Robinhood Chain
ROBINHOOD CHAIN · ID 4663 · 10 BLOCKS / SECOND

THE EDGE IS REAL.
IT DIES IN
ONE SECOND.

One wallet on the Robinhood Chain buys launchpad tokens and is right often enough to matter: 42% of its entries double. Six benches read the chain, scored 147 of them, and brute-forced 852,096 exit rules over the price paths. Then the same desk measured what copying it actually costs — and the answer is the whole story.

▸ SEE THE DELAY CLIFF OPEN THE LIVE DESK GITHUB @fluixoo
BLOCK —· — BLK/S MEASURED· GAS —· ETH/USDG —
P(X2) PER ENTRY SNAPSHOT 0.00147 SCORED ENTRIES
BEST RULE · EV+0.0% 852,096 COMBOS TESTED
COPY AT +1s+0.0% HALF THE EDGE, GONE
CHAIN LIVE —ROBINHOOD · ID 4663
01

THE LIVE DESK

Six benches, one operator log, real RPC calls leaving your browser. The block stream and the price are live; every computed number carries the timestamp it was computed at.

hexley ~ /desk/robinhood NEW TAB ↗
BOOTING THE DESK
6 BENCHES · LIVE RPC · 10 BLOCKS / SECOND
CLICK inside the frame to interact FULLSCREEN — it is built for 1920×1080 ESC to come back
02

THE DELAY CLIFF

The same rule, the same entries — only the delay between the leader's print and yours changes. This is the finding the rest of the desk exists to support.

EV OF RULE A BY ENTRY DELAY SNAPSHOT
Rule A — take profit at ×4, time-stop at 15 minutes. Entry is the first print at least N seconds after the leader's. Fills modelled at −4% on take-profit, −6% on stop, 2% fees. At 0.1 s the rule pays +29.0%; one second later it pays +15.3%; a minute later it loses money.
MEASURED COPY LAG09-05
leader — 3 wallets, same block0 blk
GMGN copy · 0xd11C…+1 / +4 / +1
Telegram bot copy · user+5 / +7 / +10
Measured on the same three tokens: UNCLES, CORGI, DOGE. At 100 ms per block a ten-block lag is one second — exactly the point where the cliff has already taken half the edge.
WHAT THAT MEANS
The edge is not in knowing what he buys. It is in being in the same block. A copier that routes through a bot pays a tax measured in blocks, and on a chain that produces ten blocks a second that tax is the entire expected value.

Verdict from the clerk bench: fix routing before sizing up.
03

SIX BENCHES

Each bench does one job and hands its output to the next. Click one to read what it actually does.

BENCH 01

TAP

Reads the chain.

04

THE SCORECARD

147 entries scored from the leader's own fill price, over 156 tokens and 588 transfers, between 2026-08-29 and 2026-09-05.

P(×1.5)0.5986 OF 147
P(×2)0.4262 OF 147
P(×3)0.2740 OF 147
P(×10)0.046 OF 147
MAX MULTIPLE AFTER HIS BUYn=147
The spike is real and it does not last: the median entry peaks at 1.76×, then trades down to 0.65× six hours later. The leader himself realises 0.89× — he wins on the tail, not on the median. An exit rule is not optional here, it is the entire strategy.
TIMING & HONESTY
MEDIAN TIME TO ×20.9 min
MEDIAN MAX1.76×
MEDIAN LOW WITHIN 1H0.68×
MEDIAN CLOSE AT 6H0.65×
P(×2) IF IT GRADUATED0.88
P(×2) IF IT DIED0.33
HIS OWN WIN RATE32%
HIS MEDIAN REALIZED0.89×
05

852,096 EXIT RULES

Every ladder × stop × time-stop × trailing combination replayed over the recorded price paths, in 80 seconds. Ranked three ways, because ranking by expected value alone rewards fragile rules.

LEADERBOARD COMPUTED 09-05 23:40 UTC
RANKRULEEVMEDIANWINPFMAX DDPICKED FOR
EV is per entry, after fills and fees. PF — profit factor, gross win over gross loss. Rule A (take profit ×4, time-stop 15 min) bootstraps to a 90% interval of +14.6% … +45.4%: positive across the interval, but wide — 147 entries is a small sample and the desk says so.
DOES A STOP HELP?
Best EV at every stop level, over the whole grid. A hard stop never beats no stop here — it cuts the tail that pays for the losers. That is a property of this leader's path shape, not general advice.
06

THE CHAIN

Everything below is read from the Robinhood Chain in your browser as you look at it. No backend, no cache — if a number moves while you read, that is the chain moving.

ROBINHOOD CHAIN
An Arbitrum Orbit L2 that settles to Ethereum. It produces about ten blocks a second, prices gas in ETH and quotes its launchpad in USDG. That block time is the reason the delay cliff above is so steep: a copier “a few blocks behind” is already a fraction of a second late, and a second is half the edge.
CONNECTING … CHAIN ID — RPC rpc.mainnet.chain.robinhood.com UPDATED —
BLOCK HEIGHT— SAMPLED ONCE PER SECOND
BLOCK TIME— — BLOCKS / SECOND
TRANSACTIONS / S— — PER BLOCK, MEASURED
LOGS / BLOCK— THE DESK'S RAW MATERIAL
GAS PRICE— eth_gasPrice
ETH / USDG— POOL 0x52e6… slot0
LEADER BALANCE— 0x5fa5…4cfb
LEADER NONCE— TX COUNT, ALL TIME
BLOCK STREAMLIVE
One sample per second, not every block — at ten blocks a second the full stream would be noise on a page like this. The desk reads every block.
RAW LOGS eth_getLogs
Straight out of the sampled block, undecoded except for the topic name. This is the material every bench upstream works from — no indexer sits between the chain and this list.
WHAT THE DESK WATCHES CLICK A ROW TO READ IT
Every address the pipeline touches, and nothing else. Click one and the inspector below reads it live off the chain.
07

ADDRESS INSPECTOR

The genealogist bench's first question about any address, runnable by you: is it a person, a contract, or a wallet that delegated itself to code under EIP-7702?

READ AN ADDRESSIDLE
TRY:
BALANCE—
NONCE · TX SENT—
CODE SIZE—
READ AT BLOCK—
WHAT IT IS
—
paste an address
An account with no code is a person holding a key. Code means a contract. Code that starts with 0xef0100 is an EIP-7702 delegation — an ordinary wallet that pointed itself at a program, which is how several of the launch wallets in cluster #1 are set up.
08

METHOD & LIMITS

Which numbers here are live and which are computed once?
Anything with a LIVE badge is read from the chain as you look at it: block height, block time, gas, the pool price, the leader's balance and nonce, and everything the inspector returns. Anything with a SNAPSHOT badge was computed by a script over a fixed window and carries the timestamp it was computed at. The desk never re-labels one as the other.
How were the entries scored?
The tap bench pulls the leader's transfers by topic filter over the window; the decoder reads the receipts and separates a launchpad curve buy from a Uniswap v4 swap, using the sign of the quote word to tell a buy from a sell. Each entry is then priced from his fill, not from the token's open, and followed forward on candles.
Why is 147 entries not enough?
Because the payoff lives in the tail. Six entries out of 147 reached ×10, and those six carry the expected value of every rule on the leaderboard. Drop them and the table changes. That is why rule A's bootstrap interval is +14.6% to +45.4% rather than a single number, and why the desk ranks by robustness and profit factor next to raw EV.
Is this a copy-trading signal?
No. It is the opposite conclusion. The measurement says the leader's edge is real and that almost all of it is consumed by the delay between his print and a copier's. Before anyone sizes up, the thing to fix is routing — and if the routing cannot be fixed, the honest read is that there is nothing left to copy.
Can I read the code?
Yes — the desk is open source at github.com/fluixoo/hexley. The Python package chaindesk holds the six agents, the RPC layer and the decoder, with 27 offline tests. The dashboard in this page is the same desk rendered from that pipeline's snapshot.
What is the chain?
The Robinhood Chain, id 4663 — an Arbitrum Orbit style L2 that produces about ten blocks a second. The short block time is exactly why the delay cliff is so steep: a copier that is "only a few blocks behind" is already hundreds of milliseconds late.