Reviving this because the problem danth identified hasn't gone away.
I'm a collector who got tired of manually refreshing eBay across multiple countries for 5 sets, so I ended up building something (https://scoutloot.com). Relevant to this thread because I took the opposite approach to BrickEconomy, instead of predicting what a set will be worth (which as Mylenium correctly pointed out requires sample sizes that don't exist for expensive vintage sets), I focus on what's actually selling right now across 12 eBay marketplaces + BrickOwl.
The reference benchmark depends on what we're measuring. For the "X% below market" you see on deal cards, we use BrickLink's current stock price (specifically the 2nd cheapest listing, not the inflated average). For the Deal Quality Score that gates whether a notification is actually worth sending, we use BrickLink completed sales weighted by recency, plus our own recorded market data from eBay/BrickOwl over the past 7-90 days. Two different lenses: "is this cheaper than what's available right now?" and "is this genuinely a good deal based on what people have actually paid?"
On the regional pricing issue that's one of the things that frustrated me most. A "deal" on eBay.com is meaningless to a European buyer once you add shipping and duties. We calculate total landed cost per country, which sounds obvious but apparently most tools skip it.
We do have an investment analysis feature (Crystal Ball) that scores sets on 11 signals, peer set appreciation, BrickLink supply trends, demand scores, price momentum, etc. but every signal is shown individually so you can see the reasoning and disagree with it. Transparent data > black-box predictions.
BrickLink Price Guide is still the gold standard for actual sale data on vintage sets. For current/retiring sets, automated scanning across marketplaces fills a different gap.