Citadel, Jane Street, Hudson River, Susquehanna, Two Sigma and Virtu filled half their retail orders close to mid-price, while for the entire market 28% of trades were executed worse than NBBO during a volatile period for equities.
US retail equity investors were rewarded handsomely for the predictability their flows provided to top market makers during the tariff turmoil of April, according to analysis of Rule 605 filings. The $666 million price improvement versus national best bid offer (NBBO) compares with far worse execution for the market as a whole.
In a new data visualisation, Global Trading examines the distribution of execution quality by Citadel Securities, Jane Street, Hudson River Trading, Susquehanna through its G1X subsidiary, Two Sigma Securities, and Virtu, comparing their effective-over-quoted (EFQ) spread distributions to the overall market.
Out of the six wholesalers, Susquehanna/G1X presents the most competitive execution with nearly 40% below 0.25 EFQ indicating that 40% of the trades they fill for retail are within a quarter of the best bid or offer to mid-price.
Jane Street fills were 55% of the time above 0.5 EFQ, Two Sigma securities fills were 52% above 0.5, Citadel Securities the largest retail market maker had a normal looking distribution of its EFQ with its median EFQ at 0.43. Virtu’s EFQ were spread across 0.2 to 0.8 EFQ with its median at 0.5 while Hudson River trading EF were concentrated between 0.2 and 0.5 EFQ.
Expressed in dollars, Citadel provided the greatest price improvement of $251 million during April, followed by Virtu and Jane Street. The total $666 million improvement is more than double the monthly price improvement provided by the same six firms over the previous year, demonstrating the importance of retail flow during volatile market conditions.
Read more: US market makers improved retail equity pricing by $3.2bn compared with exchanges
Data source: S3, BMLL
A more surprising finding of our analysis related to the EFQ was that for the entire US market (defined as all trades reported through the Security Information Processor and the associated Consolidated Tape) a proportion of more than 15%, when share weighted, or more than 27% when notional weighted, of trades happen outside of the NBBO.
According to Andriy Shkilko, professor of finance at Wilfrid Laurier University in Canada, who has researched execution quality, “Large trades executing on dark pools, for example, or crossing networks… sometimes execute at prices that are not exactly the NBBO.”
Our analysis specifically focuses on price improvement, assessed through the EFQ ratio, which compares the executed price to the quoted price at the order placement time. The EFQ ratio is computed as twice the difference between the executed price and the concurrent mid-price (the midpoint between the National Best Bid and Offer, NBBO), divided by the prevailing NBBO spread. Under best execution rules, covered orders must execute within the NBBO. Thus, an EFQ of 0 indicates execution precisely at mid-price, a ratio of 0.5 is at the midpoint between the mid price and the prevailing bid or offer. A ratio of one corresponds to executing on the bid or offer.
Another relevant measure within Rule 605 disclosures is the realised spread, which captures price deviation five minutes post-execution. However, this current approach to measuring realized spreads is insufficient for accurately assessing high-frequency execution quality. To address this, substantial changes to Rule 605 disclosures are expected by the end of 2025, introducing more granular metrics measured at intervals of fifty milliseconds, one second, fifteen seconds, one minute, and five minutes.
The plots above compare the EFQ distributions of prominent US retail wholesalers with corresponding stocks reported through the Securities Information Processor (SIP). SIP-listed securities are consolidated through the SIP, which provides standardised NBBO data.
For an accurate comparative analysis, our dataset filters out SIP trades to include only continuous trading hours, specifically trades where the bid-ask spread was not crossed and both bid and offer sizes exceeded one lot. We also only considered trades without any specific flags, odd lot trades, bunched sold trades and bunched trades.
Additionally, we matched trades for wholesalers and the SIP in terms of comparable size and stock selection.
For the all-market comparison, EFQ ratios are computed relative to the NBBO prevailing up to 25 milliseconds before the recorded execution timestamp. Our analysis presents two types of distribution weighting: Share-weighted EFQ distribution: EFQ ratios for each trade are weighted by the number of shares traded, Notional-weighted EFQ distribution: EFQ ratios are weighted using the monthly Volume Weighted Average Price (VWAP) assigned to each stock and calculated with all the trades of the month reported on the SIP.
