Every trading system generates two kinds of outcomes: the trades you took and the trades you didn't. Most traders track the first category carefully — win rate, average gain, average loss, maximum drawdown. Almost nobody tracks the second category with equal rigor. This was a mistake I made for the first six months of running the system. I was logging every trade I took, but I was letting passed signals disappear without record. When I finally started tracking them, what I found was uncomfortable and instructive in roughly equal measure.
This post is about the signals I passed on, what the data showed when I analyzed them systematically, and what three specific cases taught me about the difference between good judgment and excessive caution.
Why Tracking Passed Signals Matters
The instinct not to track passed signals is understandable. If you didn't take the trade, there's no financial outcome to record. The decision is complete. Moving on feels natural.
The problem is that without tracking passed signals, you have no way to evaluate whether your pass decisions are adding value or subtracting it. You might be correctly filtering out low-quality signals — in which case your pass decisions are a genuine edge. Or you might be incorrectly filtering out high-quality signals — in which case your pass decisions are costing you more than your stop-losses. Without data, you can't tell which one is true.
I started tracking passed signals in month seven of live operation, logging the stock name, the signal grade, the reason I passed, and the subsequent price action. By month nine, I had enough data to start seeing patterns.
The overall finding was mixed: my pass decisions were better than chance but not as good as I had assumed. About 65% of the signals I passed on would have resulted in a loss if I had taken them — which sounds like good filtering. But about 35% would have been profitable, and the average gain on that 35% was significantly larger than the average gain on the signals I did take. My filtering was eliminating more losing signals than winning ones, but it was also eliminating some of the best signals, not just the mediocre ones.
The Three Cases That Taught Me the Most
Case One — The Semiconductor Equipment Stock I Passed Because the Sector Was Down
On a Tuesday morning in month eight, a signal fired on a mid-cap semiconductor equipment manufacturer at 09:47. The grade was A. Trading value was strong. Trade intensity was well above threshold. The pattern was clean — a sharp initial move, a measured pullback that respected the ATR threshold, and a breakout with strong volume confirmation.
I passed. The reason I logged at the time: the semiconductor sector index was down about 1.2% on the day, and I had a general rule against taking signals in stocks whose sectors were clearly underperforming the broader market.
By 11:30, that stock was up 18% from where the signal had fired. The sector had recovered to roughly flat by then. The stock's individual move had been completely disconnected from the sector's broader weakness — driven by a company-specific order announcement that had come out overnight and that I hadn't checked before making my pass decision.
What this taught me: my sector performance filter was poorly calibrated. I was applying it as a binary rule — sector down means pass — when the more useful question was why the sector was down and whether that reason applied to this specific stock. A sector decline driven by broad macro concerns is a different situation from one driven by a single large company's earnings miss that has no relevance to a smaller company in the same sector. I hadn't been making that distinction.
After this case, I changed the sector filter from a binary rule to a qualitative check: if the sector is down, investigate why before applying it as a pass reason.
Case Two — The Consumer Stock I Passed Because the Crossing Count Was Too High
In month nine, a consumer goods stock generated a confirmed primary signal at 09:33, followed by a secondary signal with a crossing count of 67 by 10:15. Based on the pattern I had identified — high crossing counts typically indicate consolidation rather than momentum — I passed.
The stock was up 11% by the close.
When I reviewed the case afterward, the high crossing count was there, but I had missed something in the timing distribution. Of the 67 crossings, 61 had occurred in a twelve-minute window between 09:40 and 09:52, when the stock was clearly consolidating around the moving average just after the primary breakout. After that cluster, there had been only 6 crossings in the subsequent hour and twenty minutes — each with strong volume — as the stock made a clean second leg higher.
I had looked at the aggregate count (67), seen a number above my informal "high count = consolidation" threshold, and passed without examining the timing distribution of those crossings. The rule I had developed — and described in an earlier post — was that high crossing counts combined with low volume multiples signal consolidation. The volume multiples on this stock's later crossings were strong. I had applied the rule based on the count alone and ignored the volume data that was supposed to be part of the evaluation.
This case reinforced something I already knew intellectually but clearly hadn't internalized in practice: rules designed to filter noise can themselves become sources of noise if applied without the full context they were designed to be used with.
Case Three — The Stock I Passed Because I Had Already Taken Two Losses That Morning
This case is the most uncomfortable to describe because it reveals something about my psychology rather than my analytical process.
In month ten, on a morning when I had already taken two stop-losses in the first hour, a grade A signal fired at 10:08. The setup was as clean as any I had seen that week. Trading value was exceptional. Participant breakdown showed strong institutional buying. The pattern had developed over 34 minutes with textbook structure.
I passed. The reason I logged was vague: "market feels uncertain today."
The stock gained 14% by the close.
When I reviewed this case a week later with the benefit of distance, what had actually happened was clear. The two earlier losses had created an emotional state that I had labeled "market uncertainty" to make it sound analytical. The market wasn't uncertain in any objective sense — the KOSPI was up 0.4% on the day, volatility was normal, and the signal that had fired was objectively strong by every metric I used. What was uncertain was my confidence, not the market.
This is the loss aversion pattern I described in the first post of this series — the same mechanism that causes traders to hold losing positions too long also causes them to avoid taking new positions after experiencing losses, even when those new positions are genuinely attractive. I had built the system specifically to eliminate this pattern from execution, and here I was reintroducing it through the manual review step.
After this case, I added a specific check to my review process: before logging a pass decision, I ask whether the reason I'm giving is descriptive of something observable in the market data or whether it's descriptive of how I feel. "Market feels uncertain" is the second kind. It's not a valid pass reason.
What the Passed Signal Data Showed After Six Months
By month thirteen of tracking passed signals, I had a dataset substantial enough to draw some conclusions.
The pass decisions that added value most consistently were those based on participant composition: signals where I passed because the participant breakdown showed exclusively retail buying with no institutional or foreign participation produced continuation moves less than 30% of the time — well below the base rate for confirmed signals overall. The composition filter was genuinely discriminating.
The pass decisions that added the least value were those based on market feel and recent loss experience. Signals I passed because I "felt" uncertain after previous losses performed almost identically to the signals I took in similar conditions. My loss-state-induced caution was not predictive of signal quality.
The pass decisions based on sector performance were intermediate: better than chance, but only when I had correctly identified that the sector weakness was relevant to the specific stock rather than applying the filter mechanically.
The most valuable outcome of tracking passed signals wasn't any of these specific findings — it was the discipline of treating the pass decision as equally subject to analysis as the entry decision. Before I started tracking, passing on a signal felt like a neutral act, a default. After tracking, I could see that pass decisions have outcomes too, and those outcomes need to be measured and evaluated the same way trade outcomes are.
How This Changed My Process Going Forward
The practical changes were three. First, I now require a specific, observable reason for every pass decision — not a feeling, not a vague impression, something I can point to in the data. Second, I review my pass decisions in the monthly review process with the same rigor I apply to my entry decisions. Third, I specifically flag any pass decision made within two hours of a stop-loss and apply additional scrutiny to the reason given, because I've found that post-loss pass decisions are most likely to be driven by emotional state rather than analysis.
The goal isn't to pass on fewer signals or more signals. The goal is for the pass decisions I make to be based on factors that are actually predictive of signal quality rather than factors that are predictive of how I'm feeling at that moment. The data has made clear that these two categories are not the same.
Today's Investing Insight — Opportunity Cost in Trading
Most traders think about trading risk in terms of the downside: the maximum loss on a position, the stop-loss distance, the worst-case scenario. Opportunity cost — the profit missed by not taking a trade — is harder to measure and rarely tracked with equal rigor. But in any strategy where the hit rate is above 50% and the average winner exceeds the average loser, the cost of missed opportunities is mathematically real and potentially larger than the cost of realized losses. This doesn't mean you should take every signal — discipline about selectivity is part of what makes a strategy work. It means that excessive caution and insufficient selectivity are symmetric errors, not one-sided ones. A process that accurately measures both — tracking not just what you lost on bad trades but what you missed on good ones you didn't take — gives a more complete picture of where the strategy's real costs are being incurred.
---
This post documents a personal journey of building and running an algorithmic trading system and reflects personal experience and perspective. Past signal outcomes do not predict future results, and all investment decisions and their outcomes are the sole responsibility of the investor.
Comments
Post a Comment