Robinhood Chain
Hold $PLINKO and earn $BALLS every day. Drop them on the board to win real tokenized stocks — paid for entirely by trading fees.
Live numbers · the board animation is a simulation
Real prizes switch on at $5,000. Until then you can practise in Free play, and holders earn $BALLS from the first block — drop them in Ball mode, or bank them for the moment the vault opens. Before the threshold, 100% of trading fees go to the prize pool: no buyback, no team take.
How it works
Same pegs, same odds, same physics. What changes is what you drop and what you can win.
Connect a wallet and take 10 free balls. They play against a pretend $45,000 pool, so you can learn the board and feel a win before you own anything.
Nothing won here is real.
Hold $PLINKO and $BALLS arrive every day. Drop them to win more $BALLS — or save them up for when real prizes switch on.
$BALLS are play credits, not a token: they can’t be bought, sold or sent.
Once the prize vault holds $5,000, every lane outside the centre pays real tokenized stock — SPY, GOOGL, META, TSLA, NVDA, MSTR — and the outer lanes pay the jackpot in SPCX.
Prizes are a share of the live pool, never a fixed promise.
Proven on-chain: a win paid out as NVDA →Getting $BALLS
Holding is the whole job. The rest is four rules.
2,000 $BALLS are minted each day and divided among holders. More holders means a thinner slice each — never a bigger bill.
Hold twice as much, earn twice as many. No curve and no cap — so splitting your bag across wallets gains nothing.
Your share grows over 20 days of holding. Selling or sending any $PLINKO resets it to zero.
Earn 0.4 a day and you drop on day three. Nothing is lost to rounding, and saving up for the real board is a strategy.
Where the money comes from
Pons takes 1% of every trade and routes 70% of it back to the token's creator. That 0.7% of all volume is the only money in the system.
At launch the split is suspended — 100% of revenue goes to prizes until the pool reaches $5,000. Until then drops pay $BALLS, and the Real prizes board unlocks the moment it’s funded.
Made to be shared
Your phone renders a short vertical video of that exact drop — the real path, the slow-motion finish, the payout counting up — ready for X, TikTok or the group chat. Made on your device; nothing is uploaded.
Every drop resolves from a seed committed before you played. When the round closes the seed is revealed, and anyone can recompute where any ball should have landed.
Free play is open now. Connect a wallet, take 10 free balls, and see how it feels.
Changelog
Including the things we got wrong and reversed. A design that only ever moved forward would mean nobody was checking it.
The prize vault is live, and it has paid a win end to end: $9 of USDG in the vault, a d6 win credited, and the claim swapped through the real NVDA market and landed 0.0000438 NVDA in the winner’s wallet at $229.52 a share. See the transaction.
The win was a cent, on purpose: prizes are a share of the pool, and the pool was $9. Same contract, same swap, whatever the size — the size comes from volume.
The app is one screen now: pick a game, see one number, drop. The drop button floats under your thumb, the whole board fits without scrolling on a phone or a laptop, and the button always says the next thing to do — connect, grab free balls, or drop. Wins pop up over the board and count into your balance; big ones throw confetti. Space drops on a keyboard.
The app now opens on a switch: Free play with test balls against a pretend pool, Ball mode where the $BALLS you earn by holding are dropped for more $BALLS, and Real prizes — shown locked, with the vault’s distance to $5,000 on the tab itself. Your $PLINKO balance and your $BALLS, counting up to the sixth decimal, sit at the top of the page instead of down a menu.
Test balls used to land in the same balance holders earn into, so free play could be spent on the real board. They are separate balances now, and every ball from before this change — there was no token yet — moved to free play. The balance counter also ticked upward for wallets earning nothing; it only moves for a holder now.
Tap Share win and your phone renders a nine-second vertical video of that exact drop — the real path the ball took, the slow-motion close-up as it lands, the payout counting up — ready for the share sheet. It is made on your device; nothing is uploaded. Wins from the free trial are stamped “free play” on the video, because a clip that gets passed around should not be able to pass for real money.
Found on the way: a ball balance could drift to 0.9999999999999989 after a few centre refunds, and the server then refused a ball you could see on screen. Balances are rounded now.
Claiming now converts your winnings into the stock the lane names, at the moment you claim. The price floor is derived from the pool’s own thirty-minute average, so a claim refuses to execute if someone has just pushed the price — and the floor is set by the contract, not by us.
Checked against the live NVDA pool rather than a stand-in: spot sits four ticks from the half-hour mean, against a guard that trips at a hundred. Taking plain dollars instead is always available and always will be, because a claim that cannot fail is worth more than one that pays in the prettier asset.
Holding earns balls from the first block, and the page said those balls wait for the machine to open. They did not — you could spend them against a pool with almost nothing in it.
One test drop landed a 1-in-59 lane and paid $0.0001, because the vault held 46 cents at the time. The odds and the arithmetic were both right, which is what made it easy to miss. Dropping is now refused until the pool crosses the opening threshold, and the ball stays yours.
Every lane now names its stock on a small screen, prices scroll underneath as a ticker, and full-screen mode gives the board more than half the display instead of collapsing it to nothing.
Seventeen lanes across 375 pixels leaves nineteen pixels a lane, which fits a ticker and not a price. Casino boards manage it by showing a four-character multiplier and no stock name at all — the same budget, spent differently.
The progress bar toward the opening threshold was tracking the simulated pool that drives the board above it, so it read threshold reached on a page whose vault was empty.
Anyone checking whether this had launched would have been told yes. It now counts only a live balance, and sits under the board with the multiplier ladder it feeds.
The app is hosted now, so the button at the top of this page actually goes somewhere. Playing requires connecting a wallet — either a browser wallet or Robinhood Wallet by QR code. You can still paste any address to look at a portfolio without connecting; that data is public on chain anyway.
Connecting asks you to sign a message, never a transaction: no gas, and it cannot move funds. Until today the app took an address on trust, which meant anyone could spend anyone else’s balls by typing their address. Prizes were never at risk — those are claimed on chain by the winning wallet itself — but the ball ledger was.
The three innermost prize lanes paid SOFI, AMD and RBLX. They now pay SPY, GOOGL and META. SOFI was the reason: it sat on the lane hit roughly one drop in five, with $7,988 of liquidity on chain — about fifty times thinner than anything else on the board, and thin enough that a single winner cashing out in a quiet hour would move the price against themselves.
Every lane now has at least $1.6M of depth behind it, so no lane needs a special case. It reads better too: the most common thing you win is a slice of the S&P, and the ladder climbs from index fund to MicroStrategy.
Found by checking the actual pools rather than assuming they existed. Worth doing before building the payout path, not after.
This page and the live app each had their own copy of the board, and they had drifted twice over: different proportions, different easing, different chip sizes, and a green ramp that was legible here and nearly invisible there. Both now run the same code, so they cannot diverge again.
Merging them nearly cost the jackpot sequence — the slow-motion camera move only existed in this page's copy, and the shared version had to learn it before either page could use it. Deleting something by consolidating is the failure mode worth watching for.
The vault collects fees in USDG and the board names a stock for each lane. Those two facts do not meet yet: the tickers name what a prize is worth, not what is currently sitting in the vault.
Being explicit because it is the whole pitch. Closing it means converting a prize into the winner's ticker at the moment they claim — the pools to do that all exist and are deep, which is what the lane change above was really for. Until it ships, treat the tickers on the board as denominations.
Each round commits to a hashed seed before any drop is resolved, then reveals it after, so anyone can recompute every outcome and confirm nothing changed mid-round. Verified against 400,000 simulated drops — chi-square 20.0 against a 39.3 threshold.
Honest limit: this proves the server didn't tamper after committing. It doesn't prove the seed wasn't chosen favourably beforehand. Chainlink VRF replaces it in phase 2, at which point nobody knows the seed in advance.
26 tests, including 512 randomised runs confirming the vault can never owe more than it holds. The server submits which lane a ball landed in and never an amount — the contract derives the prize itself, so a stolen server key can misattribute a prize but cannot invent one.
Trading fees are routed to the vault contract directly and never touch a personal wallet. After launch the key that could redirect them gets burned, so the fee stream becomes permanently locked to prizes — including against us.
We had balls scaling to the 0.75 power of your balance, to stop whales dominating. Modelling showed it was solving a problem the ball count had already fixed — it moved a $100 holder from a drop every 4.2 days to every 2.1, at the cost of a formula nobody could read.
Reverting it also removed the minimum balance, the per-wallet cap and the anti-sybil snapshot logic, because linear weighting makes splitting a bag across wallets pointless by construction. Three mechanisms deleted, one sentence gained.
Earlier the same day we moved it off the board to an independent roll, because as a lane its cadence was coupled to ball count — more balls, more frequent hits, a smaller headline number.
Then we animated it, and watching a ball land on an ordinary lane and pay out a jackpot looked broken no matter how the odds were explained. The coupling is one constant to remember when tuning; the confusion was every single view. It's the two outer lanes again — 1 in 32,768, about every 16 days, and roughly half the pot hitting twice as often.
Every payout is a fixed fraction of whatever is actually in the pool, so prizes rise and fall with real trading volume. No dollar figure is written into the contract, and it can never promise more than it holds.
At 79% of drops, a centre hit would have paid about six cents of stock — costing more in fees to move than it was worth. Centre lanes now return 0.2 balls instead, which concentrates the entire stock budget into the drops people care about and made those prizes roughly five times larger.
Originally the pool had to be funded out of pocket before the machine could open. Instead, 100% of early revenue goes to prizes until the pool crosses its opening threshold — so the game funds itself from trading fees and opens in days.
Robinhood Chain has 194 tokenized stocks trading as ordinary tokens, and no one has built a game that pays them out. Unlike collectible prizes, they're liquid at full value the moment you win them.