Every signal detection system faces the same fundamental tension: the more conditions you require to fire a signal, the fewer signals you get, but the ones you get tend to be more reliable. The fewer conditions you require, the more signals fire, but they include more noise. After running the system for several weeks and reviewing which signals led to meaningful price moves and which ones didn't, I built a grading framework to make this tradeoff explicit rather than leaving it implicit. This post explains how the A, B, and C grade system works, what it's based on, and what it actually revealed in practice.
Why a Single Pass/Fail Threshold Wasn't Enough
The initial version of the system used binary filters: trading value above a threshold, trade intensity above a threshold, tick acceleration above a threshold. If all three were met, a signal was generated. If any one was below its threshold, no signal.
The problem with binary filters is that they treat a stock that barely meets all three thresholds identically to a stock that crushes all three by a wide margin. A stock with 110 billion KRW in trading value, trade intensity of 121, and tick acceleration of 5.1x would generate a signal. A stock with 450 billion KRW in trading value, trade intensity of 180, and tick acceleration of 15x would generate the same signal. Both pass. But anyone looking at the raw numbers knows these are completely different situations.
After reviewing several months of signal data, the pattern was clear: signals where all three indicators were significantly above their thresholds led to sustained price moves far more often than signals where the indicators were just barely above the cutoff. The grade system was designed to capture that distinction.
How the Three Grades Are Defined
The grading framework uses the same three indicators — trading value, trade intensity, and tick acceleration — but applies them in an AND structure: all three conditions must be met for a grade to qualify. Meeting only one or two conditions, no matter how strongly, doesn't move the grade up.
An A-grade signal requires trading value above 30 billion KRW, trade intensity above 150, and tick acceleration of at least 10 times the recent average. These thresholds represent a level of simultaneous activity that I found, through actual data review, to coincide with the strongest subsequent price moves. When all three are at this level at the same moment, it typically means a large amount of real capital is moving aggressively into the stock right now — not drifting in gradually.
A B-grade signal requires trading value above 10 billion KRW, trade intensity above 120, and tick acceleration of at least 5 times the recent average. This is still a meaningful signal — real money is moving, buying is dominant, and execution is accelerating — but the intensity is lower than A-grade. B-grade signals produced good subsequent moves in trending market environments but were more likely to produce reversals in choppy conditions.
Everything below these thresholds receives a C grade. C-grade signals are worth watching but not acting on without additional confirmation. In practice, I found C-grade signals nearly as likely to reverse as to continue.
What the Data Actually Showed
After tracking signals and their subsequent outcomes for several weeks, the grade distribution told a clear story.
A-grade signals were rare — typically one to three per day across all tracked stocks. But when they appeared, the price continued in the direction of the breakout more than two-thirds of the time in the hour following the signal. The average subsequent move on A-grade signals was roughly three to four times the size of the average move on C-grade signals that happened to meet the binary filter threshold.
B-grade signals were more frequent — usually five to ten per day. Their hit rate was lower than A-grade but still meaningfully above random. The key variable for B-grade signals was the broader market environment: in days when the KOSPI was trending strongly, B-grade signals performed nearly as well as A-grade. In sideways or declining market sessions, B-grade signals frequently produced brief pops followed by reversals.
C-grade signals were the most common and the least useful for direct action. What I found C-grade signals most useful for was as early warning — a C-grade signal on a stock early in the session sometimes preceded an A or B-grade signal on the same stock later in the day, as conditions built up.
The Insight About Tick Acceleration
Among the three indicators, tick acceleration turned out to be the most time-sensitive and the hardest to interpret in isolation. A reading of 15x means the stock is executing trades fifteen times faster than its recent average — which sounds dramatic, but can also mean the recent average was extremely slow and the current rate is merely normal.
To manage this, the tick acceleration calculation uses a rolling window of recent trade timestamps, comparing the most recent interval to the average of a longer historical window. The key insight from real data was that tick acceleration readings above 10x were meaningfully predictive only when they coincided with high absolute trading value. When a stock with minimal trading value showed 20x tick acceleration, it almost always reflected a few large orders hitting a thin book rather than genuine mass participation. When the same 20x reading appeared alongside 50 billion KRW in trading value, it was a genuinely different situation.
This is why the grade system requires all three conditions rather than any one or two. Each indicator alone is susceptible to misreading. Together, they filter for a specific combination that's much harder to produce from noise.
Today's Investing Insight — How Institutions Signal Their Presence Through Volume
One reason trading value and trade intensity matter so much as combined signals is that large institutional participants can't hide their activity as effectively as they sometimes try to. An institution building a large position in a stock tries to spread its buying across many small orders to avoid moving the price. But above a certain size, the sheer volume of those orders starts showing up in aggregate — trading value rises, trade intensity tilts toward the buy side, and execution speed increases as limit orders get absorbed. Experienced traders learn to read these footprints not as individual data points but as patterns: a stock where institutional accumulation is happening often shows a rising floor of trading value over multiple sessions, with occasional acceleration spikes when the institution becomes more aggressive. The grade system, particularly the A-grade threshold, was calibrated partly around the profile of what institutional buying actually looks like in the data.
---
This post documents a personal journey of building an algorithmic trading system in the Korean stock market and is not a recommendation of any specific stock or strategy. Signal grades are quantitative summaries of market data at a specific moment and do not predict future price movements. All investment decisions and their outcomes are the sole responsibility of the investor.
Comments
Post a Comment