Who publishes Sports Card Radar
This site is published by Vast Flow, LLP, a small software studio, and it is written and maintained by Merey Tleugazin, who started it after an evening spent trying to price a shoebox of cards and finding that the simplest question in the hobby — what is this worth right now — still did not have a simple answer.
The problem behind all of it
Every collector knows the ritual. You hold a card, type its year, set, player and number into a search bar, scroll past asking prices nobody ever pays, squint at blurry slab photographs, and ten minutes later you have a guess. Multiply that by a binder and pricing a collection becomes a weekend project. Multiply it by a storage unit and it never happens at all.
The data to answer the question properly already existed: public sold listings, auction records, grading population reports, the tolerance tables the grading companies print on their own websites. It was scattered, unstructured and slow to read, and nobody had connected it to the fastest sensor every collector already carries. That is the gap the app was built for, and the site grew out of the same observation applied to the questions a camera cannot answer — whether a submission is worth the fee, what a marketplace actually takes, what “1 in 24” means when you have opened twenty-four packs.
Who makes this, and what that is worth
Vast Flow, LLP is a software studio. It built a card-scanning app, which means the working day is spent inside sold-comps data, grading scales, population reports and the gap between what a card is listed at and what it actually cleared for. That is a real and specific kind of familiarity, and it is the honest basis for everything published here.
It is not the same thing as being a grading company, a dealer or a certified appraiser, and none of those claims is made anywhere on this site. Nobody here grades cards, buys cards, sells cards, takes consignments or earns a commission on a transaction. Merey Tleugazin is a software founder who collects, not a professional in the collectibles trade, and that is stated plainly because the alternative — implied credentials — is exactly what makes so much writing about card values untrustworthy.
So the authority here has to come from somewhere else, and it comes from two places. The first is showing the arithmetic: every calculator prints the working, every table names its source, and any figure that decides something can be recomputed by hand. The second is citing primary sources rather than summarising the same ten blog posts everybody else summarised. When PSA publishes a centering tolerance, this site quotes PSA and links to PSA. When a marketplace fee is quoted, the fee schedule is linked and the number is an input field you can change. A reader who disagrees with a conclusion here can find out exactly where it came from, and that is a better guarantee than a credential.
| Not this | What it does instead |
|---|---|
| A grading company | Published tolerances are transcribed here with the date read and a link. No grade is issued, implied or predicted |
| A dealer, auction house or broker | Seven routes to cash are compared net of fees. Nothing is bought, sold or consigned, and no commission is taken |
| A certified appraiser | The arithmetic is shown so you can check it. An appraisal is a professional opinion, and this is not one |
| A price feed | Anything that moves — fees, postage, marketplace percentages — is an editable input with a dated default and a source link |
| Affiliated with the companies named | PSA, BGS, SGC, CGC, TAG, eBay, Topps, Panini, Upper Deck and the leagues appear because a card cannot be described without them |
How a number gets onto this site
The editorial method is the differentiator, so it is written down rather than implied. Four rules decide what may be published, and each of them exists because the alternative failed somewhere first.
- A reference value is transcribed, dated and linked. Every published tolerance, era boundary and structural figure carries the date it was read — currently 2026-09-11 — and a link to the publisher’s own live page rather than to a copy of it. Where the two disagree, the publisher is right and this site is stale, and the link is there so you can find that out in one click.
- Anything that changes quarterly is an input field, not a constant. Grading fees, postage, insurance rates and marketplace percentages are all editable, with a dated default and the fee schedule linked. A hardcoded fee is wrong within a quarter and still looks maintained, which is worse than publishing none — so no grading fee schedule is published here at all.
- There is exactly one copy of the arithmetic. The formulas live in a single module that the browser tools load, that the JSON endpoints import at build time, and that the machine endpoints run. A site with a calculator in a page script, a second implementation in an API and a third in an agent service has three answers to the same question, each individually plausible, and nothing comparing them. Here a build gate fails if they diverge.
- Every figure is built from data the page already states. Charts are generated from the tables printed beneath them, never from numbers typed under a drawing. A diagram that cannot be checked against anything is decoration at best, and a plausible-looking chart of nothing is worse than no chart — particularly on a page about money.
| The rule | What it prevents |
|---|---|
| A reference value is transcribed, dated and linked | A number that quietly goes stale while still looking maintained |
| Anything that changes quarterly is an input field, not a constant | A fee schedule hardcoded in 2026 being quoted to somebody in 2027 |
| There is exactly one copy of the arithmetic | The browser, the API and the machine endpoints each drifting into their own version of the same formula |
| A figure is built from data the page already states | A chart that cannot be checked against anything, or worse, a plausible-looking chart of nothing |
| What | How many | Notes |
|---|---|---|
| Free calculators | 24 | Every one of them runs in your browser and transmits nothing |
| Open datasets | 7 | Keyless JSON under CC BY 4.0, with an OpenAPI description |
| Scouting reports | 12 | Value guides and grading explainers, written and signed |
| Primary sources cited | 8 | Linked to the publisher’s live page, never to a copy of it |
| Sports the app covers | 5 | NBA, NFL, MLB, Hockey, Soccer — the app, not the site |
Every page also has a markdown twin — append index.md to any URL — so a model reading this site gets
the text rather than the layout. That is published for the same reason the datasets are: a source that is cheap to
read correctly is a source that gets quoted correctly.
What the app does
Valdar is a recognition engine wrapped in a camera. Point it at a card, raw or slabbed, and it identifies the player, set, year and parallel, then shows what that exact card has actually been selling for: the low, the median, the high. Sold prices rather than asking prices, and no book values from a decade ago.
Computer vision
The camera reads the card the way a sharp-eyed dealer would: design, borders, foil pattern, print details — down to the parallel.
Sold-comps engine
The identified card is matched against eBay sold listings and third-party pricing databases — transactions that actually cleared, refreshed continuously.
AI condition read
Centering, corners, edges and surface get an estimate, so the submission decision can be made before the fee is paid.
The third step is the one that needs its limits stated out loud, and they are the same limits this whole site is built around. An estimate from a photograph is not a grade. Only PSA, BGS, SGC, CGC and TAG issue grades, the app is not affiliated with any of them, and a professional grade may differ from anything the app predicts. What the estimate is good for is deciding whether to pay the fee at all, which is a cheaper decision to get right than the submission itself. The expected-value calculator does the same arithmetic in the open, on numbers you set yourself.
How it got here
The question that started it
A revisited childhood collection, an evening of manual price lookups, and the decision to stop guessing and start building.
Teaching a camera to read cardboard
Months of training runs on card designs, parallels and slab labels. The hard part was never “is this a baseball card” — it was telling a base card from its refractor at an angle, in bad light.
Valdar ships on iOS and Android
The recognition engine, the sold-comps read and the condition estimate, shipped as one app. The free tier: scan and identify cards, see market value estimates, start a collection.
Scouting reports
What the sold-comps data shows, written up as 12 value guides and grading explainers — honest numbers, no hype.
The arithmetic, in public
24 calculators, the reference layer they compute from, and 7 datasets published as open JSON so anyone can check the working or build on it.
What we will not do
“A card is worth what it sold for — not what somebody is asking.”
The rule behind every value figure here“If the honest answer is that your card is worth a dollar, the page should say so.”
The rule behind the era checker and the junk-wax pages“An estimate is a decision tool, not an appraisal. We say that out loud, on every page that produces one.”
The rule behind the disclaimers in the termsThose are not slogans, they are constraints, and they cost something. They are why no page here ranks for “free card appraisal”, why the value tools report a range rather than one confident number, and why several articles end by telling a reader that the card they are holding is common and grading it would lose money. A site that never delivers bad news is a site optimising for something other than the reader.
The publisher
Vast Flow, LLP is the legal entity behind both the app and this site, and it is the entity named in the privacy policy, the terms of service and the App Store and Google Play listings. It is run by Merey Tleugazin, who writes the reference layer, maintains the calculators and answers the support address at [email protected].
Corrections are the most useful thing anyone sends. A transcription that has gone stale, a source link that has moved, a formula that disagrees with the table above it — all of them get fixed in the one place the number lives, which means the page, the dataset and the machine endpoint move together. The contact page sets out what to include so a correction can be checked the same day, and accepted corrections are credited.
The fastest way to see what any of this means is to use it: start with the 24 free calculators, read what the data actually shows in the scouting reports, or take the underlying numbers from the open API and check them yourself.