pt. 2: execution quality and routing bugs here is a filtered sample of trades over the last week that are likely to be retail and not bots the table below shows average fill quality if you get routed to an active liquidity dex (dark amm or clob) vs. passive based on trade size ($)

the first thing to notice is that if you get unlucky enough to get routed to a passive AMM your fill is likely to be about 10x worse it turns out this is primarily driven by the original raydium v4 amm that gets far more flow than it should based on the fill quality

here is an example to illustrate. a user buys on jupiter and it gets simultaneously routed through raydium v4, raydium clmm, and solfi the fill is almost 20 bps worse on raydium v4 than the remaning two dexes tx:

its unclear to me why this is happening. possibly the sim is just incorrect. or maybe the raydium dex looks good for a split second but is arbed out far before the user tx lands there is a lot of room to improve user execution quality just by routing fills to the best dexes