CAG 4.0 CAGBot - Game price prediction and deal evaluation

The place for CAG 4.0 feature discussion and bug reporting.

John

Administrator
Staff member
Feedback
10 (100%)
I'm posting a series of threads to go over new features in the CAG update, if you have any CAGBot suggestions or bugs to report this is the place for it.

What is CAGBot?
It's a safe bet that after a game is released, the price will gradually decrease over time. But how quickly the price drops can be affected by different criteria in varying amounts. For example, a Nintendo developed, Nintendo published game on Switch with a 97 Metacritic score is unlikely to drop in price anywhere as quickly as an Ubisoft PC game with a 62 Metacritic score. CAGBot is a system that takes a whole bunch of these criteria (developer, publisher, rating, initial price, age, time of year, etc.) and makes a series of predictions for future dates based on historical price tracker data.

Where is it used?
You can see CAGBot icons on the front page, on game pages, next to prices in the price tracker, and even dynamically in posts. A red icon indicates a bad deal, yellow a good deal, green a great deal.

CAGBot predictions are visible on game pages, both with a graph of predicted future prices, and a summary of target prices. (These targets are what CAGBot uses to determine if a price is recommended or not.)

Screenshot 2023-11-28 at 20.32.53.png Screenshot 2023-11-28 at 20.33.00.png

How can I use it in posts?
You can use the "Game Release" button in the post editor to search for a game and embed it in your post.
Screenshot 2023-11-28 at 20.16.45.png

This will embed a stylised link to the game page, like so:
The Witness XBO

But if you do this next to a price, a CAGBot evaluation icon will appear next to the link. For example, at $20, a bad deal:
The Witness XBO
$20

At $10, getting better:
The Witness XBO
$10

And finally, $5 is a great price:
The Witness XBO
$5

We'll be making some improvements to how this looks in the editor in a future update. (So you'll link as-is rather than just as raw bbcode.)

Future updates
With CAGBot, the predictions are only as good as the data we train it on – at the moment we have a limited dataset based on prices scanned shortly before we switched to new CAG. We'll be updating the model monthly with new data to tighten this up.

Confirmed issues/bugs
  • Have noticed an issue where new prices fail to update the price cache, meaning CAGBot sometimes makes a recommendation on a game price that is no longer current.
 
bread's done
Back
Top