How the numbers are measured
From the reports brokers must publish, refreshed weekly; last updated November 2, 2026. Prices: September 2026 (Rule 605). Payments: Apr–Jun 2026 (Rule 606).
Where the numbers come from
Rule 605: brokers with 100,000+ accounts publish monthly how their customers' orders were filled; the trading firms that fill them file the same report. Rule 606: brokers publish quarterly where orders go and what they are paid (payment for order flow). FINRA lists every 605 report and holds each broker's BrokerCheck record.
What “share of the spread” means
The gap between the best buy and sell quotes is the spread; halfway across it is the fair price. With quotes of $99.97 and $100.03, fair is $100.00; a buy at $100.01 is a cent above fair, a third of the way to $100.03, so it paid 33% of the spread ($1 on 100 shares).
“Extra cost per $10,000 traded” is that share in dollars, at the spreads each broker's customers faced.
Same stocks, same order sizes
Averages flatter brokers whose customers trade big, liquid stocks in large orders.
So each broker is also scored against the others on market orders in the same stock, lot type and order size: the “same-trade gap”, below zero = better on identical trades. The gaps persist with the same trading firms filling the orders: they price each broker's flow differently.
Common questions
Is my broker registered?
Every broker here publishes the Rule 605 price report and all but one are matched to a FINRA BrokerCheck record; each broker's page links it and says whether we found SIPC membership (up to $500,000 of securities and cash if the firm fails).
Registration is not an endorsement: SIPC does not cover a poor price. Pick a broker to see its record.
What is payment for order flow?
Most brokers send your stock order to a trading firm that fills it and pays the broker a fraction of a cent per share; Rule 606 makes those payments public.
The worry is that a broker paid more lets you keep less of the spread. Here that link does not hold up (see below); Schwarz, Barber, Huang, Jorion and Odean (Journal of Finance, 2025) reached the same answer from 85,000 real trades. The explorer shows both sides.
Does $0 commission mean free?
No: your order fills somewhere between the best buy and sell quote, and that gap is a real cost that never shows on a statement. Brokers differ by more on it than on any fee.
The home page shows it in dollars; the calculator adds the fees you do see.
Why does Fidelity look best?
Fidelity's customers paid the least above the fair price per $10,000 of S&P 500 market orders, and it holds on identical trades (−15.2 points vs other brokers). It is also paid less than most brokers.
Low payment alone doesn't explain it: Cash App actually pays its trading firm and Wealthfront is paid nothing, and both get worse prices than most on identical trades (Cash App +45.7 points, Wealthfront +27.4 points). The strongest pattern is clientele: brokers whose customers place bigger orders get better prices even at the same order size, and the same trading firms price each broker's orders differently.
Which broker is best?
It depends on what and how much you trade: with a few trades a year, idle-cash interest or a transfer-out fee matters more than a dollar per trade, and options aren't in the price reports at all.
Put your habits into the calculator, then compare your two finalists. It's an estimate from averages, not a bill.
Key definitions
- Broker is paid (% spread)
- What trading firms pay the broker (Rule 606, weighted by where its orders go), as a share of the quoted spread. Check: Robinhood's own report says 12.35%; ours gives 12.1–13.5%.
- Average execution speed
- Rule 605's average time from order to fill, in milliseconds (the unit the SEC's report specification sets). Shown, never scored: here slower fills went with better prices.
What the data cannot show
- Options, limit orders, your own trades. Rule 605 covers stocks only (though about 69% of what brokers are paid is for options) and only orders that fill at once; everything is an average, so your own trade can land better or worse; taxes, promotions and paid tiers are left out.
- Some payments. Albert, Avenue, eToro, WF FiNet, Stash and Coinbase file no payment report of their own (another firm routes their orders). SoFi reports only part of its orders. moomoo's is reported by Futu Clearing.
- Few brokers. Only 19 brokers publish both sides (TradeStation, Ally and others aren't in yet), so small differences can't be told from chance; a pattern is not proof of cause.
Fees, and when they were checked
Each broker's published schedule for a standard self-directed account, last checked November 2, 2026; re-checked every quarter.
An unconfirmed fee is left blank and out of every total, never counted as $0; promotions and paid tiers are left out, so a broker can be cheaper for you than shown.
Does getting paid make fills worse?
Not in any way the data can confirm. For S&P 500 stocks, brokers paid more do not reliably give worse prices. For other stocks, a link appears but fades once share price is held equal: cheaper shares, not payment, are the likelier reason. Who the customers are predicts the price far better than payment.
The statistics
Strength from −1 to 1; a line crossing zero could be chance.
The statistics
Strength from −1 to 1; a line crossing zero could be chance.
| Claim | Strength (−1 to 1) | Chance it's random | Brokers | Verdict |
|---|---|---|---|---|
| Paid more → customers pay more of the spread | +0.40 | Could be chance | ||
| Paid more → customers pay more, at equal share price | +0.17 | Could be chance | ||
| Paid more → worse prices on the same stocks | +0.37 | Could be chance | ||
| Bigger typical order (clientele) → better prices on the same stocks | −0.65 | Holds up | ||
| Slower fills → better prices on the same stocks | −0.69 | Holds up | ||
| Routing mix predicts a broker's same-stock prices | +0.02 | Could be chance | ||
| More orders price-improved → customers pay less | −0.21 | Could be chance |
Spearman rank correlation; permutation p-value (10,000 shuffles); 95% bootstrap interval (3,000 resamples); 5% or more = “could be chance”. With 19 brokers, every test has low power: correlations, not causes.