The first two lanes act on the same traders' moves and started with $10,000. The US shadow lane takes only what a US account could trade on Coinbase, at Coinbase's prices and fees. It also uses less borrowed money per bet (what Coinbase allows overnight; none on spot), so its swings are smaller by design. The lanes started at different times: compare them from the shadow lane's start. Prices are as of each lane's last poll.
The order-book lane copies no one: it started with the same money and bet size, and trades BTC and ETH futures on Coinbase on its own reading of the live order book, in whole contracts, at Coinbase's futures fees. Its prices are as of its last few minutes.
The plain wheel lanes run the same rules on the same stock list and differ only in starting cash ($50,000 and $100,000), so the gap between them is the effect of size: a put needs the whole strike in cash, so the smaller lane reaches fewer names. The Wheel $100k quality lane is the $100,000 wheel with quality checks before every put (a safe dividend in the filings, no earnings before expiry, smaller where the picture is mixed), so it enters in bursts between earnings seasons: the Wheel $100k lane is its control, and the gap between those two is what the checks are worth after costs. They trade only while the New York market is open, on Alpaca's data; nothing is ordered anywhere.
The Wall St desk reads the Fed's, Goldman Sachs', JPMorgan's and BlackRock's own publications, turns them into a lean per ticker, checks each lean against a week of news scored by a decision model and the price trend, has a second model approve it, and buys US stocks and ETFs, long only, on Alpaca's prices; nothing is ordered anywhere.