‘Gold is for the mistress — silver for the maid —
Copper for the craftsman cunning at his trade.’
‘Good!’ said the Baron, sitting in his hall,
‘But Iron — Cold Iron — is master of them all.’
This post attempts to explain elements of the Crowfall crafting and gathering system using Iron bars as the primary example. (I normally play a confessor. As such, I often need to craft fiery gold bars. My primary gathering loop on Sorrow has three iron nodes in close proximity to two gold nodes. I require two gold for every one iron. As such, I typically have a surplus of iron which I used to do the following experiment).
An iron bar is constructed by combining three stacks of iron. Each stack consists of three pieces of iron. The final iron bar has a statistic known as “Attack Power”. In constructing your iron bar, your primary goal is typically to maximize the amount of attack power that the final iron bar will possess. The attack power will be determined by two key parameters
First, the quality of the iron bar that you create. When you craft a piece of gear, that item will have a quality level ranging from “Failure” à “Amazing Success”. In order, these are
· Moderate Success
· Good Success
· Great Success
· Amazing Success
In addition, when you construct your iron bar, you have the ability to spend experimentation points. The total number of experimentation points that you have is determined by your skill as a craftsman and – in the Big World setting – by consuming crafting potions. However, you can only spend experimentation points if an item has experimentation “slots”. When crafting metals bars, you have experimentation slots that can be used to impact both the quality of the resulting bar as well as the durability of the resulting item.
The number of experimentation slots is determined by the quality of the materials that you are working with. Grey ore – the lowest quality ore available in PVP worlds – gives four experimentation slots. (two each for Quality and Durability). Green ore – the best quality ore – gives eight experimentation slots (four each for Quality and Durability). White ore – the intermediate quality – seems to give six slots. Mixing white and grey ore randomly gives you either four slots or six slots (Using one stack of white ore and two stacks of grey ore seems to give six slots 50% of the time and four slots 50% of the time). I didn’t have any luck getting eight slots using a mixture of green and white ore, however, given how rare green ore is dropping these days, I didn’t carry out a significant number of trials.
Hopefully, this all makes sense. For our next step, I am going to try to describe the relationship
Attack Power = F(Success, Experimentation Points)
In doing so, I am going to follow worst practices and treat a categorical variable (Success level) as a numeric value. (I’ll code a Failure as a 0, a Success as a 1, …) The R^2 of the resulting models is essentially 1 which means that I’m capturing all the variance so I am pretty sure that I have things figured out properly. Even so, it rankles. Here’s a raw data I was playing with. (For the record, all of the bars were generated using three crafting potions)
Let’s start with the case where I was spending two experimentation points on product quality.
In this case, the equation governing the relationship between Attack Power and Success Level looks to be (approximately)
Product Quality = .0016 * (Success^2) + .0022*Success) + .4
(I ran a linear regression using a second order polynomial and constrained the intercept term to be .4) The R^2 of the resulting model is .993 which is pretty good, though I expect that there is some rounding order going on)
In the case where we have three points to spend on experimentation
Product Quality = .0039 * (Success^2) + .0033*Success) + .405
In the case where we are lucky enough to have four points to spend on experimentation
Product Quality = .0038 * (Success^2) + .0093*Success) + .41
Some of you may have noticed (maybe even cared) that I modeled this as three curves rather than as one surface. Given that we only have three points to define experimentation points, doing a surface fit didn’t seem as if it would add that much.
Why is any of this interest (especially since a lot of these numbers are likely to change come launch)? The main reason that I was doing this was to get a feel for how difficult it was to figure out what type functions that ACE is using.
From a crafter’s perspective, the key problem is determining how to transform a steady stream of raw materials into finished goods. In particular, guilds will need to understand the optimal ratio of Harvesters:Crafters:Fighter. In turn, this requires a nice linear programming model that we can use to transform raw materials into desirable finished goods.
During the early stages of the game, collecting data is going to be key. We’re going to need to understand what ratio different qualities of resources drop in and how we expect these numbers to change over time. In addition we need to know the success rates in crafting different qualities of goods.
Personally, I am trying to figure out what type of information that I need to observe during the harvest and crafting stages. Once the game goes live, I need to help make sure that my guildmates not only harvest the right resources and craft the right stuff, but also that they are collating the right information so we can generate accurate predictive models.
Edited by narsille, 25 December 2016 - 10:48 AM.