HOW IT IS MEASURED
The rules, the arithmetic behind them, and the things about this game we would rather you heard from us.
Books are ranked on their own percentage growth, never on size. Ranking by value instead put the leader at +58,151% against a +149% median — the season was decided in week one. Percentage ranking gives a leader at +98.3% against +8.6%, which is a live ladder.
Week-over-week rank correlation is ρ = 0.031, and a top-half book stays top-half 51.7% of the time against 50% for a coin flip. A book that beat the field last week is no likelier to beat it next week. We make no claim of skill.
Every book is $100,000 of paper. No tokenised stock is bought, held, or transferred on anyone's behalf, and no value moves between books. The only transactions you ever sign are a mint, a fee you chose to pay, and a claim if you win.
278 books are cut each week on that week's return. Seven cuts leave 276 into the final week, which selects ten finalists ranked on their chained 8-week score — so every week you survived counts toward the payout.
How a week is scored
- Two marks. A cycle opens Saturday 00:00 UTC and closes Friday 21:00 UTC. Each mark is the last price recorded at or before that instant — never interpolated, never reached forward. Scoring a week on a price that arrived after it ended is not scoring that week.
- From a published tape. Every mark is a line of an append-only price record, and each settled cycle names the exact timestamps it used, so anyone can recompute the result rather than take it on trust. If a mark is more than 6 hours stale the cycle will not settle at all: a week scored on another session's price is worse than a week scored late.
- Your book's return is the weighted sum of its holdings' moves, which is exact for a book rebalanced to those weights at the open — and the lock-in is exactly that.
- A holding we cannot price is held flat for the week and named on the board. Renormalising onto your remaining weights would quietly change your allocation at settlement, which is the worse error.
Positions that cannot lose, and why they are not allowed
A field scored on weekly percentage growth has one structural weakness: a book that cannot move beats every real portfolio in any down week, and so survives every cut without picking anything. Four versions of that were found and closed, and they are worth naming rather than hoping nobody looks.
- Invented tickers. A book allocated to a symbol nobody prices never moves. Allocations are checked against the tradeable universe.
- Never setting a book. A seat with no allocation is not scored as flat — it ranks below every seat that played, including one that lost money, and the cut reaches it first. Buying a seat is not playing it.
- Sitting out a bad week. A seat that buys back in rejoins at the median season score of the survivors, not at the score it left with. Otherwise being knocked out before a down week and returning after it would beat staying in and taking the loss.
- Cash and near-cash. BND, SGOV, SHY are excluded. Over the last 52 weeks the field's median weekly volatility was 7.93%; SGOV's was 0.012%, with a worst week of +0.05% and not one down week in the year. Holding T-bills does not reduce risk here, it removes the cut — and this is a stock-picking tournament.
What remains, and we would rather say it: low-beta index exposure is a strong strategy under these rules. SPY moved 1.70% a week against the field's 7.93%, and had a down week 40% of the time. That is real risk and a real trade-off, so it stays — but it is a genuine property of ranking on raw percentage growth, not an oversight.
Where the data comes from
- Universe. 182 tokens. Not a hand-picked list: it is the set of Robinhood Chain stock tokens priced by Robinhood's own public feed, less the stablecoin, less the cash-equivalents named above, less anything halted.
- Prices. Read from Robinhood every fifteen minutes and adjusted for corporate actions via each token's multiplier — the raw feed quotes the underlying, so an unadjusted price would misprice any token that has split.
- ETH pricing. Fees are quoted in dollars and charged in ETH at the rate read from Chainlink at the moment you pay, never from a stored constant.
About the board below the fold
Until the first Friday close there are no real results to show, so the board, the tape and the simulator render a backtest: 2,222 generated portfolios run over 41 trading days of real daily closes, 2026-06-24 → 2026-08-20. It is labelled as a simulation everywhere it appears and it is never blended with live standings — a table mixing simulated seats with real ones could not be checked against anything. One year is also one regime: a window containing a memory-stock melt-up is not evidence the design survives a flat or falling market.
The backtest charges no fees, no slippage and no spread. Real rebalancing costs money and this pays none of it.
What this is not
It is not an investment, a fund, or a security, and nothing here is financial advice or a solicitation. You are not buying exposure to any stock: the book is paper, and the seat entitles you to compete, not to any underlying asset. Prizes are funded by fees, and the team's share is fixed in the contract at 10% and cannot be raised after a season opens.
Ranking is computed and declared by the operator, not by a contract. Three things bound what that buys us: results are public the moment they are declared, the inputs are public so anyone can recompute them, and a timelock sits between declaring and paying so a false result can be seen before a cent moves. That is an honesty model, not a trustless one, and you should weigh it as such.
The seat contract is 0x9b30760f17CD2B989Db2c6128e3379e961B11852 on Robinhood Chain Testnet. See the desk for what a seat costs and what it does.
Backtest artifacts built Thu, 20 Aug 2026 23:48:08 GMT.