Five Painful Lessons From a Decade of Losing Trades
"The most expensive lessons are not the ones where you did something obviously wrong. They are the ones where you did most things right and got exactly one thing wrong, and that one thing turned out to be structural. Those lessons are worth writing down."
I wrote a broader decade-in-the-markets post a week ago that covers the arc of ten years of trading in one sitting. This one is narrower and more specific. It is five particular losing trades I actually took, in enough detail that you can see the shape of the mistake without me having to abstract it into a rule.
None of the five are the 2019 options blowup, because that one is already covered in the decade post. These are the quieter losses. They did not break my account. They just each cost me something I was not planning to pay, and each of them taught me something I did not know I needed to learn. In roughly the order that I learned them:
📋 The five lessons
- 2018 — Sizing on conviction, not on structure (KOSDAQ small-cap)
- 2023 — Trading through work exhaustion (MNQ futures)
- 2020 — Copying a mentor without their bankroll (crypto altcoin)
- 2020 — Holding past the peak for "one more rally" (XRP on Upbit)
- 2020-2022 — Confusing a good year with a good system
Lesson 1 — 2018: Sizing on Conviction, Not on Structure
Late 2018, KOSDAQ. I had spent about three weeks building a workbook on a small-cap Korean industrial name whose Q3 earnings were about to come out. The workbook was thorough. Peer comparison, order-book analysis, a scan of the company's IR presentations going back four quarters. By the time earnings week arrived I was, in my own head, an expert on this specific name.
Because I was so certain, I sized the position at roughly three times my normal risk. My normal position at the time was maybe 1% of the account. This one went in at closer to 3%. I told myself the sizing was justified by the depth of the research. What I was actually doing was sizing on conviction rather than on structure, and those are two very different things.
Earnings missed. Not catastrophically — the miss was about 8% below consensus, in a name that had already run up 22% into the print. The stock gapped down 14% at the open. I was down more than a normal week of risk in a single overnight session. I closed the position at the open because holding was going to compound the sizing error, not fix it.
The loss itself was survivable. What was not survivable — and what took me another two years to actually internalize — was the mental model that had allowed the sizing to happen in the first place. I had confused having done the research with having a structural edge. Research tells you what you know. Structure tells you what the trade will actually do to your account across a distribution of outcomes. They are not interchangeable. The three-week workbook did not change the base rate for post-earnings gaps on small-cap KOSDAQ names in the direction of the miss. It just made me feel like it should.
The rule I eventually wrote down: sizing is a function of structure, not conviction. If I cannot express why one trade deserves 3x sizing in terms of expectancy, R-multiple, or Kelly fraction — not in terms of how much I like the setup — then it does not deserve 3x sizing. This rule has cost me a lot of upside on trades I felt very good about. It has also stopped me from ever running the 2018 trade again.
Lesson 2 — 2023: Trading Through Work Exhaustion (MNQ)
A Tuesday afternoon in autumn 2023, Ingolstadt. I had been at my desk since 8 AM debugging an embedded systems issue that had been open for three days. Around 15:30 CET — right at the NY Open — I opened a MNQ (Micro E-mini Nasdaq) long on a setup I had marked the night before. The setup was fine. The pre-market plan was fine. Everything about the trade would have been fine if I had been paying attention.
I was not paying attention. I was in the second monitor debugging a race condition, and the MNQ chart was on the third monitor, and I was context-switching every 90 seconds. I saw the price move against me about eleven ticks and my first thought was not "check the invalidation level," it was "I do not have bandwidth for this right now." I moved the stop to break-even to shut off the noise. The next candle stopped me out at break-even. The candle after that ran to my original target.
The dollar cost was almost nothing. I did not lose money on the trade. What I lost was the expectancy of a valid setup because I was not present enough to let it play out. Over a large sample, that is the same as losing money — it is losing the win rate that justified the setup in the first place.
There is a category of embedded-systems and defense-industry work where the debugging session carries context in your head the same way a chess position does. You can technically be at the desk, technically watching the chart, and still be unable to think about anything except the shape of the bug. That is the state I was in. I should not have been trading. The setup was correct. The trader was not.
The rule I wrote down after this one: if I am mentally holding an open work problem, I do not have an open trade. Not "I trade smaller." Not "I set tighter stops." I do not trade. This is a hard rule and it costs me perhaps 30% of the setups I would otherwise take in a demanding work week, which is exactly the cost of not repeating the 2023 mistake at scale. I wrote about the fuller version of this dynamic in Why I Paused Live Trading.
Lesson 3 — 2020: Copying a Mentor Without Their Bankroll
Spring 2020, the crypto bull market was starting to become obvious even to people who had been ignoring it. I found a YouTube channel run by someone who had clearly been in crypto for years, whose analysis I found sharp, and whose portfolio disclosures — voluntary, on stream, roughly quarterly — showed him running a mid-six-figure crypto book across maybe forty positions.
When he took a position in a specific altcoin and explained his thesis, I took the same position. That part is not the mistake. Reasonable people learn from reasonable people, and following a demonstrably competent analyst into a specific trade is a defensible starting point. The mistake was copying the trade at the wrong fraction of the wrong portfolio.
What was 4% of his six-figure book was more like 35% of mine. The absolute dollar amount was smaller. The relative exposure was catastrophically larger. When the altcoin behaved exactly the way his thesis said it might — a 40% drawdown before an eventual recovery — he sat through it easily because it was 4% of his portfolio and he had thirty-nine other positions to think about. I could not sit through it, because 35% of my portfolio was in flames, and I did not have the psychological cushion of thirty-nine other positions dampening the noise.
I closed at roughly the low. He held to full recovery and beyond. Same trade, same thesis, opposite outcome. The trade was not wrong. The proportionality was wrong. And the proportionality was wrong because I had copied the setup without copying the two things that made the setup survivable for him: the bankroll depth and the position count.
The rule from this one: never copy a trade without also copying the portfolio context that makes the trade survivable. If I cannot replicate the mentor's sizing as a percentage of my own book, I cannot replicate their psychology under drawdown either. The trade is not the same trade if it is 35% instead of 4%.
Lesson 4 — 2020: Holding Past the Peak for "One More Rally" (XRP)
🔼 The peak I did not sell (November 24, ~880 KRW) and the eventual exit five weeks later (~365 KRW). The trade was profitable. The opportunity cost was not.
By early November 2020 I had been dollar-cost-averaging into XRP on Upbit for four months. Eleven buy orders, spread across late July through early November, at prices between roughly 260 and 370 KRW. Weighted by size, my average cost worked out to around 289 KRW — the single largest order was 4 million KRW at 260 KRW, which pulled the true cost basis toward the low end of the range. It was a meaningful position for me at the time, but sized to survive a full drawdown.
Then November happened. XRP tripled in three weeks. My screen went from "quietly patient accumulation" to "best trade of the year, on paper." The peak on November 24 was around 880 KRW — a paper gain of roughly 204% on the weighted cost. My pre-defined exit rule was clear: trim at least half if the position doubles within a month. The position had done more than double. It had, briefly, tripled.
I did nothing.
Every candle at the top I told myself the same thing: "Just wait for one more push. The fundamentals are here. The news flow is here. Selling now leaves money on the table." That is the same "one more candle" logic that ruins short trades that ignore stops. On a long trade it manifests as "one more rally." The underlying error is identical — the in-the-moment version of me overrode the pre-committed version, because the pre-committed version had never seen a number this pretty on the screen.
November 25-26 the reversal started. I still did not act — "it will come back." Through early December the bounces got weaker each time. Then on December 22, 2020 the SEC filed suit against Ripple, and XRP gapped down on the news. The next morning — December 23 — I finally closed the entire position in a single sell order at 365 KRW, four months of patient accumulation liquidated on a headline I had not planned for. A realized gain of roughly 26% net of Upbit fees, against a paper high of 204%. The trade was, technically, profitable. In expectancy terms, it was catastrophic — roughly nine-tenths of the available reward was left on the table because I refused to execute an exit rule I had written down before entering, and the eventual exit was triggered by a news reaction rather than the plan.
The rule from this one, cleaner than any I have written since: exit signals are non-negotiable, whether they are stops on shorts or take-profit targets on longs. The narrative for holding past either is always the same shape — the "one more" argument. The version of me setting the rule beforehand knew better than the version of me watching the peak in real time. The whole purpose of a written exit rule is to protect the former from the latter.
Lesson 5 — 2020-2022: Confusing a Good Year With a Good System
This one is not a single trade. It is a pattern that ran across two years, and the lesson only became visible when the market environment stopped being cooperative.
Through most of 2020 and the first half of 2021, my crypto positions were up substantially. I was executing what I thought of as a "system" — a rough combination of altcoin selection based on GitHub activity, project narrative, and technical breakout levels. It worked. Every month, more or less. When I compared my P&L to the benchmark I had picked at the start of the year, I was ahead of it. When I compared my P&L to my own expectations, I was well ahead of those too.
I updated my mental model of myself accordingly. I was, in my head, someone who had figured out crypto. Not perfectly, but well enough that the results should be repeatable. I started planning around this — position sizing that assumed continued outperformance, spending patterns that assumed continued gains, conversations that quietly implied I knew something other people did not.
The second half of 2021 through most of 2022 corrected every one of those beliefs. The same setups I had been running for eighteen months stopped working. Not gradually — abruptly, as the environment shifted from "everything crypto-related trends up" to "everything crypto-related bleeds sideways or down." The system did not change. The environment changed. And once the environment stopped covering for the system's weaknesses, the weaknesses were suddenly visible everywhere.
The takeaway from this one — and it is the takeaway that shows up in almost every long-form piece I have written since — is that a good year in an obvious environment tells you almost nothing about your skill. It tells you about the environment. The signal about your skill lives in the neutral or hostile years, when the tide is not lifting all boats. If your process only produces good outcomes when the environment is friendly, then the "process" was mostly the environment.
I did not know this in 2020. I know it in 2026. The intermediate step — a two-year period where the record was corrected, publicly to me if not to anyone else — was not fun, but it was probably the most useful thing that has happened in my trading career. It is also why I now weight the boring, neutral, non-parabolic years more heavily than the exciting ones when I try to evaluate whether my current framework actually works.
What Actually Links These Five Lessons
🔼 Eleven buy decisions across four months. One sell decision — the entire position, in a single order, five weeks too late. What the log records is the actions. What it does not record — what should have been written down alongside it — is the pre-committed exit plan I refused to follow.
Looking at the five together, they are all variants of the same underlying failure mode: the trader in the moment is a different person than the trader who designed the trade. Every one of these losses happened because I let the in-the-moment version override the pre-committed version.
- Lesson 1 — the in-the-moment version thought conviction should override the sizing rule.
- Lesson 2 — the in-the-moment version was cognitively depleted and skipped the "am I present" gate.
- Lesson 3 — the in-the-moment version copied the trade but not the survivability of the trade.
- Lesson 4 — the in-the-moment version bargained past the pre-committed take-profit.
- Lesson 5 — the in-the-moment version generalized from a small friendly sample to a big claim about skill.
The pre-committed version — the one that had time to think, and no money on the line — always had the correct answer. The in-the-moment version had different priorities. This gap is not solved by "being more disciplined." It is solved by designing the process so that the in-the-moment version has less scope to override the pre-committed version. Written stops. Fixed position sizing. Pre-defined go/no-go checklists. Time windows outside of which live trading is not allowed. These are all attempts to shrink the moment-to-moment discretion available to the version of me that is worst at using it.
None of that is original. Most of it is in the standard trading-psychology literature — Douglas, Steenbarger, and others cover this territory well. What is not in the literature is the specific loss you had to take to actually believe it. Each of these five is that loss for me.
A Note on Why I Am Publishing This
I hesitated over whether to write this post at all. The trading corner of the internet has enough content that leans on other people's losses for engagement, and I did not want to add to it. What tipped it into "worth writing" was the realization that every one of these lessons cost me real money, and if any of them save one reader from taking the same specific mistake at the same specific stage of their journey, that is a useful thing to have published.
None of these five losses should be read as universal warnings. Sizing on conviction is only a mistake if you do not have a structural reason for the sizing. Trading through work exhaustion is only a mistake for people whose work exhausts them cognitively rather than physically. Copying a mentor is only a mistake if you do not copy the portfolio context. Ignoring a stop is only a mistake if you have not pre-defined an override protocol. Confusing a good year with a good system is only a mistake if you draw conclusions from a small sample.
You will have your own five, or twenty, or fifty. Some of them will overlap with mine. Most of them will not. The point is not to internalize my list. The point is to write your own down after they happen, in enough detail that the version of you who reads them a year later can actually see the shape of the mistake.
For a broader map of where these lessons fit in the ten-year arc, A Decade in the Markets is the long-form companion. For the technical framework that eventually replaced most of these mistakes, The Institutional Volume Framework pillar post is where the process is documented. And for a general primer on the psychology of risk that overlaps significantly with what these five failures point at, Investopedia's overview of risk management is a reasonable starting point.
Thanks for reading, and see you at the next chart.
📚 Related Reading on This Blog
If these five landed for you, the companion pieces on this blog continue the same threads:
- A Decade in the Markets — the ten-year arc these losses sit inside
- Why I Paused Live Trading — the 2025 audit that made all five visible together
- The Institutional Volume Framework — the process that replaced most of these mistakes
- The 5-Layer Chart Framework — the setup filter that catches Lesson 1 and Lesson 4 before entry
⚠️ Educational Disclaimer
This post is a personal reflection for educational and informational purposes only. It is not investment advice, financial advice, trading advice, or a recommendation to take, avoid, or manage any specific trade or trading style.
Trading involves substantial risk of loss and is not suitable for every investor. My personal losses are specific to my own situation, and nothing described here should be read as a warning that generalizes automatically to your instrument, timeframe, or risk profile.
Always do your own research, consult a qualified financial advisor licensed in your jurisdiction, and never risk capital you cannot afford to lose. See our full Disclaimer and Privacy Policy.
About the Author
Dongmin Park is a software engineer with over 15 years in embedded systems (automotive and defense industries) and 10+ years of active trading across Korean equities, US options, MNQ futures, and crypto. He started trading on a Kiwoom Securities account in Seoul in 2016 and now lives in Ingolstadt, Germany, after relocating in 2022.
Coder Trader is an ongoing project to document where systematic engineering discipline meets discretionary trading. Say hi on X, look at the code on GitHub, or email hello@codertrader.com.
Comments
Post a Comment