Does CVD by Trade Size Tell You Where NQ Is Headed?

NQ Futures  ·  last 12 months of data (May 2025 – Apr 2026)  ·  249 trading days, every one starting 08:30 ET (1hr warmup before the 9:30 RTH open)  ·  every size from 1 to 500+ contracts

• Split every NQ trade over the last 12 months into size groups.
• Tested one question per group: does its flow predict price, confirm it, or tell you nothing?
• Read the takeaways box below and you have the whole report — everything after it is the proof, in charts.

First, the setting that changes everything — Cumulative Trades ON vs OFF

• ATAS's CVD Pro / Multi Market Powers has a checkbox: "Cumulative Trades." We ran this whole study both ways to show why it matters.
• A single large order often hits the tape as several smaller prints, not one.
• ON glues those pieces back into the real trade size (right side below).
• OFF treats every piece as its own tiny trade (left side) — a real 53-lot block trade disappears into 3 unrelated small readings.

One more thing before the findings — what does a "correlation" number mean?

You'll see numbers like +0.72 or -0.22 throughout what follows — a correlation, on a scale from -1 to +1:

• +0.72 means: net buying tends to line up with price going up at the same time (and net selling with price going down).
• It does not mean this group predicts price — and it's not a percentage or a win rate.
• Below, we don't just give you the number — we show you what it looks like.

Does the flow actually track price? — see it, don't just take our word

• Scatter dot = one 15-minute stretch. Left/right = how much that group bought or sold; up/down = what price did over that same 15 minutes.
• Cloud leaning bottom-left to top-right = a relationship you can see with your own eyes.
• Gauge = a simpler, different measure of the same thing: out of every 15-minute stretch, what % of the time did direction actually match price?
• The percentage won't equal the r-value — r factors in how big each move was, the percentage only counts direction — but a strong r and a high percentage move together, as you'll see below.

44%

of the time, 1 contract went the OPPOSITE way within that same 15-min window(worse than a coin flip)

69%

of the time, 5-9 contracts agreed with price within that same 15-min window(usually agreed)

70%

of the time, 10-19 contracts agreed with price within that same 15-min window(usually agreed)

80%

of the time, 2+ (drop 1-lots) agreed with price within that same 15-min window(usually agreed)

68%

of the time, 50+ agreed with price within that same 15-min window(usually agreed)

68%

of the time, 100+ agreed with price within that same 15-min window(usually agreed)

Which size range works best on its own — the sweet spot

Each size range tested alone, three ways: the same trailing 15-minute window used everywhere else in this report, a slower 60-minute window, and the running session total. Longer bar = tighter relationship to price.

The best practical filter — what to actually watch

Set a minimum size, count everything above it. "5+" = every trade of 5+ contracts, summed together. Same three windows as above: 15-min, 60-min, session total.

Does any size see the future? — leading vs. lagging, and how far we looked

• This is trend vs. trend, not instant vs. instant.
• Compares net buying/selling over a trailing 15-minute stretch to price's move over that same stretch.
• Slid that comparison from 60 min before to 60 min after, one minute at a time, to check every other offset.

• Below: all 6 size groups, plotted continuously from -60 to +60 minutes, not just a few points.
• Each line = % of 15-min windows where that bucket's trend matched price's trend, at that offset.
• Dashed line at 50% = coin flip, no relationship at all.

Two ways to watch this — running session total vs. a rolling window

• Session total = running tally since the 08:30 ET open.
• Rolling 15/60-min = only a recent window, resetting constantly.
• Common trade sizes: barely matters which you use. Rare/big sizes: matters a lot.

Who was actually buying and selling this year

NQ rose about 6,525 points over the 12 months tested. Net buying vs. selling by size group, same window:

When size groups disagree — the divergence test

• Every finding so far looked at one size group at a time. This tests what happens when two disagree.
• Paired 1 contract (works against price alone, r=-0.23) against 50+ (works with price alone, r=+0.44) — the two buckets that already behave the most differently.
• Split every 15-minute window into 4 states by what each group was doing, then checked what price did in that same window.

Does this hold up at the open too?

Split the same 4 states into the first 90 minutes of RTH (9:30-11:00 ET) vs. the rest of the session, to check whether the read changes at the open.

20 real examples — see it for yourself

• 4 real trading days × 5 size filters.
• Each chart stacks all 3 readings under price: rolling 15-min, rolling 60-min, session-cumulative.
• Click any chart to enlarge. Click a group below to expand.

1 contract — The odd one out — moves opposite to price

2025-05-15

2025-08-29

2025-12-15

2026-03-31

10-19 contracts — The single strongest size band on its own

2025-05-15

2025-08-29

2025-12-15

2026-03-31

2+ (drop 1-lots) — The best overall reading — everyone except 1-lots

2025-05-15

2025-08-29

2025-12-15

2026-03-31

50+ contracts — Bigger size — rarer, but net-accumulated all year

2025-05-15

2025-08-29

2025-12-15

2026-03-31

200-499 contracts — Genuinely large prints — a handful per day

2025-05-15

2025-08-29

2025-12-15

2026-03-31

How to actually use this

• Turn Cumulative Trades ON in ATAS's CVD Pro / Multi Market Powers, or any custom version you build.
• Don't chase the biggest prints expecting an edge — 200+ and 500+ contract trades are rare and, on their own, one of the weakest trackers of price minute to minute.
• The 5-20 contract range is the sweet spot if you want to watch one size band closely.
• If you want one simple filter, use "everything except 1-lots."
• Use the running session total for big/rare sizes, a rolling window for common sizes.
• None of this predicts price ahead of time — use it to confirm what's already happening, not to front-run a move that hasn't started.
• Watch for disagreement between size groups — small selling while big buys (or the reverse) reads price more cleanly than either group alone. Still won't tell you what happens next.