Systematic trading focuses on a necessary, but insufficient question: does this edge work?

The right question should be why it works.

Of course, absolute certainty in a noisy and probabilistic system is impossible, but the difference between the two questions is what determines if an edge is real, if an edge will persist, and whether you can identify it will end before data tells you so.

A strategy that cannot explain itself is difficult to distinguish from a statistical accident.

Distinguishing the two requires a mixture of both qualitative and quantitative analysis. The two are not clashing ideologies, but rather complementary stages of the same process. A strategy built on both is better than one built on either alone.

It’ll Work Eventually

During my experience working at a trading desk, I recall an experience where a new trader came in and was explaining a proprietary indicator he had developed that allegedly marked out support and resistance levels more accurately than any discretionary method.

I looked at the indicator myself, and there was a support or resistance level at every 10 point interval on the NASDAQ futures contract. He pointed out every instance where price barely touched a level and perfectly reversed. He seemed so adamant that there was a true edge to be found in his indicator and pushed me to further collaborate on it in order to develop a rule-based strategy for it.

What made the experience so memorable was the fact that I had an inside joke with one of my colleagues at the time: if you put a line anywhere on a price chart where price has previously traded before (i.e. not at an all-time high) it will eventually act as a perfect support or resistance level.

The point of this story is to demonstrate the idea of false-positives: if you test enough ideas, enough rules, and enough strategies, then solely by chance some of them will appear to have a statistically compelling p-value. At a 95% confidence level, testing 100 will, in theory, give you five false positives: strategies that have compelling quantitative evidence to support capital deployment, but are nothing more than statistical accidents that survived chance.

Quantitative researchers recognized this and implemented the Bonferroni correction. The issue is that it is a patch on a wound that does not solve the underlying issue. Bonferroni adjusts the threshold after the fact. The correct methodology would then seek to subvert any post-hoc adjustments: one pre-specified theory, one test, and measure what the data supports.

Pre-specification is a commitment to what you are testing before you see the data. Pre-specification only works if you have a plausible theory as to why an edge should exist before you look for it. Theory turns the arbitrary into honesty.

Pre-specification is not a statistical technique, but rather a qualitative discipline that requires a theory to be meaningful.

The Arbitrage Problem

There is an inherent decay risk in publicly known and replicable trading signals. EMA crossovers, Bollinger Band touches, RSI levels, MACD crossovers: these signals are documented, widely implemented, and algorithmically replicable.

If such an edge were to be found solely because a sufficient number of participants are reacting to the same signal, the edge is visible to participants who can exploit the predictability of it. This is the crowding problem in trading. The mechanism is not structural. It is behavioral, and behavioral edges are contingent on the continued presence of the specific behavior.

As such, the distinction must be made: obligation versus convention and discretion.

Options dealers hedging a short gamma position are not discretionarily choosing to create directional flow. It is a mechanical consequence of the market. It does not go away because price is touching an EMA level. It does not get arbitraged out because it cannot be predicted far enough in advance to front-run it at scale. It exists because of what the market is, not because of what traders have agreed to do.

I have used options hedging as an example, but virtually any mechanism of market microstructure would fall into the realm of obligation, rather than convention and discretion.

The honest qualification is that some indicator strategies do work because the indicator happens to approximate a genuine structural variable. A moving average might coincidentally track volatility regime changes. There is potential predictive power, but the power comes from an underlying variable, not the indicator itself, so the honest response is to identify the underlying variable and measure it directly.

The question to ask of any signal is not whether it has worked historically. As previously discussed, enough trials and enough chances will produce false positives. It is whether there is a plausible reason it should continue to work, a question which is answered by a theory.

The Decay Problem

On the nature of obligation versus discretion, I would hypothesize that two types of edge decay would stem from these, differing in their predictability.

Behavioral edge decay happens when participants stop using the signal, when the participant population changes, or when algorithmic participants learn to exploit predictable behavior. Unpredictable in timing, with no warning, and the edge disappears from the data without a structural reason you could have identified in advance.

Structural edge decay happens when the underlying market structure changes. For example, the options dealer hedging mechanism is contingent on the continued existence and popularity of short-dated options. If expiration calendars change, regulatory treatment of 0DTE options changes, or if options market structure evolves in any way, the mechanism, and thus the strategy, changes with it. But these are observable and measurable developments that happen on regulatory timelines. Because theory specifies the mechanism, you know what to watch.

Strategy performance tracking is limited to quantitative methods, but a structural theory gives you a qualitative method for tracking performance. You are not waiting for the equity curve to tell you something changed. You are monitoring conditions that make the edge possible and can identify deterioration before it appears in returns.

Robustness and Falsifiability

One honest pushback on my idea here: theory-first development could just be sophisticated post-hoc pattern finding. Generating a plausible-sounding story constructed after observing a pattern is indistinguishable from a genuine theory.

Theory-first pre-specification establishes a sequence. A theory written before data is collected is accountable in a way that post-hoc rationalization is not. Timestamp on development is the first line of evidence against this. Theory, rules, data collection, then examination.

Theory can also have an independent verifiability of the mechanism in question. A genuine structural claim makes predictions about market structure that can be verified through channels other than a chart or the strategy’s performance. Theory can be corroborated by independently observable market history.

Lastly, a genuine theory would be able to make predictions on performance based on new and untested conditions. What happens to dealer hedging when options open interest is low? When the VIX is elevated? In instruments with low options activity? A data-mined pattern cannot generate these predictions because it has no mechanism to reason from. A theory can, and the accuracy of those predictions is the strongest possible evidence that the mechanism is truly real and robust.

As the nature of all speculation, complete certainty is not achievable. What is achievable is a methodology that makes coincidence and chance increasingly less likely to attribute to performance as evidence accumulates.

The claim is not that theory-first strategies are always vastly profitable and have significant edges. They are, however, testable, always revisable, and always generating knowledge, especially when they fail. A data-mined strategy that stops working will only tell you that it stopped working.

Falsifiability is not a technique layered on top of strategy development. It is a natural consequence of having a theory that can be independently substantiated. If you cannot specify conditions where an edge should not work, you merely have a description.

Before testing any strategy, ask whether you can plausibly explain why it exists and why you are able to extract alpha from it. Who is on the other side? What obligation, incentive, or other reason do they have to be there? And under what conditions would this disappear.

Inability to answer those questions makes a strategy indistinguishable from a statistical accident regardless of what a backtest says.

The European Securities and Markets Authority (ESMA) requires that CFD brokers based in Europe must post the percentage of clients that trade with them and lose money. Between 74 to 89% of all traders with European brokers lose money.

Brokers are not interested in seeing their clients fail, rather the opposite. The profitable client is the client that trades more, and is eventually the client that will pay more commissions to the broker. One can argue that the infrastructure to trade has never been better: many brokers are offering extremely tight spreads, with low margins required, and the accessibility for knowledge, thanks to the internet and social media, has never been more free.

Why then, after improved infrastructure, is the failure rate so drastically high? Contrary to popular belief, institutions do not see to it that the retail trader fails. What I propose in this article is that the individual retail trader, despite meaning well, is actually beating themselves out of the market. And unfortunately, the current trading education space does little to mitigate the beatings. It has put the cart before the horse, and at times, monetized a cart with wheels that do not work.

The Brain in Hell: Trading is Cognitively Extreme and Unique

Do you remember your first trade? I do: it was on a small cap stock named GRRR. I was on the lucky end of a limit up market where I made my first $100 trading stocks. The euphoria from that trade was unimaginable. I imagine, if you won your first trade as well, that there is an equally strong euphoric experience from your end. If you didn't win your first trade, I imagine that you might not be here at all, and you will have said to hell with the market and gone about your day.

I remember there was a day I sat in a 9AM college class and had been staring at the market, waiting for it to open. I found a biotech company with the ticker RNAZ and once again found myself on the lucky end of a limit up market where I made my first $1,000 trading stocks. I immediately walked out of the class. I vividly remember the walk back to my dorm room, and I felt as if at any moment, a group of personal servants would pop out from the bushes and start rolling red carpet where my feet would land on my triumphant walk back to a twin bed where I slept in for the day after leaving class.

Why do I tell you this? I want to demonstrate how extreme the environment of trading is. I have experienced complete euphoria from trading, where at any moment, it feels as if the entire world were plotting in your favor, and conversely, experienced complete dysphoria, where it feels like not a single thing in the world can go your way.

The emotional reactions produced from trading are extreme because the cognitive environment of trading the financial markets with leverage is extreme. And not just extreme, but uniquely extreme and quite hostile to normal, human evolutionary behavior. I'd dare to venture and say that a normal human being, with normal thought patterns and behaviors, might fail to successfully trade the financial markets. Not that they are unintelligent, or lazy, but rather the human brain was specifically designed, by evolution, to abhor the conditions the financial markets present.

Knightian uncertainty is not the same as risk. Risk is able to be quantified: I risk, and I will lose at most, 2% from this trade. Knightian uncertainty is genuinely unknowable. No model can account for it, and the destruction that it runs on the brain is largely responsible for the extremes of trading. Out in the wilderness, there was no room for probabilities. There was only life and death. If you hear a rustle in the bush, you will likely survive more if you assume there is a predator there rather than if it was just the wind. Your brain is inherently not designed for uncertainty, as hundreds of thousands of years of evolution have equated uncertainty to a lesser chance of survival.

"Nobody, and I don't care if you're Warren Buffett or Jimmy Buffett, nobody knows if a stock is gonna go up, down, sideways, or in fucking circles. Least of all stock brokers." — Matthew McConaughey, The Wolf of Wall Street

Every trade you place is under an element of uncertainty. You can compute all the statistics you want, but you will never definitively know if a trade will win or lose. That is simply a fact that the human brain abhors. The brain will equate the discomfort of a trade to a threat to your survival. When your mind is uncomfortable and angry, what does that manifest into? Self-destructive behavior:

These are all physical manifestations of a mind that is in deep discomfort. All people are going to be unprepared for this because there is no prior experience that can truly emulate how this feels in real time, with money on the line.

That is another thing that makes this pervasive uncertainty even more destructive: leverage. From my trading today, a 0.35% move in the underlying NASDAQ represented a gain of 2.5% on my account. The same move downward would have represented an equal 2.5% loss. When someone is able to take $10,000 and control, in theory, millions of dollars worth of notional exposure, gains and losses accumulate rapidly. When losses accumulate, the brain must ease the pain of financial loss, which as empirically documented by Kahneman and Tversky, is twice as painful as an equivalent gain.

The operative idea between uncertainty and leverage is understanding that your brain does not process information. It does not process the numbers. It does not process the market you see in front of your eyes. It processes the emotional magnitude of whatever narrative you have sold yourself, which is likely a narrative of certainty, simply because that is the natural human tendency.

Between the emotional heaven and hell, it is virtually guaranteed that any trader's brain will start reinforcing feedback loops for particular behaviors. The probability of heads or tails on a coin flip is 50/50. That does not mean that each flip will alternate perfectly. You might get streaks of five heads. Statistically speaking, if you flip the coin long enough, you might see outlier streaks of ten or eleven heads. The distribution of these streaks is random, meaning that if you start flipping the coin right now, you could easily encounter that outlier streak. As the number of coin flips reaches infinity, the distribution will even out to 50/50.

A trader who has a strategy with zero edge, a 50% win rate where it is mathematically impossible to make money in the long run, might encounter a streak of six wins. This trader is feeling fantastic. No one can tell them no. The feedback is immediate: the money is in their account and they are withdrawing it as I write this, about to buy themselves some new clothes. The feedback is also, unfortunately, delayed. Their brain has now been conditioned to run the strategy because of the emotional feedback of this lucky fluke, and the delayed feedback manifests itself as ruin over the long run. They will fail to produce a profit over the course of 200 or 300 trades because their brain has conditioned itself to associate this strategy with profits, despite the fact that the math shows otherwise.

Slot machines exploit this same feedback loop psychology: bright lights and loud sounds reinforce the behavior of hitting spin, despite the fact that someone may conceptually understand that the house always wins. The failure rate is not an issue of knowledge, but rather traders being dropped into a cognitive environment that they could not have possibly known existed.

I Blame Stochastics: What Education Is Selling

Virtually all available trading education is filled with how-to's. How do I use the MACD and RSI? How do I trade a hammer pattern? How do I mark support and resistance? It is, at best, pattern recognition with a specific narrative attached to it. The TA provided by current trading education is not evidence-based and is cherry-picked to produce a marketable narrative.

A broker explaining the MACD will equate the 9MA crossing above the 21MA as a bullish signal and will cherry-pick examples while leaving out every example on the chart where the crossover does not work. The RSI is notorious for equating trending conditions as consistently overbought or oversold, with the narrative that the dominant move should be faded, despite the fact that a market in extreme trending conditions will likely persist with the same trend. A hammer candlestick pattern looks like a good tool to time every top and bottom, until you go back to charts and realize it occurs more frequently than you imagine and is not predictive of price at all.

There is a characteristic of the human brain called apophenia: the tendency to find patterns where no patterns exist. Again, a product of evolution. The education market has done a great disservice to traders because it reinforces aggressive pattern-finding with virtually no statistical rigor attached, but rather an emotional and certainty-filled narrative that does not hold up in live markets. Your brain and my brain will do their best to form some type of story around whatever chart they are looking at. Do not listen to it.

Well-meaning brokers with a genuine interest in client success still peddle these flimsy how-to's because the education itself is structurally flawed. They are not being purposefully dishonest. The issue is that many of the patterns and technical setups that are sold online and in books lack statistical validation that would make them distinguishable from randomness on any finite sample. What would the alternative look like? It starts with a simple question: how do I know whether or not this pattern is a real pattern against noise? Not how do I identify this pattern, but how do I test if this pattern has any predictive value?

The Role of Quantitative Analysis and Evidence-Based Technical Analysis

Before I write this section, I must preface that there is no feasible way that anyone can rid themselves of the uncertainty of the markets. It is its premier feature. Quantitative analysis only serves as a way to make the gap between knowing and not knowing a little more narrow. It is not a cut that is sutured fully.

The goal with any form of quantitative analysis or evidence-based technical analysis is answering one question: does this approach have a positive expected value over a large number of repetitions? Forget this trade. Forget the next trade. The mind should shift to a longer horizon. Day by day as we bob on with our lives, it is hard to see that the current is actually inching us ever so slightly closer to one direction. When you lose today, your brain will immediately and erroneously conclude that you have regressed. When you win today, the brain will think of how far it has gotten. The short-term horizon is dominated by noise and endless bobbing. Quantitative analysis shifts things toward the long term.

In all of my strategies, I employ a statistical method called binomial significance testing. It is not exotic mathematics and a high school math student can do it. It simply answers one question: given the sample of trades I have and the observed win rate, what is the probability that this result happens by pure chance?

Take a sample of 700 trades with a 54% win rate: 378 wins and 322 losses. The question is how often a breakeven strategy at 50% produces a result this extreme by pure randomness. The answer comes from how far 378 sits from the expected 350 wins of a coin-flip strategy, measured in units of standard deviation. In this case, 378 wins sits 2.12 standard deviations above baseline. That distance corresponds to a probability of 1.7%. If the strategy had no real edge, this result would occur by luck once in every sixty times. This strategy is worth committing capital to.

This one formula is the entry point for whether or not a particular pattern or strategy should have a single dollar committed to it. If one cannot statistically prove that a pattern can be exploited with results better than chance, it should have no money dedicated to it. Such testing never makes it to the education space, simply because it is easier to sell a pattern and a narrative rather than formulas that are imperative and foundational to profitable trading.

There is an erroneous belief floating around the trading space that psychology, discipline, and fortitude are what is needed for profitable trading. I agree to an extent. But no one leaves out the fact that even if your mindset is fortified with obsidian, it would not matter if you were trading a strategy with zero detectable edge.

Quantitative analysis shifts the questions and narratives. If I can statistically prove that a strategy is unlikely to be profitable by pure randomness alone, how will I feel when I come across a losing streak? Sure, I will be in pain when I lose five or six trades in a row, but I will be much more confident going into the next trading day because I have a degree of certainty in my strategy. I am not operating on blind faith. I do not have my hands bound to useless prayer that I will hopefully make money. Rather, I have put in the work to validate that it is likely I will make money over the longer term, and I should not worry about what this day or this week has brought me.

I must reiterate: quantitative analysis does not make the psychological harshness of trading disappear. It does however reframe the pain narrative. When the pain narrative changes from "this strategy does not work" to "this is just a losing streak in an otherwise profitable strategy," you are far less likely to deviate and engage in self-destructive behaviors.

Knowing and Doing

Richard Dennis allegedly turned $400 into $200 million. He turned novice traders into millionaire traders with his Turtle trading system. He reportedly said that he could post his exact strategy on the front page of the New York Times and still people would fail to make money. Dennis did not have a deep cynicism for human intelligence or discipline, but he understood that a profitable strategy means virtually nothing without the correct behavioral techniques. Are you able to take a loss and not override the framework? Are you able to tolerate periods of drawdown without abandoning the approach? Are you able to maintain an emotional equilibrium, win or lose?

Dennis understood that even a statistically validated strategy is still prone to the human tendency to override it in the moment. If a perfectly validated strategy has produced five losses in a row, the brain does not register a binomial significance test. It registers, and quite strongly, the pain of five losses in a row. Cognitively, the tendency to want to deviate from a perfectly sound strategy will always be present.

In my own experience trading an account at $3M for an office, I saw a drawdown across several statistically validated strategies. In the span of two trading days, I had gone one win out of twelve trades. I was ripping my hair out and began feeling as if the size I was trading was somehow large enough to where someone, somewhere was targeting and fading my moves. There was, of course, nothing wrong with the strategy, and I was not trading a meaningful enough size for any of my positions to be targeted. Even with experience and intellectual understanding that my strategies were statistically sound, the narrative feelings of explaining why something has lost will always persist.

These behavioral techniques, being able to take a loss, overriding your impulses, acting equanimously in the face of immense uncertainty and pain, are not things that are learned through a trading course or a broker. These are things that are learned through exposure therapy coupled with deep introspection. The act of thinking does not sell well, which is why there is virtually no trading book that will tell you that introspection is a necessary skill.

You can still fail with a validated strategy. Knowing is not doing. Doing is far harder than knowing. Having a statistically validated framework can remove some of the ambiguity on how to operate a profitable system. But it does not remove the requirement for behavioral discipline. It does not teach you how to be equanimous when faced with a losing streak. It cannot teach you to sit with an inhuman amount of pain. No book, teacher, course, or article can. Only you can.

Market information is neutral. Candlesticks are visual representations of price on a time axis. There is no narrative attached to a candlestick. The narrative is made by the human brain. When a narrative fails to play out as the brain has predicted, the brain will experience the pain of financial loss as well as intellectual loss. You were wrong, and no one naturally likes being wrong. The operative idea to neutralize the pain of being wrong is understanding that you do not need to be right or wrong on this trade. You need to be right in the long run. On each individual trade, the result is largely unimportant.

Sit down with yourself and think about your character traits. How do you handle being wrong? How do you handle being confused? How do you handle sitting on your hands and doing nothing?

I was, for the longest time, a person that was unable to look at their mistakes. After a big losing day, I could not open the chart and see where my trades were placed. It absolutely disgusted me. This is not a trait that serves anyone well. You must see your mistakes and learn from them. For a long time, I was a person that could not sit on their hands, and I would impulsively take trades on a one-minute or two-minute chart, trades that were not according to plan.

Naturally, the feeling of disgust will follow when you look at all your mistakes. This is the most powerful feeling you can experience, as disgust is one of the strongest catalysts for transformation. For the longest time, I was a person disgusted by my own body. My ribcage was visible and I was so skinny that you could see the left side of my chest beat in tandem with my heart. I discovered the gym when I was around 16. I started taking care of my diet and started to take pride in training with intensity because I was so disgusted with myself. Fast forward six years later, and I have built a body which I am satisfied with. My nutrition and training are intuition now. I trained like I was possessed and force-fed myself meals. I recall one instance where I had a 1,500 calorie dinner, then woke up several hours later for breakfast. I was still full from the dinner which had not fully digested and threw up my breakfast, which I remade without hesitation. Achieving these behavioral modifications and tolerating the discomfort of training intensely and eating beyond what my body wants were manifestations of disgust that I had over what I used to look like.

I think about it this way: I am not as enamored by the beauty of a bouquet of roses, as I am enamored by avoiding getting pricked by its thorns. Imagining my dreams and imagining what life would be like with them is cool. Imagining what would happen if I did not change and I failed is terrifying. So terrifying that I would do anything to keep it from happening. If this meant that I had to sit on my hands and tolerate the boredom of staring at a minute chart, then so be it.

There is a myth of fearlessness in trading. I do not believe that as human beings we can ever truly rid ourselves of emotion. All we can do is build up the courage through repeated exposure to act against our own fears. The pain will always be present, regardless of how well-done a strategy's framework is. The key is acting equanimously. That, I believe, is the great secret to trading. It is not so much ridding ourselves of our human emotions, but rather acting in a way where they are not distinguishable from our actions. Much of human nature, our ingrained thoughts and behavior, is counter-productive in this environment. Overcoming this through self-control is the real challenge. It has and will prove to be a non-linear process.

Between Heaven and Hell

That is the great secret in trading: no matter how much statistical work you do, the pain and fear will never truly dissipate. It puts the trader on a tightrope between heaven and hell. On one side, there is the reality that happens where you give into your human nature and start lashing out on the markets, self-destructively. On the other side is an imaginative and idealistic world in which human emotions are able to be separated from a human brain.

Quantitative analysis cannot fully remove the balancing act, though it can dissipate some of the ambiguity. The rest of the balancing is up to the trader's will.

Build a theoretical framework, run statistical tests, and most importantly, learn to sit with an enormous amount of discomfort. Without doing so, no amount of math can save you.

Evaluation-based prop firms have exploded in popularity in the retail trading space. The business model is simple: take advantage of an enormous failure rate, add rules that make trading even more difficult, and for the lucky few, provide a demo account under the guise of live capital — of which only a fraction of funded traders ever receive a return on investment.

Trading is hard enough, and many are drawn to fictitious claims of live capital for an accredited firm. Make no mistake: these evaluation-based prop firms are not in the favour of the trader.

To demonstrate this, I am going to show a legal and repeatable method that allows a trader to generate positive expected value while having zero real market edge. The mathematical proof is impossible in a live brokerage account, but demonstrable — with limitations I will explain — in an evaluation-based prop firm.

Over this article I will build the mathematical proof, show the results of a Monte Carlo simulation across 50 accounts, and draw the uncomfortable conclusion about what prop firm evaluations actually test.

Methodology

For this example I am using TopStep and their 50K evaluation. The contract has three numbers that matter.

Evaluation stage: profit target $3,000, max loss $2,000. Funded stage: profit target $2,000 (10× the total cost), max loss $2,000. The asymmetry is visible before any math is written. Your maximum downside if you fail the evaluation is $49. Your upside if you complete both stages is $2,000.

The Strategy

The strategy I will trade to generate money from TopStep is net breakeven. It does not make money. It is barely a strategy at all and can be achieved by taking random trades with specific take-profit and stop-loss parameters.

Parameters: 67% win rate · +0.4925R per win · −1R per loss · 1R = $500

Expectancy is how much a strategy makes on average per trade over the long run. It is calculated as:

Breakeven before fees. The 67% win rate was chosen deliberately. A high win rate, small and frequent winners, and large occasional losers produces an equity curve that drifts upward on most days. The losses are less frequent. This shape is calibrated for gambler's ruin dynamics, which is what the evaluation structure rewards.

The Drunkard: Gambler's Ruin

The essential statistical concept underlying this proof is gambler's ruin.

Imagine a drunk walking in an alley with a wall to his left and a wall to his right. His steps are random. The probability he reaches the right wall before the left wall depends only on where he starts and the distance to each wall. His previous step tells you nothing about the direction of his next step. We are not predicting his path. We are calculating which wall he hits first. In our context: which comes first, the profit target or the max loss.

The evaluation and funded stage are two independent events. The joint probability of completing both is:

One in five attempts produces a $2,000 payout. The strategy has no edge. The payoff structure is doing all the work.

Expected Value: The Full Cost Model

Not every attempt costs $198. You only pay the $149 activation fee if you pass the evaluation. The expected cost per cycle properly weights each fee by the probability of incurring it.

The expected value of one attempt cycle using a strategy with zero real market edge is positive $291.40. Let me be direct about what this means: because of the asymmetric payoff structure that prop firms have created, you can make money with no real market edge. This is impossible on a live brokerage account. In a live account a zero-edge strategy produces zero expected return before costs. The evaluation structure is doing all the work.

The Monte Carlo

Doing this on one account is not very profitable — the odds are 80% against you on any single attempt. Thankfully, TopStep allows up to five accounts per user. For maximum absurdity, this experiment gathers a team of ten people with five accounts each — fifty accounts total — all taking two trades per day.

10,000 simulations were run. Before the table: the most important finding comes before the six-month numbers. The mean cycle length is 70 trades. At two trades per day and 21 trading days per month, each account has 42 trades available in a given month. Most accounts cannot complete a single full cycle in month one. They spend the entire month in the evaluation stage, paying $49 each time they fail. The positive expected value exists — but it requires enough time for full cycles to complete. Month one is almost certainly a loss. This is not a flaw in the strategy. It is the lag between deploying a positive-EV approach and accumulating enough completed cycles for that EV to express.

By month six, the expected value has fully realised. Across 10,000 simulations every single trial produced a profit. Even the worst 1% of outcomes cleared +$17,468. This profit was not driven by lucky runs but by the law of large numbers operating on a large number of completed cycles.

By month twelve, the mean profit across the ten-person team is $91,742. Ten traders extracted nearly six figures on average with a strategy that is breakeven. The mean profit per person across five accounts: $9,174.

The Limitation

TopStep is not naive. What we have just discussed is not an obscure corner of mathematics. Firms employ risk managers who monitor for exactly this pattern. Common countermeasures include maximum accounts per person, withdrawal restrictions, consistency rules requiring profits to be distributed across multiple days, and pattern recognition on high win-rate, small-winner account behaviour.

This does not dismiss the math. The fact that these countermeasures exist confirms the mathematical argument. You do not build risk management infrastructure around an exploit that does not work. TopStep would almost certainly detect similar trading patterns across ten people sharing the same IP address. But the fundamental payoff asymmetry holds as long as someone can maintain the cost structure without being detected and barred.

What This Actually Means

In a live brokerage account, a statistical edge is necessary for positive expected returns. There is no structure that transforms a zero-edge strategy into a profitable one. The market is indifferent to your payoff structure.

In an evaluation-based prop firm, the payoff structure itself is the primary driver of expected value. A trader with zero edge and a well-calibrated high win-rate strategy can still generate positive expected value from the evaluation structure alone.

Passing a prop firm evaluation is weak evidence of trading ability and strong evidence of surviving a constrained random walk. Many social media traders will display payout certifications while sharing nothing about what it took to get there. It also creates a dangerous feedback loop: a trader runs a strategy with no validated edge, receives a payout through luck, and is now emotionally anchored to that strategy. Deviating from it becomes psychologically difficult — despite the fact that it has no real predictive value.

Prop firm payouts are not proof that fair value gaps, liquidity sweeps, order blocks, or any pattern promoted online have predictive value. They do not hold up in a live trading environment. The evaluation rewards a specific payoff profile. Real markets reward expected value. The gap between those two things is the difference between an unprofitable and a profitable trader. This article quantifies that gap.

If there is one idea to take away from this article, it is that quantitative analysis is fundamental to profitable trading. You must know the performance of your strategy. You must backtest. You must be able to demonstrate, to a statistically meaningful degree, that your results are unlikely to be produced by chance. That process requires hundreds if not thousands of trades logged in a backtest. Do not trade strategies you find online without verifying them yourself. Virtually everyone who teaches trading strategies and sells courses is not a trader but a business person. They will substantiate their teachings with payouts from the lucky few — but as this article demonstrates, those payouts are rather meaningless.

This article is a direct continuation of The Drunkard. If you have not read it, read it first. The argument here only makes sense with that foundation. In that piece I demonstrated, mathematically and through Monte Carlo simulation, that evaluation-based prop firms can be beaten with a strategy that has zero real market edge. The conclusion I drew was that passing an evaluation is weak evidence of trading ability and strong evidence of surviving a constrained random walk.

Since writing it, I have come across data that makes me want to go further. I am not just saying these firms are mathematically exploitable. I am saying you should not touch them. Trading is hard enough. There is no reason to make it harder with a product that is designed to extract fees from the majority and simulate trading for almost everyone who survives.

What TopStep's Own Data Says

TopStep publishes their own trader performance statistics annually. For 2025, they disclosed four numbers. Each one tells you something specific about what this product actually is and who it actually benefits.

Start with the evaluation pass rate. The random walk model, using a zero-edge strategy calibrated specifically for the evaluation structure, predicts a 40% pass rate. The actual pass rate across all TopStep evaluations in 2025 was 16.8%. That gap is not a failure of the math. It is confirmation of what I argued in Between Heaven and Hell about the cognitive environment of trading. The average retail participant is not running a zero-edge strategy. They are running something worse. They are overleveraging after losses, cutting winners early, holding losers too long, and doing all of the things that the brain does when it is dropped into an environment it was not built for. The cognitive destruction I described in Between Heaven and Hell is not theoretical. It is showing up directly in TopStep's data. The 16.8% pass rate is the empirical fingerprint of the feedback loop problem.

The random walk model holds up precisely. It is not that the math is wrong. It is that the math only works if you actually run the zero-edge strategy mechanically, without deviation, without revenge trading, without emotional override. Virtually no retail trader can do that. The 23.2 percentage point gap between the model and reality is the exact size of the gap between knowing what to do and actually doing it under financial pressure. This is what Richard Dennis understood. This is why he said the rules on their own mean nothing.

Now look at the 33.3% funded payout rate. The random walk model predicts 50% for a zero-edge strategy with equal profit target and max loss. The actual number is 33.3%. Same explanation applies: the funded participants are not zero-edge. They are negative-edge, because the same cognitive failures that cause most people to fail the evaluation continue into the funded stage. The minority who passed the evaluation mostly did so by luck or by discipline they cannot sustain. Once in the funded stage, the cognitive environment reasserts itself and the failure rate climbs accordingly.

But the most important number in that disclosure is the one that almost nobody talks about: 0.71%.

Of every participant who passed the evaluation and reached a funded account, 0.71% were called up to a Live Funded Account where actual capital is at stake. The rest are trading a simulated account indefinitely. The payouts are real. TopStep does pay them. But the capital being traded is not live. 99.29% of funded traders are, at any given time, running a simulation with real financial consequences but no actual market exposure on TopStep's end. The risk to TopStep from a funded trader is not the risk of the trade going against them. It is the cost of the payout itself, which is funded by the evaluation fees paid by the 83.2% who never passed in the first place.

This is not a trading firm that is selective about who it backs with real capital. It is a fee-collection business that pays out a fraction of its fee revenue to the traders who survive long enough to claim it. That is the business model, stated plainly.

None of this is hidden. TopStep publishes these statistics voluntarily. The 0.71% figure is sitting in their own disclosure. Most retail traders who purchase an evaluation have never read it. Source: topstep.com/our-program

The Strongest Counterargument, and Why It Fails

The obvious counterargument is that 33% of funded traders receive a payout, and some of them must have genuine edge. That is true. But the 33% is a mixture. Some of those payouts belong to traders with genuine, validated, repeatable edge. Some belong to traders who got lucky during a short observation window and will blow up in the next evaluation cycle. The evaluation process, by design, cannot distinguish between them. A constrained random walk with the right payoff profile passes at 20% probability regardless of whether the trader has any real understanding of markets. The payout certificate looks identical in both cases.

The most honest number in the entire disclosure is not 33%. It is 0.71%. Those are the traders that TopStep itself decided had genuine edge, because they became a recurring business expense that needed to be reclassified. A trader with genuine, repeatable edge who is consistently receiving simulation payouts is a liability on TopStep's fee revenue. The payout has to come from somewhere, and it comes from the 83.2% who failed the evaluation. The longer that trader stays funded and profitable, the more that revenue is consumed.

The live funded account is the resolution to that problem. When a trader on a 50K account has built their balance to, say, $65,000, TopStep has the discretionary power to call them up. They pay out the $15,000 in profits. But the live funded account is structured differently. The trader now pays their own data fees. More importantly, the trader can withdraw profits freely at any time, without TopStep processing simulation payouts from fee revenue. The business expense disappears. The trader is now operating with real capital on their own terms, and TopStep no longer bears the cost of their success.

The 0.71% live funded figure is not evidence that TopStep is selective about backing traders with real capital. It is the most accurate answer to the question of how many people trading an eval prop firm actually have genuine, repeatable edge. After running every participant through an evaluation, a simulation funded stage, and a payout process over an entire calendar year, 0.71% cleared every bar. That is not a failure of the product. It is an honest reflection of how rare genuine edge is, and of the fact that the evaluation process is not designed to find it.

The Constraint Nobody Mentions

There is one more constraint worth stating plainly because it almost never appears in the marketing. The maximum drawdown on a 50K TopStep evaluation is $2,000. That is 4% of the account. The most lenient maximum drawdown I have seen on any evaluation-based prop firm is 10%. These constraints are so tight that they would disqualify most statistically validated systematic strategies during normal variance.

A strategy with a confirmed edge at the 95% confidence level will still produce losing streaks. That is what variance means. A 4% maximum drawdown means that a perfectly sound strategy, running exactly as expected, will frequently breach the limit during a short observation window. The evaluation is not testing whether your strategy makes money over the long run. It is testing whether your strategy avoids a 4% drawdown during a brief observation period. Those are not the same question, and one of them has very little to do with trading skill.

A systematic intraday futures strategy with a validated Sharpe ratio of 3.04 would fail most of these evaluations repeatedly during normal operation. Not because the edge is not there. Because the constraint is calibrated to extract fees, not to identify talent.

The Conclusion I Should Have Started With

Do not touch evaluation-based prop firms. Not because you cannot beat them mathematically. You can, under specific conditions that most traders cannot sustain behaviourally. Not because the payouts are fake. They are real. But because the product you are buying is not what it is marketed as.

You are not buying access to capital. You are buying access to a simulation, within constraints so tight they would disqualify most professional strategies, with a payout funded by the evaluation fees of the 83.2% who failed before you, and a 0.71% chance of ever trading real money. The business model is fee collection. The product is the illusion of a trading career.

Trading is already the hardest cognitive environment a person can operate in. The uncertainty is genuine, the leverage is real, the feedback loops are destructive, and the skill required to validate a genuine edge is considerable. None of that changes inside an evaluation. What changes is that you are now paying for the privilege of experiencing all of it within constraints designed to make you fail faster.

Real trading, on a real brokerage account, with real statistical validation, is a harder and longer road than any evaluation shortcut suggests. It is also the only road that goes anywhere.