Tired of spending hours researching, so I did a thing! PT 2
Posted here last month about the scraper I was building so I could stop watching YouTubers and chasing deals. The pitch was simple: let the computer do the homework. Feel free to check it out: atlasdrake.ai/tcgboostoracle/
A lot of you weighed in, thanks. Some of the feature requests went straight into the roadmap.
Since then I shipped three more games:
- Pokemon Japan (storefront + forecasts)
- One Piece (forecast + storefront)
- Yu-Gi-Oh (forecast + storefront)
Six things I learned along the way:
1. The pain isn't finding deals. It's the legwork before finding deals. Cross-referencing TCGplayer market price vs eBay sold vs PriceCharting vs Discord opinion is the real time sink. Removing the legwork is the value. The "deals" almost surface themselves once the legwork is done.
2. Forecasts get the clicks. Buy targets get the use. Turns out people don't want predictions about Q3 2027. They want to know if $54.99 is a fair price for a Booster Bundle right now. So I made the buy target the headline feature and the forecasts a deeper layer underneath.
3. Yu-Gi-Oh broke my chase-rarity model. The Pokemon ladder (Common to Special Illustration Rare) doesn't translate. Starlight Rare blew past anything Pokemon has, and Quarter Century Secret Rare landed at a tier I didn't have weights for. Built a new 0.5 to 3.5 table just for YGO. Pokemon-trained intuition is wrong here.
4. Pokemon Japan was the hardest crawl. Set codes, JP-only product types like "Premium Trainer Box" and "Special Card Set," regional variants. I spent more time on JustTCG name normalization than I did on the actual crawler logic.
5. The feature people don't ask for but use the most: storefront comparison. Once partner stores landed with promo codes and shipping/tax info baked in, retention jumped. People care about what they actually pay, not the listed price. Surfacing the real total before checkout is more useful than I expected. I crawl hundred of storefronts daily, surfacing best deals + analytics all on same page.
6. The price hunt extends beyond TCGplayer. Half the TCG investor workflow happens on Facebook Marketplace, eBay sold listings, and Google Shopping. Added one-click buttons for each, pre-filtered to the calculated buy target. Buy target on FB Marketplace means "show me only listings at or below fair value." Buy target on eBay Sold means "show me what flippers actually paid." Cross-platform price reasoning with the math already done ended up being one of the stickier features.
One example from this week:
TCGBoost Oracle flagged that Destined Rivals Booster Bundles were rapidly climbing in market price while some LGS listings still hadn’t updated yet.
I grabbed 5 bundles at $72 each.
Current market price is already around $85 👀
The biggest value honestly isn’t just “finding deals.” It’s identifying lag between market movement and storefront pricing updates before inventory disappears.
Instead of checking:
- TCGPlayer
- eBay sold
- dozens of storefronts
- Discord alerts
- YouTube market videos
…the system surfaces:
📈 Momentum shifts
💸 Under market listings
📊 Historical pricing
🛒 Storefront lag opportunities
Still early, but it’s already saving me a ton of time.
Feel free to check it out: atlasdrake.ai/tcgboostoracle/
Magic and Lorcana are next on the roadmap. Curious what other signals you'd want me to build. I've been going by gut and what folks ask for, but happy to take direction.