On Monday 21 September 2026, the S&P 500 futures went up for three hours in something close to a straight line, and a tight trailing stop underneath them never got touched.
Linda Raschke posted the two-minute chart with a short verdict: a very efficient mark-up. Tom McClellan replied with a translation. A line that clean, he argued, is usually the footprint of a large amount of money being fed into the market in small pieces by execution algorithms, because the firms doing the buying do not want to move the price against themselves.
The idea deserves more than a thread. This guide turns “that looks clean” into a number you can compute, explains the machinery that produces it, shows what the research says about why these moves persist and why they stop, and sets out how to trade one. It also covers the uncomfortable part: straight lines have other causes, and the same signal fires on plenty of days that go nowhere.
The chart that started it
Raschke’s chart shows the December E-mini S&P 500 contract (ESZ26, listed as EPZ26 in CQG) on two-minute bars, from the 09:30 cash open to just before 13:00. Price opens near 7,756, climbs in a stair-step that barely pauses, and trades above 7,820 by 12:30: a move of more than 60 points in roughly three hours (levels read from her chart). At $50 a point on the full-size contract, that is over $3,000 per contract, on a day when the pullbacks were shallow enough to hold with a stop most traders would call suffocating.
Three things on her screen matter:
- A volatility channel (the blue bands) with a mid-line. Price spends almost the whole morning in the upper half and touches the mid-line only briefly.
- An ATR trailing stop (the red stepped line), ratcheting higher beneath price and never tagged during the advance.
- Her 3/10 oscillator in the lower panel, a short-term momentum measure: a 3-period simple moving average minus a 10-period one, with a 16-period average of the result as its signal line. After a brief dip below zero at the open, it prints one positive swing after another.
Raschke is one of the traders Jack Schwager interviewed for The New Market Wizards, and her own story is worth reading in full. McClellan, whose parents Sherman and Marian McClellan created the McClellan Oscillator in 1969, supplied the engineering vocabulary: he described the session as having a high “signal to noise” ratio.
That phrase is the key to everything that follows, because signal-to-noise can be measured.
Sources: Linda Raschke on X and Tom McClellan on X, 21 September 2026.
Putting a number on “clean”
Traders use the word clean as if it were a feeling. It is closer to arithmetic. Perry Kaufman defined the Efficiency Ratio in Smarter Trading (1995) as the net distance price travels divided by the total distance it covers getting there:
Efficiency Ratio (ER)
ER = | last close − first close | ÷ sum of | each bar’s close-to-close change |
1.0 means every bar moved the same way. Close to 0 means price travelled a long way and ended roughly where it began.
Two sessions can finish with the same gain and have nothing else in common:
| Illustrative session | Net move | Total path | ER |
|---|---|---|---|
| A: the staircase | +40 points | 80 points | 0.50 |
| B: the washing machine | +40 points | 400 points | 0.10 |
Session A paid anyone who bought and held. Session B paid the same amount on paper and stopped out nearly everyone who tried. Kaufman built his adaptive moving average around this ratio. For our purposes it is simply the best single-number summary of a session’s signal-to-noise.
Two cautions before using it. First, ER depends on bar size and window length: a two-minute ER across one session cannot be compared with a daily ER across a month. Second, there is no universal threshold. What counts as high for the E-mini is set by the E-mini’s own history, so calibrate any threshold against your market’s recent sessions rather than taking one from a textbook.
What a random market scores
Before calling a session unusual, you need to know what ordinary looks like. We simulated 20,000 cash sessions of 195 two-minute bars (09:30 to 16:00 ET) in which price is a pure random walk, then repeated the exercise with a steady one-sided push added. The push is expressed as signal-to-noise: the average drift per bar divided by the typical size of a bar’s move. We also opened a long position at the open with a trailing stop set a fixed multiple of the typical two-minute move below the highest close, and recorded how often it lasted the day.
| Signal-to-noise per bar | Median ER | 90th percentile ER | Tight trail (3×) survives | Wide trail (6×) survives |
|---|---|---|---|---|
| 0.0 (random walk) | 0.06 | 0.15 | 0.0% | 1.2% |
| 0.1 | 0.13 | 0.24 | 0.0% | 9.5% |
| 0.2 | 0.25 | 0.35 | 0.1% | 31.6% |
| 0.3 | 0.36 | 0.46 | 1.4% | 61.2% |
| 0.5 | 0.56 | 0.64 | 19.2% | 92.8% |
Simulation, not market data. 20,000 sessions per row, Gaussian noise, constant drift, exits filled at the bar close that breaches the stop.
Three things stand out.
A random session almost never looks efficient. Nine random sessions in ten score below 0.15. Only about one in 40 clears 0.20, one in 200 clears 0.25, and roughly one in 1,200 clears 0.30.
A tight trail almost never survives a full day, even with a real push behind price. At a signal-to-noise of 0.3, where the drift per bar is less than a third of the noise, a stop three typical moves behind the high lasts the session 1.4% of the time. For a tight stop to go untouched for hours, the push has to dominate the noise. That is precisely the condition McClellan was describing.
A small, persistent push transforms a wide trail. A drift one-fifth the size of the noise takes the wide trail’s survival from about 1% to about 32%. The push does not need to be dramatic. It needs to be relentless.
The point of the exercise. A session in which a tight trail survives for hours is not a slightly better day. In a random market it is close to impossible. Either something was pushing steadily in one direction, or the noise itself shrank. Both are information worth acting on.
The model is a toy. Real markets have fatter tails, volatility that clusters around the open and close, and noise that is not independent from one bar to the next. Treat the numbers as a map of the logic, not a calibration. The toy is useful for one thing: it shows how far outside normal Raschke’s chart sits.
Why big money cannot simply press buy
Suppose a large asset manager needs to add a substantial amount of S&P 500 exposure in a day, whether from client inflows, a rebalance or a change of view. The E-mini is one of the deepest futures markets in the world, but the size resting at each price in the order book is finite. An order large enough to matter would sweep through several price levels at once, pay more at each one, and announce itself to every high-frequency participant watching the book. The cost of that is called market impact, and for a large fund it can dwarf commissions. It is also why displayed liquidity can vanish exactly when you lean on it.
The standard solution is to split the parent order into many small child orders and release them over time through an execution algorithm. The main families leave different footprints:
| Algorithm | How it slices | Likely footprint |
|---|---|---|
| VWAP | In proportion to the day’s expected volume curve, aiming to match the session’s volume-weighted average price | Steady pressure, heaviest near the open and close where volume concentrates |
| TWAP | Equal slices at equal time intervals | Metronomic; easy for other algorithms to detect unless randomised |
| Participation (POV) | A fixed share of whatever volume actually trades, such as 10% | Buys harder when the market is busy, eases off when it is quiet |
| Implementation shortfall | Trades off impact against the risk of price drifting away, usually front-loading | Strongest push early in the order’s life |
| Iceberg orders | Shows a small visible quantity that refills as it is hit | A price level that keeps absorbing the other side |
Implementation shortfall comes from the optimal-execution framework Robert Almgren and Neil Chriss published in 2000, and VWAP is the benchmark most institutional desks are judged against, which is why it matters so much to intraday traders. Our VWAP trading strategy guide covers how to read it.
Each child order is small enough to look like noise. The sequence is not. When a buyer keeps lifting offers every few seconds for hours, sellers who step in get absorbed, dips are bought before they develop, and the chart draws the staircase on Raschke’s screen. Some of that buying never touches the visible book at all: it runs through dark pools in the underlying stocks, and index arbitrage carries the pressure across to the futures. If you want the plumbing in detail, how an order travels from click to fill is the place to start.
The research: order flow has a long memory
McClellan’s explanation is not folklore. Market-microstructure researchers have studied it for two decades, and three findings line up with what the chart shows.
Buying follows buying. Studying the London Stock Exchange, Fabrizio Lillo and J. Doyne Farmer found that the sequence of buy and sell orders has a long memory: a buy tends to be followed by more buys for far longer than a random market would allow (The Long Memory of the Efficient Market, 2004). The leading explanation is order splitting, large parent orders executed piece by piece, and later work by the same group, including a 2015 paper titled Why is equity order flow so persistent?, pursued it further.
Impact is real but concave. Across many studies, the price impact of a large order grows roughly with the square root of its size relative to daily volume, so doubling the order raises the impact by around 40%, not 100% (Almgren and colleagues, 2005; Tóth and colleagues, 2011). A big buyer moves the price, but in a grinding, persistent way spread over the life of the order. That is a description of a staircase.
Some of the move fades when the buying stops. Studies that track orders to completion find that price gives back part of the impact once execution ends. Bacry and colleagues documented this decay across roughly 400,000 investor metaorders on European markets (2015), and a 2018 study of metaorder data found the lasting impact settling at about two-thirds of the peak. The move is not all permanent information. Part of it is the pressure of the order itself.
Put together, the research supports McClellan’s reading as a mechanism. Persistent one-sided flow exists, it comes largely from split orders, and it produces steady price pressure for as long as the order is being worked. What the research cannot do is tell you, on any given morning, that this is what you are looking at.
Wyckoff had a name for it: mark-up
Raschke’s word choice points straight at Richard Wyckoff. His market cycle runs through four phases: accumulation, when large operators quietly build a position; mark-up, when supply has been absorbed and price is allowed to rise; distribution; and mark-down. Wyckoff personified the large money as the Composite Operator and taught traders to read its intentions from price and volume.
The mark-up phase has a recognisable signature: higher highs and higher lows, shallow reactions that fail to attract fresh supply, and pullbacks that are bought before they reach meaningful support. Compress that into a single session and you get the 21 September chart. The Composite Operator in 2026 is less a person than a parent order and the algorithm working it. Our complete guide to the Wyckoff Method covers the full cycle.
How to spot a trend day while it is happening
Recognising a trend day at 15:00 is easy. The skill is recognising one by 10:30, while there is still a trade to take. Market Profile traders call this day type a trend day: an open near one extreme of the range, a one-sided auction, and a close near the other extreme. The evidence builds in layers.
Trend-day checklist
1. An opening drive that does not come back. Price leaves the open in the first 30 minutes and the opening price is not revisited.
2. Price holds one side of session VWAP. Pullbacks find buyers at or above it rather than slicing through.
3. Pullbacks stop at the channel mid-line. They do not reach the far band.
4. Momentum stays positive. Short-term oscillators dip only briefly and shallowly below zero.
5. First-hour ER is high for this market. Judged against the instrument’s own history, not a textbook number.
6. Range expands early. By late morning the session has already covered a large share of a normal day’s range.
7. A short-timeframe trail is intact. Raschke’s own test, and the one that doubles as your exit.
Now the uncomfortable number. We built a second simulation in which 15% of sessions carry a steady push (signal-to-noise 0.3) and the rest are random, then flagged a session as a trend day whenever the first hour closed up with an ER above a threshold.
| First-hour ER above | Days flagged | Flags that were real | Real trend days caught |
|---|---|---|---|
| 0.2 | 28% | 41% | 78% |
| 0.3 | 18% | 53% | 62% |
| 0.4 | 10% | 66% | 44% |
Simulation, not market data. 200,000 sessions; trend days assumed to push at a constant rate all day.
Even in a world built to contain clean trend days, a first-hour filter at a sensible threshold is wrong about half the time, and tightening it simply swaps false alarms for missed days. Real markets are messier than the model. That is why the checklist has seven items rather than one, and why the stop matters more than the diagnosis.
Trend day trading through Mind, Method and Money
Mind: stop arguing with the tape
The hardest part of a trend day is psychological. Everything looks extended from 10:15 onwards, and every instinct trained on range days says sell the stretch. Fading an order that may have hours of buying left is one of the most reliable ways to give back a month on a single morning. The opposite error is just as real: chasing the upper band at noon because you missed the open.
The shift is to stop asking how high it can go and start asking whether the evidence is still intact. The checklist answers the second question. Nobody can answer the first.
Method: buy the pause, trail the trend
Enter on pullbacks towards the channel mid-line or session VWAP, not on breakouts at the upper band. Then let a volatility trail do the managing. On a trend day the trail is doing two jobs at once: it protects the position, and it is the live test of the thesis. When the trail is hit, the trend-day read is suspended, whatever your view of the reason.
Resist fixed profit targets. Our trailing stops guide shows the same entry returning very different results depending purely on the exit rule, with wide trails winning in trends and failing in ranges. Trend days are the sessions the wide trail exists for, which is the point made in the edge is in the exits.
Money: a wider stop means a smaller position
A trail that can survive a trend day sits further away than a scalper’s stop, so size has to come down to hold risk constant. On the full E-mini, each point is worth $50. As an illustration, a 12-point stop risks $600 per contract before slippage. A trader risking 1% of a $50,000 account has $500 to lose, which is less than one full contract. The Micro E-mini, at $5 a point, makes the same trade possible: eight micros risk $480. Run your own numbers in the position size calculator, and see our guide on how to trade futures for contract specifications.
One caution for funded accounts: a trailing drawdown limit caps how much open profit you can give back, which quietly caps how wide your trail can sit. Check the rule before the trade, not during it.
When the algorithm finishes
On Raschke’s chart the advance flattens after about 12:15. Price moves sideways near the top of the channel, the oscillator loses its rhythm, and the bars turn mixed. Nobody outside the executing desk knows whether an order finished at that moment. It is equally consistent with the lunchtime volume lull that affects most sessions. On 21 September it turned out to be the lull: the advance resumed in the afternoon.
The research on completed orders offers a useful warning, though. If part of a trend day’s move is the pressure of the order itself, part of it can fade when the order is done. A morning staircase does not promise an afternoon one. Practically:
- Do not add to the position late in the session because the morning was clean.
- Watch a rolling ER over the last 30 to 60 minutes. A sharp fall in efficiency while price stalls is the chart telling you the push has paused.
- Let the trail make the exit decision. Its job is to be right about the end of the trend without you having to be.
What happened next
Raschke posted her chart just before 13:00 ET. The staircase did not stop there. After a midday pause, the December contract climbed again through the afternoon and peaked in the mid-7,840s late in the cash session, roughly 90 points above the morning’s starting level. Then the rest of the week unfolded like this:
| Session | What price did |
|---|---|
| Mon 21 Sept | Cash-session staircase from about 7,756 to about 7,845 |
| Tue 22 Sept | Sideways, roughly 7,810 to 7,850. No follow-through |
| Wed 23 Sept | Turned lower in the morning and fell through the session to the 7,760s |
| Thu 24 Sept | Overnight low near 7,710, below where Monday’s rally began |
Approximate levels read from a two-minute chart of the continuous E-mini contract (ES1!) on TradingView.
Within two sessions the entire mark-up had been handed back, and then some. That does not make Monday’s read wrong. It shows exactly where the read ends.
The research on completed orders predicts that part of an order’s impact fades once the buying stops. A full retracement is more than that, which suggests fresh selling arrived on Wednesday for reasons of its own. We have not attributed a cause, and the chart cannot supply one. What the week does show is the scope of a trend day. It is an intraday event with an intraday edge. A trader who bought Monday’s pullbacks and let a trail manage the position would have been stopped out somewhere in the Tuesday chop, with most of the move banked. A trader who turned Monday into a swing thesis, because the buying looked so determined, would have watched the whole gain disappear.
The footprint tells you what someone is doing today. It says nothing about what the market will do on Wednesday. Trade the staircase while it is being built, and let the trail decide when it is finished.
What else draws a straight line
McClellan’s reading is an inference from the shape of the chart. Nobody posting about 21 September saw the order tickets. Several other forces can produce a staircase, and some leave almost identical footprints.
- Short covering. Forced buying by traders caught short looks like accumulation on a price chart. It tends to be sharper and to stop more abruptly, but not always.
- Options dealer hedging. In the widely used dealer-positioning framework, dealers who are net long options gamma sell rallies and buy dips, which damps volatility, while hedging adjustments as options decay can create steady drift. The framework is popular. Evidence that it explains any specific session is contested.
- Calendar and systematic flows. 21 September was the first session after the September quarterly futures and options expiry on the 18th, and quarter-end was nine days away. Rebalancing and volatility-targeting funds trade on schedules that can create persistent one-sided flow with no view behind it at all.
- Corporate buybacks. Companies repurchasing shares typically use execution algorithms and operate within the SEC’s Rule 10b-18 safe harbour, which limits daily volume and trade timing. That spreads the buying through the session in the underlying stocks, and the index carries it. See how passive and mechanical flows move prices with no headline at all.
- A thin, quiet market. ER measures shape, not size. A low-volume session with little two-way trade can drift smoothly with very little money behind it. Read ER alongside volume.
- Hindsight. The charts that get posted are the ones that worked. Mornings that looked exactly like this and reversed at 11:00 do not go viral.
None of this makes the trade wrong. It makes the story optional. You do not need to know who is buying to trade a trend day well. You need the evidence to stay intact and a stop that tells you when it is not.
A note for Smart Money traders
If you trade ICT or Smart Money Concepts, this is the mechanism beneath the vocabulary. Institutional order flow, in practice, is rarely one order parked at an order block. It is a parent order being worked by algorithm over hours, and its footprints are persistence, dips absorbed before they develop, and a trail that does not get hit. Those footprints are observable, measurable and falsifiable.
Order blocks and fair value gaps can still be sensible places to locate entries inside a trend day. What they cannot do is tell you who is buying, how much is left, or when they will stop. Our advanced Smart Money Concepts guide covers the entry models; this article covers the context they sit inside. For a broader view of where human traders stand against the machines, see algorithmic vs manual trading.
The honest summary
A trend day where a tight stop survives for hours is rare in any market driven by noise, and the likeliest explanation is a large order being worked patiently by algorithm. The research on order splitting and market impact supports that mechanism. It does not confirm it on any particular day.
What you can do is measure. The Efficiency Ratio turns “clean” into a number, your instrument’s own history tells you whether today is unusual, and a volatility trail both manages the trade and tells you when the read has failed. Take the checklist, respect the base rates, size for the wider stop, and let the market decide how long the staircase runs.
You do not need to know who is buying. You need to know when the evidence stops.
Size the wider stop before the next trend day, not during it.
The Complete Trader's Edge
The full Mind · Method · Money framework. 70 chapters.
View on Amazon →
Market Mayhem
400 years of bubbles, crashes, and the pattern that keeps repeating.
Buy on Amazon →
Greatest Companies
How the world's greatest companies were built — and what traders learn from them.
View on Amazon →
Greatest Traders
Eighty-six lives that explain the markets — Livermore to Madoff, told with the losses left in.
View on Amazon →
