot_tier = 0
cumulative_probability = 0

for idx, probability in enumerate(loot_probability_distribution):
    cumulative_probability += probability
    if random_number < cumulative_probability:
        loot_tier = idx
        break

loot_tiers = [
    [('Common Item 1', 1), ('Common Item 2', 1)],
    [('Uncommon Item 1', 1), ('Uncommon Item 2', 1)],
    [('Rare Item 1', 1), ('Rare Item 2', 1)],
    [('Very Rare Item 1', 1), ('Very Rare Item 2', 1)],
]

loot_items = []
for item, quantity in loot_tiers[loot_tier]:
    loot_items.append((item, quantity * loot_value))

print("Loot items:", loot_items)

Mr. GPT says:
Yes, you can think of the loot system in Entropia Universe as being similar to a slot machine, but with variables that the user can manipulate to improve their chances. By increasing their looter profession levels, using higher efficient weapons for better loot value returned, and optimizing their Damage Per Pec (DPP) for loot composition, players can actively influence the outcomes to some extent.

This dynamic nature of the loot system makes it more engaging and rewarding for players, as their efforts and investments in improving their skills, equipment, and strategies can lead to better loot over time. In this sense, the loot system in Entropia Universe provides a more interactive and skill-based experience compared to traditional slot machines, which rely solely on chance.
Click to expand...
 
some close to real alghorytm )
 
to analyze:
~44% of loot its not multiplified returns (return < cost to kill)
all other formed from multiples - small or bigger ones
 
You realize that the values put in are just from other people on the internet right? ChatGPT can't hack Mindarks servers or some shit.

Unless you tell it to site it's sources, and it says "From Mindark code monkeys" This mean even less than Rattspatooties Mining shit.
 
Angela Draniie Cloud said:
You realize that the values put in are just from other people on the internet right? ChatGPT can't hack Mindarks servers or some shit.

Unless you tell it to site it's sources, and it says "From Mindark code monkeys" This mean even less than Rattspatooties Mining shit.
Click to expand...
It's not meant to pull the actual algorithm. It's an exercise in using it to produce something similar using what the community believes to be the variables which contribute to it. If anything, it's simply to help better understand what it could be.

The variables in the Hunting Loot algorithm?

What are all the variables you can think of which are present in the Hunting Loot algorithm? Damager Per Pec Efficiency Looter Profession Level Timing (waves?) Mob (?)
www.planetcalypsoforum.com www.planetcalypsoforum.com

It took many prompts to get it to resemble what it does above. Feel free to use it in your own prompts as a starting point or to experiment yourself.
 
The Great Adventurer said:
It's not meant to pull the actual algorithm. It's an exercise in using it to produce something similar using what the community believes to be the variables which contribute to it. If anything, it's simply to help better understand what it could be.

The variables in the Hunting Loot algorithm?

What are all the variables you can think of which are present in the Hunting Loot algorithm? Damager Per Pec Efficiency Looter Profession Level Timing (waves?) Mob (?)
www.planetcalypsoforum.com www.planetcalypsoforum.com

It took many prompts to get it to resemble what it does above. Feel free to use it in your own prompts as a starting point or to experiment yourself.
Click to expand...
But without sources, there's some people in the community who are stupid as shit.

And then it could also be taking loot 1.0 into account, could randomly be grabbing mining or crafting info as well.
 
It would likely be able to provide a decently working algorithm if you supplied it with data gathered with different parameters as well as observable data instead.

You could then ask it to simulate with other parameters, for which you also have data representation and check accuracy.

That could be an interesting exercise.
 
R4tt3xx said:
Please don't ask Mr GPT3 to do maths... It has the memory of a goldfish.
Click to expand...
It was actually done with GPT-4 while I subscribed for a month to test out the difference in capabilities between that and GPT-3.5. You can always recenter on the task at hand if it starts to hallucinate or forget what you were originally talking about if you know how and what to prompt it with.
 
Unless you have mindarks code you questions to the chatgpt will be what is already know, Eff = tt return, dpp dmg per pec = what you loot.
 
ProActive Mango said:
Unless you have mindarks code you questions to the chatgpt will be what is already know, Eff = tt return, dpp dmg per pec = what you loot.
Click to expand...
The AI has been trained on tons of code, which includes that of similar systems, like slot machines. Adding in what the communities knows, tweaking it, running simulations then comparing to what you and others have experienced can help produce a close match to what is the actual algorithm.
 
The Great Adventurer said:
Adding in what the communities knows, tweaking it, running simulations then comparing to what you and others have experienced can help produce a close match to what is the actual algorithm.
Click to expand...

The community knows nothing. We make assumptions based on current best data fit. Using what the community thinks as input is a bad idea, for obvious reasons.

You could produce a close match to the actual algorithm, but you wouldn't know it is a close match.

You would more likely produce an algorithm that is able to produce results that are close matches to your/communities beliefs of what is reasonable results. Something that reliably produces results that are within the bounds of your beliefs does not necessarily mean it is "close to" the "real algorithm". It may just as well mean that the bounds of your belief is off.

This is a fun exercise, but that is all it is. A toy
 
i love the smell of Bullshit in the morning...
smells like ......Victory
 
  • Haha
Reactions: SSN
Ferial said:
The community knows nothing. We make assumptions based on current best data fit. Using what the community thinks as input is a bad idea, for obvious reasons.

You could produce a close match to the actual algorithm, but you wouldn't know it is a close match.

You would more likely produce an algorithm that is able to produce results that are close matches to your/communities beliefs of what is reasonable results. Something that reliably produces results that are within the bounds of your beliefs does not necessarily mean it is "close to" the "real algorithm". It may just as well mean that the bounds of your belief is off.

This is a fun exercise, but that is all it is. A toy
Click to expand...

This logic is flawed and goes against the scientific method. There would not be computers, let alone Entropia had those of the past not theorized and tested outcomes.

It's actually so flawed one has to wonder if you're a MindArk employee...
 
ChatGPT
This dynamic nature of the loot system makes it more engaging and rewarding for players, as their efforts and investments in improving their skills, equipment, and strategies can lead to better loot over time. In this sense, the loot system in Entropia Universe provides a more interactive and skill-based experience compared to traditional slot machines, which rely solely on chance.

Arbitrary PCF User
a Slot machine should be a Slot machine

Other Arbitrary PCF User
Random loot no weaves randomize process that way it will be more intresting !

:eek: ChatGPT is already better at making sound game design arguments and has a better growth mentality than maybe a quarter of Entropians! :eek:
 
Did you get ChatGPT to write that code? That's positively amazing to me that this is a possibility today.
 
The Great Adventurer said:
This logic is flawed and goes against the scientific method. There would not be computers, let alone Entropia had those of the past not theorized and tested outcomes.

It's actually so flawed one has to wonder if you're a MindArk employee...
Click to expand...

Avoiding bias is a large part of the scientific method. The scientific method also makes it very clear that "we know nothing" and that the current standpoint is what current data suggests.

Everything I posted there is very much in line with the scientific method.

I guess I could translate my previous post for you quickly, since you seemed to not understand it:

Assuming that the current best understanding of something is accurate is dangerous, as it can lead you down a path that produces erroneous results that you assume are correct (if you are not careful and include this as a potential risk).

Whatever result you end up getting is not verifiable as you have no data that is positively an accurate representation of what you are testing for.

What you think is right based on current best understanding may be wrong and as a result you may come to conclusions that end up being wrong.

With all of the above in mind what you are doing, at best, is toying.

I hope this makes more sense.

Have fun.
 
Ferial said:
It would likely be able to provide a decently working algorithm if you supplied it with data gathered with different parameters as well as observable data instead.

You could then ask it to simulate with other parameters, for which you also have data representation and check accuracy.

That could be an interesting exercise.
Click to expand...

Ferial said:
The community knows nothing. We make assumptions based on current best data fit. Using what the community thinks as input is a bad idea, for obvious reasons.

You could produce a close match to the actual algorithm, but you wouldn't know it is a close match.

You would more likely produce an algorithm that is able to produce results that are close matches to your/communities beliefs of what is reasonable results. Something that reliably produces results that are within the bounds of your beliefs does not necessarily mean it is "close to" the "real algorithm". It may just as well mean that the bounds of your belief is off.

This is a fun exercise, but that is all it is. A toy
Click to expand...
 
I don't see your point. There is nothing contradictory in those posts.

I am saying that using data to build a model is more likely to produce a more robust solution than going by the assumption that current understanding is correct.

You said this was an exercise, and as an exercise it is fun and interesting. When you started defending the validity of the results I felt the need to step in and point out potential risk factors that you seemed to have missed.
 
You must log in or register to reply here.
Back
Top