Making Markets While the Market Sleeps: Automated Market Making for Tokenized Stocks When the Reference Market Is Closed

Replication package for the article "Making Markets While the Market Sleeps: Automated Market Making for Tokenized Stocks When the Reference Market Is Closed" (Zhengdong Zhu; submitted to FinTech, MDPI). The archive rwa-closure-study_replication_2026-09-26.zip (487 files) reproduces every table, figure, and in-text number of the paper from the data of record. It contains: the Python analysis scripts and the SQL of the 17 Dune Analytics queries used for the on-chain data; the input data as retrieved (Dune query results, one-minute bars from Binance, MEXC, and KuCoin, Hyperliquid candles, daily prices of the underlyings, ex-dividend dates, and the fee rates and fee shares read from the Solana pool accounts with revision/read_pool_fees.py); all generated outputs, tables, and figures; the LaTeX sources with the compiled manuscript and its Word version; the response letters of the review rounds; and REPLICATION.md, which documents data provenance, computational requirements, the order in which to run the scripts, and an end-to-end replication check (REPLICATION_TIMING.txt). The scripts are deterministic (fixed bootstrap seeds). The accompanying manifest lists every file in the archive with its MD5 checksum. The Dune API key is read from an environment variable and is not included. Abstract of the paper. Tokenized stocks trade around the clock, but the stock market that anchors their prices does not. Using 15 months of on-chain xStocks trades, centralized-exchange minute data, and 30-day to eight-week accounts of the main automated market maker pools ending in September 2026, we study trading outside the regular session, when most of it takes place. Token prices absorb the information revealed at the next open in three steps, the first before any U.S. venue trades. In the weekend dead zone, when no U.S. price exists, token prices move almost twice as much as the information in order flow warrants and partly reverse. The overreaction survives instruments built from prices in other hours and is absent when the stock can be hedged. A model combining informed trading, arbitrageurs' inventory risk, and loss-versus-rebalancing explains these patterns. Liquidity providers earn at least as much per dollar while the market is closed as in the regular session, mostly on coins routed through the tokens' pools, and lose when it reopens, to the largest traders. Halting pools while the listing exchange is closed would forgo that profit and, unless pools resume at the reference price, charge liquidity providers the halted hours' variance at the reopening.