After looking at your Reddit post about the idea:
1: Judging by the fact that you're trying to get the pricing APIs yourself, the web dev you consulted is screwing you with the $31,000 quote.
This is something an experienced developer with Magic experience could prototype for you in a month. Hell, I'm not a particularly experienced developer, and I'm confident I could have a working concept in two months.
You mentioned before that you can't program and don't want to learn, but at the very least find a partner who is willing to help you with the programming in exchange for equity in the company. Sinking $31k is insane.
2: You need to be able to account for price manipulation, which means the easily accessible pricing API (tcgplayer) is probably not going to cut it, and StarCityGames is notoriously difficult to scrape. They've sent me enough angry emails that I've given up on getting their pricing data.
TCGplayer low and mid prices are very easy to game on low-stock low-visibility cards. If you're offering a guaranteed copy of, say, the SDCC 2013 black-on-black promos for current TCGPlayer mid prices, it wouldn't be too difficult for someone to purchase every available copy for around $2000 and relist them (with no intention of shipping) in the area of $20 . They could continue listing and selling nonexistent copies until your pricing was impacted, then have a group of people each reserve four copies at twenty bucks each on your dime.
Now they've bought out an extremely rare card, and you're forced to go looking for copies you agreed to sell for $20 + your booking fee.
Pricing APIs also have a lot of errors. Tcgplayer doesn't update as often as it should sometimes; for nearly a week, Shahrazad was listed as being ~$100 mid despite the cheapest copy being in the area of $180.
If you can somehow scrape SCG without them repeatedly threatening to sue, you still run into the issue of out-of-stock prices not being updated. I don't see a feasible way to run this without a single-vendor pricing metric unless you're manually verifying every "investment", but that means your users will only be able to book cards that are in-stock at StarCityGames or ChannelFireball.
3: This is the most difficult issue: userbase.
I imagine you envisioned a platform where the small percentage of successful predictions would be payed for by the vast majority of terrible speculators.
The issue is getting the terrible speculators to use your service in the first place, and proving that you're not going to screw them over by keeping the booking fee.
Your Reddit post was not well received, and after pucatrade caused point hyperinflation by selling their imaginary currency for cash, I really don't blame people for being distrustful of MTG startups. I have absolutely no idea how you're going to solve this, short of making introductory transaction fees practically free.
4: If you want to abandon this idea, repay the loan, and find another MTG startup to work on instead, PM me. I have a few great ones.