Understanding Market Structure

Understanding Market Structure

Markets are a tape — buyers and sellers leaving footprints in time. Before you read flow, before you read regime, you must understand what those footprints are. This chapter teaches you how the tape is built, who’s on it, and why the structure of the market is the most under-appreciated edge in retail trading.

1.1.1What “market structure” really means

When a beginner watches SPY tick from $723.71 to $723.72, it looks like one place doing one thing. It is not. That one-cent uptick is the visible surface of an enormous, fragmented plumbing system: 16+ lit exchanges, roughly 30+ alternative trading systems (ATSs / dark pools), dozens of single-dealer wholesalers, and a securities information processor (SIP) trying to stitch it all into a single quote. Market structure is the rules and routing of that plumbing.

Every print you see has three properties most retail traders never think about: where it executed (which venue), how it executed (which order type), and who was on each side (a real institution, a wholesaler internalizing your order, or another retail account). Lose track of those three and the tape becomes noise. Track them and you start to see why a stock holds $155.40 for forty minutes — because that’s where the size sits, not because of a chart pattern.

“Price is a vote. Volume is a count. Venue is a confession.” — the unspoken rule of post-Reg NMS markets.

This lesson is the tour. By the end you will know the names, the rough volume share, the incentives, and the places where retail trading frameworks (including SomerQuant’s) get their edge or get trapped.

1.1.2The 16+ US equity exchanges and the SIP

US equities trade either on one of the registered national securities exchanges or off-exchange — through an ATS/dark pool or a broker-dealer internalizer — with every off-exchange trade reported to a FINRA Trade Reporting Facility. Reg NMS (adopted 2005) sets the quoting, access and order-protection rules that knit these venues together. There are currently 16 active equity exchanges, run by 4 holding companies plus a handful of independents:

The remainder — roughly 40-45% — trades off-exchange (dark pools and wholesaler internalization). Each exchange runs its own matching engine, fee schedule (maker-taker, taker-maker, or flat), and order types. A market order for 100 shares of NVDA at your broker may be sliced across 5 different venues in under a millisecond by the broker’s smart order router (SOR), each chasing a different rebate or price improvement.

The SIP (Securities Information Processor — CTA Plan for NYSE-listed, UTP for Nasdaq-listed) consolidates the best bid and offer from every exchange into the NBBO — the “official” national best bid and offer. This is the quote your retail platform shows. It is also slower than the proprietary direct feeds the wholesalers and HFTs run, by anywhere from 200 microseconds to a few milliseconds depending on geography. That latency gap is itself an edge that retail simply cannot close — which is why we don’t try to compete on speed in the SomerQuant framework. We compete on context.

1.1.3Dark pools and ATSs — the hidden 40%

An ATS (Alternative Trading System) is a non-exchange venue that matches orders — legally, under SEC Reg ATS. The largest are operator-owned (UBS ATS, J.P. Morgan’s JPM-X, Goldman’s Sigma X2) or independent (Intelligent Cross, Liquidnet, Level ATS). (Credit Suisse’s Crossfinder was absorbed after the 2023 UBS takeover.) A “dark pool” is just an ATS that does not display its order book pre-trade. Trades print to FINRA’s TRF (Trade Reporting Facility) after they execute, usually with a small delay, and that’s how you see them on the tape with venue code “D” or “FINRA ADF”.

Why do they exist? A pension fund that wants to buy 2 million shares of NVDA cannot post that on Nasdaq without moving the price 2-3% against itself. So they slice it through dark pools where other large naturals (other pensions, sovereign wealth funds) can match without signaling intent. Roughly 40–45% of US equity volume executes off-exchange — the larger share is wholesaler internalization (roughly 25–35% of total volume), with ATSs/dark pools around 10–15% (FINRA OTC transparency data).

This is where SomerQuant’s dark-pool flow tools come in. When you see a 485,000-share dark print in PLTR at $92.40 while the lit market is sleepy at $92.38, you may be seeing an institution accumulate without lifting the offer — but every print has both a buyer and a seller, so direction must be inferred from where prints sit relative to the spread and how price behaves afterward. A pattern of dark prints stacking near a level is a useful, but inferential, accumulation clue. Lesson 4.5 covers this in depth.

1.1.4Market makers, wholesalers and HFTs

Three big retail wholesalers handle the overwhelming majority of US retail equity flow: Citadel Securities (~35-40% of retail), Virtu Financial (~20%), and Jane Street, G1X, and Two Sigma Securities splitting most of the rest. When you click “buy” in Robinhood, Schwab, Webull, or E*TRADE, your order does not go to an exchange. It is sold via payment for order flow (PFOF) to one of these wholesalers, who internalizes the trade against their own book at or slightly inside the NBBO — offering “price improvement” of typically $0.0003 to $0.002 per share.

The economics: the wholesaler pays the broker (often $0.0015-$0.0030 per share for equities, much more for options), captures the bid-ask spread on aggregate, and avoids the adverse selection of trading against informed institutional flow. Retail order flow is uninformed on average — that’s why it’s profitable to internalize. Whether that makes retail volume itself a contrarian signal at extremes is more contested: some academic work finds one-sided retail imbalances positively predict short-horizon returns, so a flood of one-sided retail buys is context to weigh, not a standalone fade signal.

Distinct from wholesalers, HFTs (Hudson River Trading, Jump, Tower, XR Trading, Headlands) provide liquidity on lit exchanges, capturing the maker rebate ($0.0029/share on Nasdaq’s flagship tier) by posting limit orders inside the spread. They are the reason your SPY bid-ask is $0.01 instead of $0.05. They are not the enemy; they are the plumbing. But they will pull liquidity in milliseconds when the flow turns informed — which is why a $723.71 bid that looked rock-solid can vanish on a hot CPI print.

1.1.5Order types and the order book

The exchange order book is a sorted list of resting limit orders — bids sorted descending (highest first), offers sorted ascending (lowest first) — with the top of book being the highest bid and lowest offer (the BBO for that venue). The NBBO is the best of those across all 16 exchanges. Depth is everything else underneath. Here are the order types that matter:

• Market order — take whatever liquidity is at the NBBO; in a fast market this can slip dramatically. Almost always wrong for retail unless you genuinely don’t care about the next $0.05.
• Limit order — rest at or inside the spread; if at NBBO, you may earn a maker rebate (if your broker passes it through, which most retail brokers do not).
• Stop / stop-limit — conditional triggers held at the broker (not the exchange) until the trigger price prints, then released as a market or limit. Stop orders are visible to no one, but the cluster of stop levels at common chart points is highly predictable to anyone watching.
• Hidden / iceberg — institutional limit order with most of the size hidden from the displayed book. Common at round numbers and key VWAP levels.
• Mid-point peg — rests at the midpoint of the NBBO, where most dark-pool matching occurs.

For SomerQuant students, the key insight is that the displayed top of book is a tiny slice of the real liquidity. A SPY bid showing 12,000 shares may have 200,000+ hidden behind it across IEX (with its 350-microsecond speed bump), midpoint pegs, and iceberg orders. That’s why “tape reading” via Level 2 alone is incomplete. You need the prints, the dark data, and ideally GEX context to know what the book really is.

1.1.6The opening and closing auctions

Two of the largest single liquidity events of every trading day happen at 9:30 ET (open) and 4:00 ET (close). The exchanges hold formal auctions that match accumulated buy and sell interest at a single clearing price.

• Opening auction (9:30 ET) — primary listing exchange (NYSE for NYSE-listed, Nasdaq for Nasdaq-listed) collects MOO (market-on-open) and LOO (limit-on-open) orders from 7:30 ET onward. Imbalance feeds publish from 9:28. Opening crosses in large single stocks can run into the millions of shares; for ETFs like SPY the opening auction is far smaller — often under a million shares — with the close the day’s dominant auction.
• Closing auction (4:00 ET) — far larger. The MOC / LOC / IO (imbalance only) book builds all afternoon, with imbalance feeds publishing from 3:50 ET. The closing cross routinely accounts for 10-15% of daily volume — on rebalance days (quarter-end, MSCI / Russell rebal), this can spike to 25%+. The S&P 500 closing print at 4:00:01 ET is the most-watched price in finance.

The takeaway for retail: if you trade at the open or close without watching the imbalance feed, you are flying blind during the windows where institutional intent is most visible. SomerQuant’s premarket and end-of-day routines (Lessons 1.6 and 4.x) build directly on auction-imbalance reading.

1.1.7Why retail mis-reads the tape — and how SomerQuant’s framework responds

Most retail trading content treats the tape as a single transparent place. It isn’t. Here are the four most common ways retail gets fooled, and how the SomerQuant framework defends against each:

1. Confusing absence of lit volume for absence of activity. A “quiet” Level 2 in NVDA may mask 800k shares being accumulated in dark pools at $182.30. Defense: track dark-pool prints alongside lit data (Lesson 4.5).
2. Treating the SIP NBBO as the real market. Wholesalers see the real market 200µs earlier. Defense: never use market orders on retail platforms during volatile windows; always trade inside the spread when you can.
3. Mis-reading retail-heavy stocks. When 95% of PLTR volume one afternoon is retail flow internalized by Citadel, that “tape” is largely the wholesaler’s inventory rebalancing — not informed money. Defense: weight institutional signals (dark prints, options flow, MPI) more heavily than tape pace.
4. Single-source confirmation. Trading off one indicator — one chart pattern, one news headline, one Twitter call. Defense: multi-source confirmation — the SomerQuant framework requires alignment across at least three of {dark-pool flow, options flow, MPI / market regime, technical level, sector breadth}. When SPY is at $723.71 with MPI 78 (risk-on), bullish dark prints in financials, and call flow building in XLF, that’s three sources aligned. One indicator is a guess; three sources is a setup.

Common mistakes

• Using market orders on small-caps. A 100-share market in a $15 stock with a $0.04 spread is a 25 bp giveaway before the trade even happens.
• Setting stops at obvious round numbers. If your stop is at $100.00 exactly, so are 40,000 other retail stops. Use $99.78 instead, or use a time-based exit.
• Trading the first 5 minutes blind. The 9:30 auction print is the consensus opening price; the next 5 minutes is wholesalers and HFTs digesting overnight inventory. Wait for 9:35.
• Assuming a “halt” or “freeze” means the market broke. Limit-up/limit-down (LULD) bands trigger 5-min pauses on outsized moves; this is a feature, not a bug. Use the pause to re-anchor.
• Reading PFOF as “your broker is stealing from you.” The math is more nuanced — price improvement is real on most retail-sized fills. The bigger issue is that PFOF removes the incentive to route to the venue with the best fill (not just best displayed price). For sub-1,000-share retail orders the difference is usually tiny; for active traders it adds up.