Why Do Your Stop-Losses Always Die at the Lowest Point? SEC Hard Evidence: Only 3% of Orders on U.S. Stocks Actually Execute, QQQ Not Even 1% — The Order Book Is Just an Electronic Screen! Retail Traders Make Up Only 2% of the Driving Force, Stop-Loss Clusters = Free Fuel for Machines — Stop Gambling on the Future with Drawn Lines

This video provides an in-depth analysis of the underlying reasons why retail investors in the U.S. stock market often encounter phenomena such as stop-losses being triggered at the lowest point or missing take-profit targets by a hair. Using data from SEC monitoring systems, the speaker points out that retail investors account for a negligible portion of order flow drivers; the real movers are market maker hedging orders, option books (especially positive/negative gamma shifts), and algorithmic programs. The video critiques the limitations of traditional technical analysis (charting) and the so-called "smart money" concept, explains the mechanisms by which liquidity, order cancellation rates, and stop-loss clusters become fuel for machine trading, and previews a quantitative algorithm the author is developing to calculate forced order flow.

Video link:

Full text:

Retail investors base their decisions on whether a chart looks good, while machines base theirs on book constraints, so you are fundamentally not in the same physical world. Observe things as they are and reject fragmented cognition. Today, let's talk about a slap in the face that every trader has experienced: why is it that when you set a stop-loss, it always gets precisely swept at the lowest point? After it's swept, the price turns around and rallies without looking back. And why is it that when you set a take-profit, you always miss it by just a hair? It hits the target, but then slips back. I'll give you the answer right now: because you are using a tool that only describes the past to bet on a future generated by machine trading. Therefore, your counterparty is not the retail investor next door, nor is it some mysterious "big player"; it is a mathematical engine. Today, we will identify who this engine is, why the lines you draw are always ineffective in front of them, and what exactly is wrong with your take-profit and stop-loss strategies. Let's look at this layer by layer.

In fact, many people have a default premise when reviewing their trades: they think that if the price goes up today, it's because there are more buyers, and if it falls, it's because there are more sellers—some kind of "bull vs. bear" battle, or market sentiment. While this mental image is quite vivid, it is not the true structure of the market. In a market like the U.S. stock market, what percentage of short-term price drivers do retail investors account for? According to the metrics we've broken down, it's less than 2%. Authoritative academic reports suggest their influence is less than 3%. You heard that right: out of 100 units of force pushing the price, all retail investors combined contribute less than two. Who accounts for the remaining 98? It's market maker hedging orders, institutional rebalancing, passive fund rebalancing at specific times, and various programmatic trades. What are they all? They are machine-driven and rule-based orders. The logic behind their orders doesn't look at those colorful channels, resistance or support levels, or what is called "trading volume" at all—or rather, they themselves are the support and resistance levels. Their trades only look at trading rules. You see, at the end of every month or quarter, whenever the index weights change, trading volume sometimes explodes to a massive level for no apparent reason. This isn't because someone is bullish or bearish; it's because index funds worldwide are contractually obligated to rebalance their positions before the close—they can't be off by even a single share. Therefore, retail buying and selling is not a driver for price; it is merely background noise. Most retail investors think they are participating in a game, but in reality, they are just providing fuel for the game.

Oh right, let me also mention another illusion on the side, which is trading volume. Many people treat this volume as an indicator of popularity, and by extension, as real money entering the market. They say that prices might lie, but trading volume won't lie to you in the same way. Well, volume is something that can be faked; for example, if the same batch of shares is traded back and forth between the left hand and the right hand through matched orders for a few rounds, the volume piles right up. The quotes of a market maker are essentially a left-hand-to-right-hand business. Suppose you are the buyer, how does a market maker operate? They take the seller's inventory, add a few points of spread to their own quote, and then sell it to you. This is a two-way quote, which is also the job a market maker has to do. Therefore, high volume does not equal real money and real guns. I wonder if any of you have looked at the quote cancellation rate in the Level 2 order book. Orders at the Level 2 tier are all cut from the same cloth; they are just putting on a show for everyone. A classic performance across several price levels is this: they place a few astronomical orders at key price levels, and then as the price moves in that direction, they suddenly pull them, followed immediately by market orders aggressively hammering the price down. Things like this happen every single day, you could say.

Speaking of which, I will give you an official, solid piece of evidence. This number comes straight from the SEC's own market monitoring system, Midas. Out of all the orders placed on the U.S. stock market's order book, how many actually end up being executed? Among individual stock order volumes, only about 3% are executed; the remaining 97% of placed orders are just meant to be canceled. ETFs like QQQ and SPY are even more absurd, with trading volumes accounting for less than 1% of orders. If 1,000 shares are placed inside the book, 998 of those shares are not intended to do business with you at all. Counting by the number of transactions, it is equally terrifying: behind every single actual transaction, there are dozens of orders placed and then canceled, canceled and then placed. The SEC has also been updating the magnitude of this sequence on a quarterly basis. Judging by the trend over the past dozen years or so, it is becoming increasingly illusory. Just think about this proportion: the buy and sell walls on the order book seem quite thick, but what are they essentially? They are just an electronic display screen where the numbers can be refreshed or disappear at any moment. Today's order books and warehouse receipts are just like advertisements, and their effects are the same. Therefore, if you use Level 2 order book entries as the basis for your trading decisions, what is that equivalent to? It is equivalent to looking at the mannequins in a department store window and deciding whom to marry based on those models. Every single wall you see could completely evaporate the exact second before you reach out your hand. Thus, the order book entries on Level 2 serve much more of a theatrical purpose. Who placed how many lots at what price—it seems like all the hidden cards have been laid bare, but everyone must understand that what can be laid bare is never the hidden card; this is the most theatrical place in the entire market. Everyone must remember that order books can be canceled. When a massive order is jammed in there, most of the time it is just luring retail investors in, and then all of it is pulled out just a second before execution. What will institutions that actually want to buy do instead? They will chop this order into hundreds of small pieces and feed them in bit by bit using algorithms, and you won't even be able to see this part—it is called an iceberg order. And not only are there icebergs, but things like spoofing, order walls, and all important elements actually require precise calculation.

Traders have a massive misconception: they feel that intuitive indicators are the only useful ones because this mindset requires the least cognitive effort. They tell you: sell when it hits the upper point, and buy when it drops to the lower point. But the reality is heartbreaking. Academia has already done something quite ruthless, which is pulling out all 7,800 or so indicators on the market for a collective autopsy, trading strictly according to the indicator instructions. As a result, after deducting transaction fees, the total number of samples that survived was zero. Then there are those kinds of indicators that don't have direct functions—what type are they? They are auxiliary analysis indicators, such as market state assessment or sentiment indices, and you have to do derivative analysis based on that foundation. Well, academia has tested this too; factoring in this category, some indicators saw their survival rate rise to a mere 2.3%. Moreover, these 2.3% share one common trait: none of them provide you with direct buying or selling signals. That is to say, truly effective indicators do not tell you whether to buy or sell; they only tell you what state the market is leaning toward today, or how expensive insurance is. Therefore, whether an indicator is valuable or worth using depends on one thing: how close it is to forced order flow. Anything far away from it is just window dressing. What is the essence of short-term trading? It is finding potential liquidity and the inflection points of liquidity exhaustion, rather than relying on a piece of news plus your own evaluation of whether it is bullish or bearish to figure out the market.

Here is an even more brutal truth: many people feel like they are being specifically targeted by the market, with their stop-loss being triggered the moment they set it, and the market bouncing right after they cut their losses, as if it has eyes. In reality, no one is targeting you personally; it is just that everyone acts too uniformly, with behaviors as if they were copied from the exact same mold. Let's look at the underlying principle, and everyone will understand. The main model used to sweep everyone is called the Poisson distribution. What does it specialize in calculating? It calculates the anomaly degree of events, paired with a probability density distribution. Do you know what the characteristics are? For instance, when active orders suddenly flood in at a certain price level, the entire Poisson distribution and probability density become severely distorted. Then, looking at the screen, they see such a high event arrival rate, such an anomaly, and what happens? Right below that price step, it triggers the order density estimation model to measure how many shares will still be sold if that price level is breached. At this point, the algorithm calculates and realizes there is indeed a profit to be made. What happens next? The algorithmic passive absorption of liquidity vanishes instantly, and it begins to go short by operating in reverse to the retail investors. The sole objective is to breach this area to obtain liquidity. Why does this happen? Because retail investors all over the world read the exact same batch of tutorials, draw the exact same charts, place their stop-losses in highly unified positions, and even trigger panic stampedes at the same threshold. Therefore, in the eyes of the machine, ten million retail investors are just one person. It doesn't need to see you at all; it only needs to see your normal distribution and probability density anomalies—this thing is more accurate than a clock.

Now let's talk about charting, drawing lines—things like support and resistance, moving averages, Bollinger Bands, distribution channels, and all sorts of channels. I collectively call them drawing or line-drawing. Actually, the problem with charting is not that it occasionally fails; what is it? It's that its logical foundation simply does not exist. You use a ruler to measure a line on historical K-line charts. This line describes the past shape of prices; it does not describe anyone's behavior, nor does it answer the only important question in the textbook: who is forced to place an order at this position? A position where no one is forced to trade has no mechanical significance whatsoever.

So someone will definitely say, ah, sometimes it works pretty well. It's simply because a lot of people drew it, and everyone placed orders together on the same line, coupled with the fact that retail investors only remember the few times it worked and forget the times it was broken through. This explains why support and resistance levels work sometimes and fail others—this is an old ailment of technical indicators. The line you drew happens to have real pending orders behind it, so it acts as support. The moment those orders are withdrawn or quietly wiped out by the algorithm, that same line becomes thin air. Lines are dead, but orders are alive. What determines whether it bounces or falls is never the line itself, but whether there are positions behind the line that must be executed. Therefore, betting on live behavior with dead shapes means losing is the norm, and winning is just luck.

By the way, this explains an old confusion many might have: why a line-drawing strategy in backtesting looks gorgeous, but falls apart completely the moment it goes live. Besides the reasons mentioned above, there is an even deeper factor: what kind of money does the line-drawing strategy make? Actually, it makes money off the market state. There is a market state that is naturally oscillating—just like now, prices grind back and forth, and any horizontal line you draw to buy high and sell low will work. Then there is another state that is naturally unilateral, and at that time, any support is made of paper. If the market period you backtested happened to coincide with the former state, your strategy was legendary. The moment the state switches, the exact same strategy will blow up on the spot. You think the strategy is working, but actually, the state is working. And the craft of charting is completely clueless about when the state switches. What is the state? Who is switching the state? Remember this question, we will talk about it in a moment.

And then, what is the next level up from drawing charts? It is the SMC (Smart Money Concepts) crowd—the so-called "smart money," "liquidity hunting," "supply zones," "demand zones," and all that stuff. Let me be fair: their direction is correct because they have started to think about something—they have started to think about behavior. Short-term market prices are indeed driven by liquidity, and the hunting of stop-losses does happen every single day. But the root cause of their problem is exactly the same as the chart-drawers: they only speculate, they provide no evidence. You draw a box on a chart and say, "Here is a supply zone, institutions are dumping goods here." On what basis? Where is the evidence of their positions? Where are the transaction receipts? If you dig deeper, did you draw this trend first and wait for the price to validate it, or did you fill it in after the price reversed? If it is the latter, that is just wishful thinking; anyone can do that. If that were fucking effective, it would be truly strange. But if it is the former, please present the corresponding evidence. If you cannot present it, then it is just like Fibonacci or channel lines—it is just a more intimidating piece of jargon. By the way, why is it so popular? Because it sells a very special story: you aren't losing money because you are bad at trading, but because "smart money" is hunting you in the shadows, because there are bad guys, conspiracies, and insider information. It provides plenty of emotional value. But emotional value cannot replace evidentiary value. If you attribute your losses to an invisible, intangible bad guy, it is methodologically the same as attributing your losses to Mercury being in retrograde. So, my evaluation of SMC is just one sentence: it points out the right problem, but then provides an answer without any evidence. It is true that stop-losses are hunted, but who is the hunter? Why act at that specific position? Where do the orders come from? They cannot answer that at all. They just tell you there is "smart money," but even "smart money" has no answer, or they say "insider," but who is the insider?

But is there really no answer? Actually, there is, and it can be calculated using public data. Where is the answer to the right question—who is being forced to place orders? It is in the options market, and on the books of market makers. What do market makers do? Simply put, they run a casino and take a cut. They rarely bet on direction; they earn the bid-ask spread. If you buy a call option on a brokerage app, your counterparty is most likely them. Once they sell you this call, they immediately carry risk on their books because if the index rises, they lose money, and they don't want to gamble. If they don't want to gamble, they have to buy some underlying assets in the market to hedge that risk. This action is called hedging. The trouble is, it is not a one-time deal. As long as the price moves, panic sentiment changes, and time passes day by day, the amount you need to hedge changes, so they have to constantly adjust their positions from morning till night. Think about their situation: the whole market comes to buy insurance, but there are only a few global insurance sellers—SIG, Citadel, Renaissance Technologies, etc. All the risk of these policies is piled onto their books. They don't bet on direction. Is liquidity hunting betting on direction? No, it is using the certainty of a liquidity vacuum to create short-term price movements. This is different from being bullish or bearish; everyone must distinguish this.

What is the professional approach? It is delta neutrality under the central view, so they have to constantly offload risk. And when these policies pile up to a certain point, the book slides into a very dangerous state: negative gamma. So, negative gamma is not a constant, nor is it innate; it changes according to the fluctuations of risk factors because it leads to what? It leads to the recalculation of risk exposure. Suppose you have ten option hedge orders in your hand, and your risk exposure is, say, one. Then something happens. Suppose Trump acts up again, says something or does something, and causes implied volatility to explode. Even if you still hold those ten orders and your position hasn't changed, your risk exposure jumps—say, from one to two. At this point, you are forced to act to push the risk exposure back down to one. So, gamma is not a deterministic value; it changes constantly with time, volatility, and events. This thing cannot be drawn on a chart at all, so you can only use dynamic algorithms to continuously approximate it, such as using a Markov-Bayesian combination to approach the gamma value. Under positive gamma conditions, the market tends toward range-bound trading—that is the range oscillation everyone sees. This game is about ensuring that all expiring options fail to reach the strike price, allowing market makers to collect the time value.

Conversely, once the market reaches a consensus, it instantly expands the market maker's risk exposure. As I just mentioned, the exposure goes from one to two, suddenly sliding into a negative gamma state. Risk control protocols force them to chase the rise and sell the fall; the more the price rises, the more they must buy, and the more it falls, the more they must sell. What does this turn into? It turns into buying high and selling low. You heard that right. It's not that they are fucking stupid; it's the ironclad rules of compliance and risk control forcing their hands to do exactly that. After the 2008 financial crisis, regulators basically welded shut the ability for market makers to use their proprietary trading desks to bet on market direction, so they are left with only one thing to do: balance their books. Let me give you an order of magnitude so you can get a feel for it. Suppose they sell 1,000 QQQ calls; for every dollar the price goes up, they have to chase and buy several thousand shares to balance the books. The fiercer the rise, the more aggressive the chase. This explains a bunch of market moves you've seen but couldn't understand. For example, a short squeeze with no news, or an inexplicable flash crash—if you peel back the layers, many are just these market makers biting their own tails, forced to buy and push prices higher, and higher prices force even more mandatory buying. This is a gamma squeeze. The most famous case is GameStop in 2021. What was the situation then? Retail investors banded together to sweep up out-of-the-money calls, and market makers, acting as the counterparty, were forced to chase and buy the underlying stock. The more they chased, the more it rose, and the more it rose, the more they had to chase. In two weeks, the stock price increased more than tenfold, truly beating the shit out of Wall Street's big shots. I can say this: half the fuel for that fire was sprayed out by this very engine itself.

Conversely, once the book flips, the same group of people will start selling as it rises and buying as it falls, which installs what for the market? It installs a shock absorber. The manifestation at this time is that the price gets stuck at a certain level, grinding back and forth, unable to break through. Then, in the days leading up to option expiration, the price often gets pinned to a round number range, unable to go up or down. That's actually what's going on. What else? Right, when an event concludes, for example, after a major event, even if the data is mediocre, the US stock market might show a slow bull run that even the bears can't understand. That's not market optimism; it's panic insurance collectively expiring. This is market makers mechanically buying back their hedging positions. Also, what about those inexplicable rallies or sell-offs at the market close, or the common Monday gap? That's also the time value of options placing orders for the market, and this hand is becoming increasingly heavy. Now, about half of the volume in US stock options is 0DTE, or zero-day-to-expiration options. As time ticks by, the books move accordingly. None of this is metaphysics; it is the automatic execution of a balance sheet locked by institutional rules.

When I say buying high and selling low is effective for trading, it is usually in some positive gamma state, so range-bound trading can sometimes be profitable. And that is not because your drawing skills are good, but because you happen to be in that trading environment. Once it slides into negative gamma, hey, it becomes a chase-the-rise-and-kill-the-fall gear, and sessions are the one-sided state of the market. Support, resistance, all of that is bullshit and ineffective. But if market makers are selling as the price rises and buying as the price falls in this gear, the market is in a consolidating state. At that time, even drawing a horizontal line works. So when you draw your technical analysis lines, when are they zero and when are they dead? From beginning to end, the ledger calls the shots. The person drawing lines just happens to have hitched a free ride. This is what I call true liquidity supply; it is never about which line, it is about this ledger. This ledger shows at which price level he must buy and at which price level he must sell, and this can be completely calculated. Now think again about how you set this stop-loss? Hey, is it slightly below the previous low, or slightly above the previous high, right? Or taking a round number just below to act as a support level. So here is the question, retail investors all over the world learn by looking at the same batch of tutorials, so tens of thousands of stop-loss orders are concentrated in the same batch of locations, and what do they pile up into? They pile up into clusters of stop-loss orders. What is a stop-loss order? Isn't it just a pre-promised market sell order? Isn't this just a pool of ready-made liquidity? Which market maker wouldn't drool when they see this? When the market maker's ledger is right in that chase-the-rise-and-kill-the-fall gear, as soon as the price drops, he is forced to sell, and then this wave of mechanical selling pressure precisely sweeps through the entire stop-loss cluster, and your stop-loss is triggered, by the way adding fuel to his selling pressure. When he finishes selling what needs to be sold, forced sell orders dry up, buying pressure emerges, the price naturally bounces back, and you get stopped out at the absolute lowest point. It is not that the market maker or the market has a grudge against you; it is that you placed your stop-loss right in the path of the machine and personally served as its fuel. Take-profit is a mirror image of the same story, meaning that the target price you measure out according to the chart often hits precisely on the ceiling pushed through by option positions. As soon as the price climbs there, mechanical selling pressure automatically drops. And then, you just watch with your own eyes, hey, it missed by just that last little breath. The second sin of the stop-loss is width. Most people's stop-loss width is fixed, either a fixed number of points or a fixed percentage, like getting out once it drops below a certain amount. But a fifty-point pullback in a shock-absorber gear is a disaster. In the market maker's shock-absorber state, hey, it is just a very large upward or downward move, but in the one-sided amplifier gear, it is just everyday breathing. If you take a stop-loss width nurtured by a oscillating market and plunge headfirst into a one-sided market, what does it equal? Let me use an inappropriate analogy, it equals running straight into a fire while wearing a rain poncho, right? It would be a miracle if you didn't get swept. So the width thing should move with what? It should move with the gear. If you don't know the current gear, its width is fucking just a blind guess.

So the conclusion is that the action of taking profit and stopping loss itself is not wrong; what is wrong is the basis for choosing the location. Retail investors' basis is whether the chart looks good, while the machine's basis is ledger constraints. So you are not in the same physical world at all. There is only one kind of effective location, and that is to first figure out where the forced order flow is actually buried, where the ceiling gets heavier the more prices rise, where the floor gets harder the more prices drop, and where the watershed is where the ledger divides from a shock absorber into an amplifier, and then move your order away from the machine's lane, because that's where accidents are easily caused, or simply hitch a ride along its lane. This map can also be calculated, and it uses all public data. But from the raw option chain data to an entry, take-profit, or stop-loss that can be drawn on the K-line, hey, that whole set of work in the middle is truly dirty and meticulous. What do you need to do? You need to clean the data, do attribution, and verify; not a single step can be skipped here. I now write my algorithms myself, but the convergence speed is a bit slow. Once I modify the algorithm again and finish running the backtest, I will report back to everyone. Rest assured, once I tinker it out, I will definitely open it for public testing.