This summary is private. Enter the password to continue.
TableMaster / Event Planning
Decided. We are running the hybrid: One-Eyed Jack's keeps their night and their system of record, and our software takes registration, the clock, and the floor signals alongside it. This page is now the record of that choice and what it commits us to.
We are running the hybrid. One-Eyed Jack's runs their night on PokerAtlas as they always do - that stays the system of record and the safety net. Our software runs alongside it: players check themselves in at a kiosk, the window works from a live roster, our clock runs in parallel on every surface we have, and ten invited players use the app from their seats.
Kiosk check-in at the door, live roster behind the glass, dealer taps free the seat. Money still changes hands at the window.
In parallel with PokerAtlas, shown on the TVs, on every dealer tablet, on the director's tablet, and in ten players' pockets.
We run on the room's existing screens rather than rolling our own in. The truest test available, and the one that needs the most care.
This is the top of the hybrid lane, not the middle of it. Every surface we have built is doing a real job on the night - kiosk, window, dealer tablets, director tablet, screens, phones - with PokerAtlas underneath as the thing that is authoritative if ours stumbles. The safety net is real, and it is the whole reason this is a responsible choice rather than a brave one.
Two consequences follow and neither is optional. The clock sync is a job. Somebody owns keeping ours matched to PokerAtlas for the whole night, and on the room's own TVs a drift is visible to everyone. The room's AV is their equipment. Anything that goes wrong on those screens will be remembered as ours whether or not it was, which raises the bar on the display work below from "should check" to "must prove".
Kept as the record of the choice. The September options page holds the same ground in five options, A through D. The question that separated them: whose software is responsible for the night? The answer we landed on is that PokerAtlas stays responsible, and ours does the work.
One-Eyed Jack's runs the tournament exactly as it does today, on PokerAtlas and paper. Our software is not operational. The signup site and the app can still exist around the event.
If something fails: nothing can. No one would notice.
= option A
The room's own process stays in charge. We assemble the night from standalone pieces: a kiosk at the door, a live view behind the window, dealer and director tablets, maybe our clock running in parallel. Paper never stops being the fallback.
If something fails: staff shrug and do that one step the old way.
= options B1 / B2, plus the kiosk flow below
RoomDirector is the source of truth: registration, clock, seating, payouts, and optionally dealer tablets and the clock in players' pockets. The product finally does the job it was built for - with no second system already running.
If something fails: we recover live, in front of paying players.
= options C / D
The comparison that settled it. Kept in full, because the reasons we did not pick lanes 1 and 3 are the reasons to hold our nerve if the build gets tight.
| Lane 1 · Their night | Lane 2 · Hybrid | Lane 3 · Our night | |
|---|---|---|---|
| The pitch | Zero risk. The event is about the brand, not the software. | Real proof on real players, with the room's process as the net underneath. | The full story: our software ran a real tournament, end to end. |
| Pros | Nothing to build. Nothing to rehearse. All attention on hosting a great event. | Players touch the product. Staff see the value. Every piece has paper behind it. Failure is invisible to the room. | Strongest proof and demo material. Points, results and history land in the platform for keeps. |
| Cons | We learn nothing about the product. The night proves only that we can throw a party. | Two systems in the room; the receipt-to-cashier handoff is reconciled by hand; less impressive than a full takeover. | Highest stakes: live recovery in front of paying players, staff training, settlement never done for real. |
| Biggest risk | Missed opportunity - our first event teaches us nothing. | A kiosk or screen hiccup looks scrappy for a minute; the tournament itself cannot be hurt. | An untested surface fails mid-event with no second system running. The rehearsal calendar, not the code, is the constraint. |
| Dev work left | None required. | Days. One kiosk page (blocks exist), printer confirm, small choices below. | Days of build, but the real cost is rehearsal: training, dry runs, one real settlement. |
| Confidence today | Proven - it is how the room already runs. | Untested pieces, tested net - each piece degrades to paper. | Untested - built months ago, never run a real event. |
| Player app | Independent of the lane. Invited TestFlight group is ready today; public stores are weeks plus a first Apple review. Push broadcasts are live and proven. | ||
The August 24 assessment called the check-in tablet "the least explored" idea on the board and priced automatic receipt printing at weeks of Windows work. Both readings were out of date: every building block of the kiosk flow already existed in the product - a merged 80mm thermal receipt page, live-updating registration views, the seat signals - and silent printing turned out to be a browser flag rather than a Windows build. That is what moved the hybrid from an idea to a plan inside a single day.
Both questions were answered yes, so every piece below is in. This is the full inventory of what has to work on the night, and what state each piece is in today. Nothing here is speculative any more - it is the build list.
These never depended on anything. All running.
All of these are now in scope for the night:
All of these are now in scope for the night:
Running our clock next to PokerAtlas turns the night into a live demonstration of RoomDirector - pausing, level changes, moving players - things the room has never had, with their software still authoritative underneath. If our side hiccups, the tournament does not notice.
The cost is now owned rather than weighed. Because our clock also goes on the room's own TVs, a drift is a public contradiction rather than private feedback - so the sync duty gets a name attached to it before the night, and the rehearsals below exist largely to find out how a real level change feels when two clocks have to move together.
Live Untested on the real glass
A true kiosk app, not a website on a screen. Shipped as a rotating attract screen leading to three equal doors - check in, schedule, leaderboard - with the tournament featured rather than the device being tournament-only.
The data choice answered itself. It was framed as mock data with nothing to keep accurate, against real room data that lands harder but has to stay right. Neither was needed: the schedule reads the same public tournament feed as the consumer site, and the leaderboard reads the same feed as the card already serving on One-Eyed Jack's own website. Real data, kept correct by systems already running.
Live Proven end to end
The blood-work model, and it is built. Pre-registered players type their phone number and are done - the dsptour.com signup list already stores phone numbers in a matchable format. New players give name and phone, the same two fields staff capture today. The entry lands in RoomDirector and gets a seat as it saves. Verified end to end in a real browser on the live address: number in, match found, welcome shown, row written.
A kiosk check-in counts as registered, not paid - money still changes hands at the window, like every other night. The remaining choice is small: does the arrival show only on the window tablet, or also print a receipt the player carries to the glass?
One caveat worth carrying to the night: only 6 of the 14 signups hold a phone number, so most arrivals will go through the new-player path rather than the fast one. That is the path to watch in a rehearsal.
Prototyped Standalone yes/no
The branded 80mm thermal layout is merged and tuned against the Epson driver. For one night, silent printing is a Chrome launch flag - the Windows wrapper is not required. Or skip paper entirely: the window tablet already knows, and when the room runs PokerAtlas its own receipt still prints.
Works today One small addition
The staff tournament screen already lists every player and flashes a new registration within about a second. The addition worth making: the pre-registered-but-not-yet-arrived slice from the signup list, so the glass sees who is still expected.
Built Never run a real night
Each table's tablet shows the live clock and level (when our clock runs) and calls floor, drinks, or a chip runner on a rail the room side sees instantly - food is one more label on the same rail. A bust tap empties the seat on the window's view in the same second. The seat card stops walking.
Built
The same room software on a tablet in the director's hands: floor calls arrive as alerts with a standing queue, players move with a tap, and - when our clock runs - pause and level control from anywhere in the room.
Unchecked sleep fault 24 versions behind Never run on their hardware
Decided: our clock goes on One-Eyed Jack's existing screens rather than a TV we roll in. The truest possible test, and the piece that carries the most care, because anything that misbehaves on their equipment will be remembered as ours whether or not it was.
Three things stand between here and there, and none of them is hard: the screen app has never been checked for the sleep fault already fixed on the dealer tablet, its shared packages are 24 minor versions behind, and it has never run a night on a real TV at the room's resolution. Days, not weeks - but now mandatory rather than advisable.
All three were true on 26 August and all three are still true, verified on 3 September. One thing did come off this card: the screen now recovers on its own after a network drop, which shipped on 31 August. That was the fourth item, and it is the only part of this surface that has moved in nine days. This is the card to read twice.
Chosen
Channel ready today Event alerts never fired
Ten friendly players get the full app through TestFlight - no App Store submission, no public exposure of half-tested features, honest feedback from real seats. They see the clock in their pocket (needs our clock running), and check in by showing their QR card for a staff scan - that path is built. True register-from-the-phone is new.
Break and move alerts exist but have never fired in production - they need a tournament actually running in our software. The one-Mac simulation proves them before the night does.
This is the night we are running. Step through it - each step says what the player and staff experience, and what state that piece of software is really in today.
Works today verified in code, not yet in a room Untested built, never run a real event Open defect New build
| Item | Needed for | Effort | State |
|---|---|---|---|
| New build - written and deployed since the last edition | |||
| Kiosk check-in page: phone-number fast path for dsptour.com pre-registrations, name + phone for new players, register and reset. Built, deployed and publicly serving at kiosk.tblmaster.app since 27 August. Writes through one server-side function, so no credential sits in the kiosk browser | Registration | Done | Live |
| Kiosk idle surface: shipped as a rotating attract screen leading to three equal doors, rather than a tournament-only device. The data question answered itself - the schedule and leaderboard both read live feeds we already serve, so there is nothing to keep accurate by hand | Kiosk | Done | Live |
| Kiosk on the real hardware: full-screen mode on the HP touchscreen, USB keypad, idle reset and the automatic return after a check-in, on real glass rather than a developer's browser | Kiosk | An hour | Untested |
| Window roster: show the pre-registered-but-not-yet-arrived slice from the signup list beside the live entry list that already exists | Registration | Small | New build |
| Receipt printing from the kiosk browser, if we decide on paper: the 80mm thermal layout is merged and tuned; silent printing is a Chrome launch flag, not a Windows build. Ten minutes on the real HP and printer proves it | Open choice | Hours | Untested |
| Their AV, now that we are on it - the sharpest risk on the page | |||
| Display and Dealer are 24 minor versions behind on the shared database package. Both now carry real jobs on the night, and neither picks the fix up on its own - each range has to be raised deliberately and typechecked | Clock, dealers | Days | Severe drift |
| The screen app has never been checked for the session-expiry sleep fault that was found and fixed on the dealer tablet. Worse than the dealer case: the heartbeat outlives the session, so a frozen screen still reports itself online | Clock on TVs | Hours | Unverified |
| The deployed screen is three versions behind on its renderer, so the break countdown and stat-card fixes from August are not in what the room would see | Clock on TVs | Hours | Known gap |
| Stability on the actual TV hardware, at the room's resolution, driven by the room's own AV chain - not a monitor on a desk. Includes graceful recovery after a network drop | Clock on TVs | Days | Never run |
| The clock stage skin has never been reviewed live on real data, only on fixtures | Clock on TVs | Hours | Untested |
| Already built - needs proving, not writing | |||
| Window tablet: the staff tournament screen already lists every player and flashes a new registration within about a second | Registration | None | Works today |
| Seat-open signal: a dealer recording a bust already empties the seat on the window's floor view in the same second | Floor | None | Works today |
| Dealer tablet recovering by itself after a long sleep. The fix shipped; nobody has yet watched a tablet run a full session and come back | Dealers | Built | Untested |
| A dealer pressing Floor a second time inside five minutes reaches nobody. With a tablet on every table, the repeat press is the escalation that matters | Dealers | Hours | Open defect |
| Director tablet: floor calls, table moves, and clock control from anywhere in the room. Built, never used to run anything | Director | None | Untested |
| Rehearsal and configuration - the part that actually decides the night, and the part that has not moved | |||
| The one-Mac, five-surface simulation: kiosk, window, dealer tablet, screen and phone all running one small tournament together. Also the only way to prove the four event notifications that have never fired in production. Still not started nine days on, and the kiosk it needs is now built and available | Everything | Days | Not started |
| Full dry run of an event start to finish in the real room, then a real settlement and payout run once before the night. Still not started, and no dates agreed with the room | Everything | Days | Not started · 23 days |
| Room staff trained and comfortable: the window person on the roster, the director on the tablet, dealers on their tablets. Two One-Eyed Jack's staff already hold operator accounts | Everything | Days | Underway |
| Name the person who owns clock sync on the night, and rehearse a level change with both clocks moving together | Clock | Hours | Unassigned |
| Points: pick the scoring style and set the room's default profile. A configuration task, not a build - but read from the live database on 3 September, zero of 32 rooms has one, so a real event would still award zero points | Results | Hours | Config gap |
| The ten TestFlight players: pick them, get them on the build, and tell them what feedback we want | App | Hours | Channel ready |
The building went well. The last edition said four rows above were new code and none of them was hard. That was right. The kiosk was written, deployed and put on a public address inside two days, and it came out better than specified: the schedule and leaderboard read live feeds rather than the hand-loaded data the plan allowed for.
And the warning in this box nine days ago has come true. It said unproven is what bites on the night, and that it does not get better by waiting. Since then, not one proving item has moved. The room screen still has not been read for the sleep fault, its packages are still 24 versions behind, and it still has not been driven by One-Eyed Jack's AV. The rehearsals have not been scheduled. The five-surface simulation has not started. What changed is only the calendar: 31 days became 23.
The two rows to watch are unchanged, which is the point. The shared-package drift on Display and Dealer, and the unchecked sleep fault on the screen app. Both were survivable when our screens were optional. On the room's own AV, with our clock on it, they are the difference between a demonstration and an apology. Together they are perhaps two days of work, and they have been outstanding for nine.
The deadline is not the tournament - it is the last rehearsal before it. Anything we rely on must be finished early enough to run end to end at least twice, and now that the room's own AV is in scope, at least one of those runs has to be on their equipment.
The build window that was eleven days is now about three. The last edition put the cut-off at roughly 6 September and had eleven working days in front of it. Eight of those have been spent, and the work they bought was real: the kiosk exists, the dealer tablet gained shift tracking, high hands shipped, and the credential gap closed. But none of it was on this page's critical path.
What still fits, comfortably: the remaining build. The sleep-fault read is hours; the package bumps are a day or two. Three days is enough for both if they start now. What no longer has slack: the rehearsals. Two runs in the real room, at least one on the room's own AV, with staff present, and the dates are not agreed. That was described as needing agreement "this week, not in September" on 26 August. It is now September. This is the item to move today, and it is the only one on the page that depends on somebody else's calendar. What does not fit, and does not need to: a public app-store release. The ten TestFlight players need none of it, which is precisely why that was the right call.
The shape of the night is agreed. What follows are the smaller calls that the plan now depends on - each one small, each one currently unowned. All six were open on 26 August. One has been answered by the build rather than by a decision; the other five are still open, and there are now 23 days rather than 31. None of them is hard. That is what makes leaving them open expensive rather than risky.
The hybrid, with PokerAtlas as the system of record and the safety net. Registration runs through our software - kiosk check-in at the door, live roster behind the glass, dealer taps freeing seats - and a check-in counts as registered, never as paid. Our clock runs in parallel, in full, including on the room's own AV. Ten invited players use the app from their seats through TestFlight, which means no app-store release is on the event's critical path.
Still the single highest-value item on this page, and the only one with someone else's calendar in it. Two runs, at least one on their AV. The last edition asked for these to be agreed "this week rather than in September". It is now September and they are not agreed. Everything else on this list can be settled in an afternoon by one person; this one cannot, because it needs the room to say yes to a date.
A name, not a role. Our clock is now on the room's TVs, so a drift is public. Whoever it is should be in both rehearsals doing exactly that job.
The layout is merged and tuned, and silent printing is a browser flag rather than a build. The question is whether the player carries something to the window, or the window simply knows they are coming.
This was a choice between mock data with nothing to keep accurate, and real One-Eyed Jack's data that lands harder but carries an accuracy duty. The build answered it and neither option was needed. The schedule and the leaderboard both read live feeds we already publish and already keep correct: the same tournament feed behind the consumer site, and the same leaderboard feed behind the card already serving on the room's own website. Real data, with no new accuracy duty, because nothing is hand-loaded. The idle screen rotates through those alongside the featured event.
Two things, one conversation, and the first half has drifted past its occasion. The series-points proposal was built for the venue's Summer Showdown, feedback was due 30 August, and that date passed without an answer. The series itself ran from 17 to 30 August and has finished. Nothing is lost but the timing: the model is sound and the durable version is a variation of a leaderboard card we already ship, so it can point at their next series instead. It needs a decision about which version to build, or an explicit decision to park it.
Friendly faces who will actually report back. They need to be on the TestFlight build and briefed on what feedback is useful before the night, not handed a phone at the table.
The decision was taken on 25 August and has not changed. This page was refreshed on 3 September from the live issue tracker, a direct read of the production database, and a read of the version files behind every drift claim. It supersedes the September options assessment of 24 August, which is kept as the record of the five options considered. Effort figures are build time only and deliberately coarse. "Works today" means verified in code, not exercised in a live room - the rehearsal items exist precisely to close that gap.
What this refresh changed, and what it did not. The plan itself is untouched: the hybrid, the safety net, the surfaces and the ten TestFlight players all stand as agreed. What changed is the state underneath it. The kiosk moved from the only unwritten thing on the page to a live application. Every proving item stayed exactly where it was, and the calendar moved eight days. Nothing here argues for a different plan. It argues for starting the rehearsals.
Companion documents: portfolio overview, build queue, full-stack scope.