I. Collection and Preparation
Our project utilizes comprehensive data extracted from Stardew Valley, primarily sourced from the game’s official Wiki page, which is widely regarded as the most accurate and community verified representation of in game mechanics. This source was chosen due to its consistent updates, trust within the player community, and our own experiences as players who rely on it during the game. However, it also required the data to be compiled manually. While tedious, it gave us a lot of control in our process.
We focused on gathering data for the three core economic systems within the game: crops, fish, and animal products. We applied consistent naming conventions and structured the data using tidy or long-format principles to support flexible analysis.
For crops, we gathered detailed information related to planting, harvesting, and profit margins, including:
- Crop name
- Crop type ( fruit, tree, vegetable, forageable, flower, other)
- Season availability (Spring, Summer, Fall, Winter)
- Seed price
- Days to initial harvest
- Days to regrow (for multi-yield crops)
- Yield per harvest
- Maximum number of harvests per season
- Harvest frequency under different fertilizers: Speed-Gro, Deluxe Speed-Gro, Hyper Speed-Gro
- Impact of the Agriculturist profession on fertilizer-enhanced growth rates
- Base sell price by quality: Normal, Silver, Gold, Iridium
- Tiller profession price multipliers by quality
- Keg product output (Juice, Wine) and:
- Base and Artisan profession prices by quality
- Preserves Jar output (Pickle, Jelly) and:
- Base and Artisan profession prices by quality
- Dehydrator product (Dried Fruit, Dried Mushroom) and:
- Base and Artisan profession prices by quality
We included effects of relevant professions, which are selected by players upon leveling up in specific skills. Professions significantly alter item values and growth rates:
- Tiller: Crops worth 10% more
- Artisan: Artisan goods worth 40% more
- Agriculturist: Crops grow 10% faster
For fish, we collected data relevant to seasonality, pricing, and processing:
- Fish name
- Fish type (ocean, lake, pond, river, legendary)
- Seasonal availability (Spring, Summer, Fall, Winter, or Special)
- Base quality sell prices: Normal, Silver, Gold, Iridium
- Smoked product prices, with and without Artisan profession
- Sell prices under Fisher and Angler professions
- Fish Roe value (raw and aged), including Artisan profession boosts
The Fisher and Angler professions apply multipliers of 25% and 50% respectively to raw fish values. The Artisan profession adds value to processed fish products, such as smoked or aged goods.
For animal-based goods, we recorded information on raw and processed profits, along with production context:
- Product name
- Animal that produces it
- Base sell prices by quality
- Rancher profession multipliers (+20% value)
- Artisan profession multipliers (+40% for artisan goods)
- Processing time
- Machine used for processing (Cheese Press, Oil Maker, Loom)
- Required building type (Coop, Barn, Deluxe Barn)
II. Cleaning and Processing
While compiling our data, we faced several formatting issues. Our original datasets included mixed capitalization and spacing that caused problems when trying to link data with file paths or reference values dynamically. For example, image assets for items had to be referenced by name in the browser, and names like "Sweet Gem Berry" would not match correctly unless written in a file-friendly format like "sweet_gem_berry".
To address this, all item names were converted to lowercase, spaces were removed or replaced with underscores, and column names were standardized using snake_case. Additionally, all numeric values, especially prices and growth times, were manually converted from strings to proper numerical types. This was necessary to enable quantitative operations and comparisons during exploratory data analysis (EDA), especially when using tools like Tableau, which would otherwise treat the values as categorical text.
We made deliberate decisions about which data to include and exclude. While our datasets are rich in numerical detail, we chose to omit broader global multipliers, like general profession-based price boosts, from most visualizations to avoid distorting trends across categories. For example, although professions like Fisher, Angler, and Rancher affect final profit margins, they apply uniformly and do not change item relationships or patterns of seasonality, processing efficiency, or base quality profitability. Instead, we focused on underlying factors that players could experiment with directly: quality tiers, processing choices, season timing, and growth enhancements.
This decision helped us maintain the clarity and usability of the visualizations, allowing users to evaluate core economic trade-offs without needing to toggle multiple modifiers. That said, we still preserved all the profession-based pricing in our datasets to allow for more advanced comparisons if desired, and future iterations of the visualization might include display of these modifiers.
III. Integration
Initially, we planned to work with CSV files directly in the visualization pipelines. However, we encountered technical limitations when integrating CSVs into the website using Javascript. Since CSVs are flat tables, they lack nested structure needed to easily associate relative properties (matching a fish to its season and to its processing value). To solve this, we pivoted to using JSON for all frontend integration which allows us to easily query and manipulate the data within our Javascript functions.
We converted our clean CSVs into JSON arrays using, which allowed us to access data via key-value pairs, making our JavaScript logic far more efficient and readable. This structure was especially important for: dynamically filtering items based on season or category, loading related images based on name, and calculating profit per day or per season using conditional logic. When a user filters by variable (choosing crops available in Spring), our script generates a new array containing only the relevant data, which is then passed into the visualization and used to update other display elements.
In the end we developed ten spreadsheets, three .CSV files, three .JSON files, and 9+ dynamically generated arrays.