About
This is a site to browse and share Owl of Minerva search results from
MapleLegends. Screenshots are cropped and
machine-read to create a repository of owl search results. Please see the
forum post or contact geospiza to leave
feedback.
The repository history allows us to inspect the price and volume of items sold by merchants in the Free Market. Click on one of the items in the items table or visit the charting page to see more plots this this.
How is the data in table populated?
Data enters the table from the upload page. Anyone can upload in-game screenshots. The data for the table is then generated once an hour through an automated system. See the forum post for technical details.
The most recent upload for each item is chosen for summarization. The data is cleaned to remove entries that might skew the results such as auction auto-win (A/W) prices or errors from converting screenshots to text. This cleaning process removes prices that are 3 MAD from the median, and entries with corrupted values. Then, statistics like a five-number summary are computed using the cleaned data.
How do I read the table?
Hover over each column to see a description. The descriptions are also found here:
| Name | Description |
|---|---|
| Search Item | The name of the item that was searched using the Owl of Minerva. |
| Updated | The date of the owl search upload (YYYY-MM-DD). |
| Uploads | The number of distinct owl uploads. |
| Listed | The number of search results for the most recent owl search for this item. |
| Bundle | The sum of all item bundles. This field may be unreliable due to overlapping text from the price. |
| min | The minimum price in the cleaned search results. |
| p25 | The 25th percentile AKA the 1st quartile. This is the price that is greater than 25% of prices. |
| p50 | The 50th percentile AKA the median. This is the price that middle value if all prices are sorted. |
| avg | The average price of the cleaned search results. |
| std | The standard deviation of the cleaned search results. |
| Valid (%) | The number of valid rows after cleaning the search results. |
Can I trust this data?
For the most part, yes. There is validation and cleaning to ensure the quality of the data is acceptable for general use.
You may notice typos and unusual values when looking at the listing page. You
can get a good sense of market prices by sticking to to reliable fields like the
search result count (Listed), the 1st quartile (p25), or the median (p50).
The sum of bundles (Bundle) is generally unreliable due to an issue with text
overflow from the price column into the bundle column in the owl interface. For
example, if you look at the count history of Brown Work Gloves in the charting
page, you’ll notice that sum_bundle is greater than the search_results! This
is because gloves are often sold at prices greater than 10 billion mesos. The
minimum suffers similar problems due to occasional errors reading the
screenshot.
There are also issues that lie with the market data itself. Some items don’t have very many results, or suffer from a bias of not being sold. Another is the inability to see the stats on an item, which makes weapon and armor data less valuable for pricing. If there’s any doubt, just owl it.
How do I actually price my items?
Use the statistics to guide your pricing. For scrolls, I usually set my price to the 25th percentile or lower depending on my patience. This value tends to reliable in spite of wild variations in prices and errors when reading screenshots, as long as there are more than a few items on the market.
References
The website is open-source on GitHub at geospiza-fortis/owlrepo-client. The server source code is currently not public, but please reach out if you have interest in contributing outside of uploading owl searches.
The datasets for uploads and presentation are public via BigQuery at
owlrepo.owlrepo. See this starter
notebook
for more details.