The shift from a paper ledger to a website did not change how a satta matka result is made. It changed who can check it, and only on the surfaces that let them. That distinction is the honest version of the story, and it is measurable on one platform without interviewing anybody.
Most writing about this subject is unmeasurable. The standard historical account — the one this site's own history of satta matka works from — documents the analog era in detail: the New York Cotton Exchange closing its rates in 1961, which removed the numbers the game was built on; Kalyanji Bhagat's Worli draws from 1962; Ratan Khatri's New Worli panel from 1964; a 1995 press figure of more than two thousand bookies operating in Mumbai. Then it stops. The same article mentions that players moved on to "other forms of gambling such as online and zhatpat lotteries" and gives no date, no platform and no number for the move. So there is no adoption statistic in this post, no market size, and no decade-by-decade narrative of servers replacing clerks. Those would be invented, and an invented figure in a thought-leadership piece is the same failure as an invented result in a chart.
What can be measured is one live platform's stack, read on 29 October 2026: its market registry, its results archive, and the three surfaces that sit in front of them. Everything below came from a request, a source file, or the shipped Android package, and the two fixture files behind it are dated.
A digital result starts as a row in a registry
Before any number exists, the platform has to know which markets exist, when they open and close, and which days they draw. That table is the least glamorous part of a matka site and the most consequential. Read from this site's feed on 29 October 2026:
| Measure | This platform, 29 October 2026 |
|---|---|
| Market rows in the live registry | 32 |
| Distinct market codes | 32 |
| Distinct open-to-close windows | 30 |
| Codes sharing a window with another code | 4, in two pairs |
| Markets this site charts | 19 |
The pairs matter more than the counts. Two markets open at 15:05 IST and close at 17:05 IST — the codes MILAN_DAY and RAJDHANI_DAY. Two more share 21:30 to 23:30: KALYAN_NIGHT and OLD_MAIN_MUMBAI. A registry can therefore contain two live markets that are indistinguishable by timing, which means the only safe join key between the feed and a page is the code.
This site learned that the hard way, and the learning is preserved as a comment in src/data/markets.ts rather than quietly deleted:
The previous model (src/data/games.ts) carried display names only, and the chart pages matched them against the API by comparing lowercased display names at runtime. That silently failed for five markets, which then rendered invented numbers.
apiCodeis now the join key: it is stable, it is what the API returns, and a mismatch fails the build instead of the page.
That is the first generation of the digital platform reproducing a very old problem. An agency knew its own book by sight; a website matching names guessed, and when it guessed wrong it printed a number that had never been declared. Five markets, on the site's own account. The fix — make the join key stable and let a mismatch break the build — is the single most useful thing a results site can do, and it is invisible to readers.
The registry also shows what "more markets online" really means. Thirteen of the 32 codes are not charted here at all: three SITA variants, four STAR variants, two SRIDEVI variants, plus GEETA_MORNING, BALAJI_NIGHT, OLD_MAIN_MUMBAI and RATAN_KHATRI. The registry itself still held 28 rows when the verification guide counted them on the morning of 20 October, and returned 32 by 23 October, where two later measurements agree — and it is still 32 today. That is what growth looks like in a registry nobody audits: names duplicated, windows reused, and a public page choosing 19 of them and publishing a chart for each. Coverage is a decision, not a feature list.
The archive exists — and three surfaces read it differently
The interesting question about a digital platform is not whether it has history. It is how much of its history each of its own surfaces is willing to show. This site's upstream serves a per-market archive at a guessable address, and the difference between its surfaces is a measurement, not an opinion:
| Surface | Dated rows it published on 29 October 2026 | Range |
|---|---|---|
| Upstream results archive, for the 19 charted markets | 378 | 2026-09-09 to 2026-10-28, on 23 distinct dates |
| This website's pana and jodi charts | 75 complete records | 2026-09-10 to 2026-09-14, on 5 dates |
| The Android app's chart screen | whatever the archive returns, fetched when you open the screen | live |
19.8 per cent. That is the share of the available, dated, already-served history that this website actually publishes, and the gap is not a storage problem or a rate limit. Across the union of those 23 dates there is exactly one step wider than a day: the last September row is 14 September and the first October one is 12 October, so for 27 days — 15 September to 11 October inclusive — no charted market has a row at all. Two independent posts measured that same hole from the other end — the Rajdhani guide and the historical-charts piece — so it is a property of the feed, not of this site's snapshot.
Worked example: fifteen Kalyan results you cannot find here
Kalyan is the market a reader is most likely to ask about, so count it directly. The archive for its code returns 19 dated rows, from 2026-09-10 through 2026-10-28. This site's Kalyan chart holds 4. Fifteen dates exist upstream and appear nowhere on the chart page: 12, 13, 14, 15, 16, 17, 19, 20, 21, 22, 23, 24, 26, 27 and 28 October.
Two October dates are missing from that list, and they are the interesting pair. The 18th and the 25th are Sundays, and the registry gives Kalyan Monday through Saturday. So the site's absence for those two dates is a schedule, and the absence for the other fifteen is a publishing decision. Being able to tell those two apart is the actual skill — the same distinction the results-verification guide makes about blank cells.
The call the website never makes
Here is the finding that reframes the whole "digital transparency" question, and it is eight lines of source code deep.
src/data/api.ts:176 defines a function called fetchMarketCharts, which requests the archive path for a market id and returns the rows. Searching the site's source for that function name finds exactly one reference: the definition itself. No page imports it. No component calls it. The website has an archive client that nothing uses.
The Android app calls the same endpoint. Its chart screen chunk fetches the market archive on open, and if the response carries a chart array it renders it, otherwise it renders an empty state. The heading on that screen, read out of the bundle, is "Panel & Jodi Historical Records" — the archive is a first-class feature of the phone and an unused export of the website.
That is the cleanest possible illustration of what "digital" actually delivers. Nothing about the data model prevents a reader from checking a result from three weeks ago. The upstream serves it, the app shows it, and the browser does not, because the function is there and nothing points at it. A platform's transparency is a property of its surfaces, not of its database.
What the app's own client says about the business
The shipped Android package — 5,758,251 bytes, SHA-256 beginning 39d41929017e, the same file the download guide hashes and the troubleshooting guide reads — contains a single network client with 23 distinct request methods. Grouped by what they are for:
| Group | Count | Examples, read from the bundle |
|---|---|---|
| Public data | 6 | /markets, /markets/settings, /markets/game-rules/rates, /markets/notices, /markets/banners, /markets/{id}/charts |
| Sign-in and device | 3 | /auth/check-number, /auth/me, /auth/fcm-token/unregister |
| Money | 5 | a bid-placement path, /wallet/balance, a native-deposit verification path, and two bank-details calls |
| Operator | 9 | a result-declaration path, a global-betting switch, market creation, deposit approval, banner management |
Twenty-three counted calls, and a refresh path plus a logout path also sit in the same chunk, assembled from a variable base rather than written out as a literal call — so the real surface is a little wider than the table.
Read the last group slowly, because it is the whole history of the trade compressed into four path names. One of them takes a result declaration as a submission. A matka result is not computed and published; it is entered. That is not an accusation, it is architecture: the chain from a chit to the number on your screen ends at a form someone with an operator account fills in. Every arithmetic check a reader can run — and this site publishes several — catches a typo at that form. None of them catches a person.
The other eight operator paths are a global switch that turns betting on or off, a market-creation call, a deposit-approval call, a dashboard-stats read, an operator-side game-rules read, and three banner calls. That is a back office. The public chart viewer and the back office are in the same bundle, which is the most honest possible answer to what a "digital matka platform" is: the website and the wagering operation are one product with two doors, not a neutral archive that happens to have a partner.
None of those endpoints was probed. A request that places a bid or approves a deposit is an act, not a measurement, and this page does not perform acts. The evidence is the shipped client, and the claim is only that the code can make the call.
Where digital genuinely is more transparent
The ledger era had one verification method: know the agency, or know somebody who did. A platform that publishes any of the following gives a reader something the paper book never did, and this one publishes four of them.
- A dated machine-readable copy. Every chart page carries an as-of date, and the dataset sits at
/data/chart-snapshot.jsonwith its owngeneratedAtfield. - Byte identity across environments. The file in
src/, the file in the build output and the file live on the domain hash to the same value,4daa09dd82f0…. There is no separate production dataset to be confused with the one the code was reviewed against. - A publisher that refuses to pad. The script that regenerates the snapshot exits non-zero rather than writing a file with holes in it, so a gap on the page means a gap in the feed and not a silent placeholder.
- Arithmetic that is checkable line by line. Every published pana has to reduce to the printed ank, which is what lets the verification guide report 0 rule-breaking rows across 202 live records instead of asking for trust.
That is real. It is also narrow, and the narrowness is the point: transparency is a set of specific properties, each verifiable on its own, not a mood a website projects.
Where it is not
Four measured gaps, in the same platform, on the same day.
The newest result this site publishes is dated 2026-09-14. The archive it could have published from held a row dated 2026-10-28. Forty-four days, and the site's own homepage copy says results are "read directly from the Mama567 live market data feed in real time … synced continuously with the official market servers", which is true of the live board and untrue of the charts — the page that measures the mobile build documents the same contradiction and the editorial policy's own line that pages are rebuilt on a schedule, not continuously.
Thirteen registry markets are uncharted and the site publishes no reason. Seven built pages placed content past a 412-pixel viewport, where the stylesheet clipped it instead of scrolling — counted across the 95 pages built on 28 October, and a reminder that a record you cannot read on the device in your pocket is not transparent in the way that matters. And the formats the site sells were undefined on it until yesterday's guide transcribed the definitions out of the app: seven priced products, no published rules, until the rules were reverse-read from a bundle.
Five checks you can run on any matka platform
- Ask for the machine-readable copy. Request the JSON or CSV the page was built from. A site that assembles numbers in the browser and offers no file is showing you a picture.
- Compare surfaces on the same date. Open the app, the chart page and the results table for one market and one day. Three surfaces that disagree have a join-key bug, and this site has shipped one.
- Date every number you read. A result without a date and a page without an as-of line are both undated claims, however precise the digits look.
- Run the arithmetic and the schedule. Digits must add to the printed ank; declared dates must fall on days the market says it draws. Blanks that match a draw calendar are honesty; blanks that do not are holes.
- Ask who writes the row. Look for the operator surface — an admin panel in the bundle, a declaration form, a switch that stops betting globally. Every platform has one, and knowing it exists changes what "official result" means.
What the next version of a platform like this would publish
Not a roadmap, and not a promise about anybody's plans — a list of changes a reader could verify in one request each, derived from the measurements above. Publish the archive the endpoint already serves, with the same dated file. Wire the call that already exists in the source instead of leaving it unused. Replace "synced continuously" with what the poll actually does, per surface. Say why 19 of 32 markets are charted and which page shows the rest. Print the operator model — who declares, and what a reader can check about a declaration.
None of those requires more data than the platform already has. That is the uncomfortable and useful conclusion: going digital did not make a matka platform honest. It made honesty cheap to verify and equally cheap to withhold, and the difference between the two is now a decision about which calls to wire up.
The same three surfaces, read for what they can send rather than what they publish, are measured in How to Set Up Notifications for Live Market Declarations — including the notices and banner feeds the API exposes empty and the website never calls.
The record this site does publish is dated, machine-readable and checkable — start with the chart snapshot itself, then all charts, today's declared results and the editorial and data policy that says how each is built. The vocabulary needed to read that record — chart, cell, placeholder, ank, and the two anks that form a jodi — is defined term by term in the satta matka terms glossary.
Satta matka is illegal in most Indian states; check the law where you live. Mama567 is an information platform — it does not accept wagers or sell predictions, and no number here is presented as more likely than another. 18+ only.