MIR4 character NFT market analytics — source: XDRACO BY PATRIOS EU
🔑 Log in to see this — free for now
666 character(s) found
Try widening the price or power score range, or clear all filters to see the full catalog.
Within each class, a quality score (0-100) is calculated combining 10 metrics — power score, potential, equipment quality, special/unique equipment (magic stones, spectrumites, sets like Majestic/Ragnos, named weapons), mythic item count, tradeable equipment resale value, spirits, skills, codex, and training —, each expressed as a percentile against the rest of the characters in that same class. Each metric's weight isn't hand-picked: it comes from a statistical model (regression) trained against real sale prices (see the caveats below). With the score calculated, characters in the class are then grouped into 4 quartiles (from lowest to highest quality) and a reference price is calculated per quartile. Each character is compared against its own quartile's reference:
Tradeable resale value is calculated separately from overall equipment quality: some buyers look for characters specifically to strip their tradeable items and sell them separately (replacing the equipment before reselling the character) — a different business from "I want a powerful character", which weighs in as its own factor instead of mixing with the set's power. Like every other metric, its weight comes from the same statistical model trained against real sale prices, not a fixed number.
The reference price prioritizes real sales: when there are enough characters sold recently (last 60 days) in that quality quartile, the median of what people actually paid is used — not what other sellers are asking now, which can be inflated with nobody buying at it. If there aren't enough recent sales in that quartile (small classes or little activity), it falls back to the median asking price of other listings, as before.
Some classes have "key skills" configured (abilities that in practice determine whether someone buys a character or not, e.g. a healing skill at level 10 instead of 9) — if the character doesn't meet any of them at the target level, the score can never show as "Good price", no matter how well the rest of the metrics turn out. Not every class has this configured yet.
🔮 Expected price (new): besides the label, a machine learning model (trained on 1,442 real sales) estimates a concrete WEMIX range for each character. Tested against known sales, the typical error is 142.42 WEMIX and the predicted range contains the real price 49% of the time — this is still being refined, it complements the value score rather than replacing it, and its accuracy improves over time as more sales accumulate.
The blue line is the reference price for each quality quartile (real sale or asking price, whichever is available). A character below it shows as Good price; above it, High price; close to it, Fair price. Recalculated every hour along with the scraper.
⚠️ Limitations to keep in mind:
Use it as one more reference, not the only reason to decide a purchase.
We use Google Analytics cookies to understand how the site is used. You can accept or reject them — your choice is saved in this browser.