Lot-size classification is the practice of sorting executed trades in the futures market into size buckets — small, medium, large, and whale tiers — based on the observed quantity of contracts per fill, and analyzing each bucket's cumulative volume delta (CVD) separately, without attributing any specific order to a specific type of market participant. It answers a question that can actually be measured: *what did the tape print?* Not a question that, in practice, cannot: *who printed it?*
What Is CVD by Lot Size?
Cumulative volume delta (CVD) tracks the running difference between aggressively-executed buying volume and selling volume. On CME's Globex platform, gold futures (GC) trade under a Market-By-Order model — every individual resting order in the book is visible as a distinct liquidity source, and every match is a discrete execution event. Standard CVD lumps all of that into a single net-flow number.
CVD by lot size splits it. A 1-lot market buy and a 400-lot market buy both move the price the same way, but they carry very different information about conviction and intent. By bucketing fills into tiers — small, medium, large, and whale — ATLAS Sovereign builds a separate delta curve for each tier, then reads the relationships *between* those curves: when whale-tier buying persists while small-tier flow flips negative, when large-tier flow dries up ahead of a breakout, when mid-size flow absorbs a sell impulse without price giving ground.
Why Not Label Participants Directly?
This is the part skeptical readers are right to press on. Many order-flow products advertise "institutional footprint" or "smart money detection" — language that attributes detected patterns to specific *types* of traders. The problem is simple: exchange data does not identify who placed an order. An execution tells you the instrument, the price, the size, and the aggressor side. It does not tell you whether the initiating account was a hedge fund, a trading desk, a systematic strategy, or an individual.
Any product that claims to identify *who* is trading is layering an unverifiable assumption on top of the data. ATLAS Sovereign doesn't do that, for three reasons:
1. Size is observable; identity is inference
Lot size is a raw field on every fill. Trader identity is a guess. A 300-lot order could come from many kinds of accounts; a single algorithm can slice one large parent order into hundreds of small child fills that land in the small tier. Classifying by size keeps every claim we publish anchored to something independently checkable on the tape.
2. Sliced orders punish identity-based reading
Order-splitting algorithms exist precisely because large participants don't want their size visible in one place. The result: "small lots = retail" is simply false a meaningful percentage of the time. Size-tier CVD sidesteps this by never needing the assumption in the first place — we track the *behavior of liquidity at each size band*, not the biography of who might be behind it.
3. It keeps our signals auditable
Every ATLAS signal traces back to observable execution data: price, size tier, aggressor side, and the resulting delta structure. If we publish a note saying whale-tier lot-size activity was persistently bid-side into a session low, that is a statement about the tape that anyone with market data can verify. That is the standard we hold ourselves to.
Why Lot-Size Tiers Matter for Gold Futures Specifically
Gold is one of the most heavily traded futures markets in the world. On CME Group's GC contract specifications, each contract represents 100 troy ounces, the minimum tick is $0.10 per ounce — $10 per tick per contract — and the market trades nearly 24 hours a day. That tick value means size-tier differences are economically meaningful even at modest lot counts: a large-tier fill of 200 lots represents $2,000 of value per tick of movement.
For traders migrating from spot XAUUSD to GC futures, the size dimension is the biggest structural difference to internalize. Spot feeds generally show aggregated volume; futures Market-By-Order data shows discrete, size-classifiable executions. Reading gold through tiered CVD rather than a single blended delta curve can offer a more granular analytical perspective — additional detail that a blended number simply does not surface, and a lens you can evaluate for yourself against your own data.
What This Looks Like in Practice
In ATLAS Sovereign's published recaps, you'll see language like "large-tier lot-size flow rotated to the bid side while small-tier flow sold into the move" — a description of observed activity, stated plainly, that a reader can go verify against their own data feed. No participant labels. No narrative about what a mysterious class of traders is "doing." Just the structure of the tape, classified honestly.
The full technical breakdown — how we define the tiers, how tiered CVD interacts with our five-pillar confluence framework, and live worked examples from recent GC sessions — is available to Sniper and Sovereign subscribers in the ATLAS Academy.
If you want to see how size-tiered order flow reads on live gold sessions — with every claim tied to observable execution data — ATLAS Sovereign offers a 14-day free trial. It's the fastest way to judge the methodology on the tape rather than on a promise.