Individuals/Companies Publicly Verifying Habits, Time-of-Day Edges, and Seasonality by Time (Minutes/Seconds)
(Overview of Case Study Sites and Representative Evidence / Long-read with Citations)
Introduction: Why Track Time-of-Day Edges by Minute and Second?
Markets often exhibit "Time-of-Day Effects" and "Seasonality," and performance can vary depending on what time and minute you enter or exit, even with the same trading logic Academic research has long demonstrated U-shaped patterns in volume and volatility and the peculiarities of the moments immediately after the open and before the close Meanwhile, practitioners and tool vendors provide visualization and verification functions based on minute bars which individuals can also use for backtesting and verification (listed below).
1) Visualization and Verification of "Time-of-Day/Minute Bars" by Practitioners and Tools
Market Chameleon: Visualization Tool for "Time-of-Day Performance" by Ticker
Market Chameleon, famous for options analysis, provides a tool that allows you to interactively check the price movement, win rate, and average return of ETFs and individual stocks for specific times of the day For example, you can see the average profitability of specific time slots like "SPY from X:XX to X:XX" at a glance.
“See how a stock performs at different times of the trading day.” (Summary of feature description)
Point: By overlaying SPY or related futures for the same time period as your own bot output (e.g., a list of PFs from 22:32 to 23:59), you can get a bird's-eye view of the "terrain you are fighting on." It is also possible to infer the impact of real-demand opening/closing needs, auctions, and rebalancing time zones.
Seasonax: Systematic Visualization of Seasonality (Calendar Days, Days of the Week, Months, Events)
Seasonax specializes in visualizing calendar seasonality While it is stronger in cycles longer than intraday rather than pure "minute bars," it is effective for grasping biases by day of the week, date, and event.
“Seasonality analysis helps traders identify recurring patterns in asset prices.” (Summary of definition)
Point: "Instantaneous edges" in minute bars are often hierarchically intertwined with longer cycles (e.g., month-end, week-start, employment statistics days, etc.) If you first grasp the tailwinds/headwinds of the higher-level cycle and then apply them to the minute bars, reproducibility is more likely to increase.
2) Academic Side: Representative Research Supporting the "Existence" of Time-of-Day Patterns
U-Shaped Patterns (Volume and Volatility)
Classic research on market structure has confirmed in numerous studies that intraday volume and volatility form a U-shape, being high at the open and close and low during the day.
“Intraday patterns show high activity near open and close.” (Summary of survey)
Implication: Near the open and close, slippage and volatility coexist. Because profits are easier to make but losses can also be larger, it is a "double-edged sword" where the PF (Profit Factor) tends to become extreme.
Intraday Trends in "Time-of-Day Commonalities" and Correlations in High-Frequency Data
Research using 1-second data suggests that "idiosyncratic factors are strong immediately after the market opens, and market-wide common factors become stronger toward the latter half."
“Opening hours are dominated by idiosyncratic risk; a market factor emerges later.” (Summary of abstract)
Implication: First half = strong single-name habits / Second half = strong index-linked characteristics. "Afternoon convergence and synchronization" can sometimes be seen in minute-bar-based time-of-day PFs.
Volatility prediction and "time-of-day effects" in machine learning
It is reported that Time-of-Day factors are useful even in volatility prediction using machine learning.
"We highlight time-of-day effects that aid the forecasting mechanism." (Abstract summary)
Implication: Seasonality of minute bars × volatility can be an important feature for predictors, so it is effective not only for PF improvement but also for risk allocation and size adjustment.
Decomposition of "Intraday vs. Overnight" (Relative evaluation of intraday edges)
In practitioner blogs and surveys, the phenomenon that much of the long-term return occurs overnight is repeatedly reported (e.g., SPY).
"A considerable portion of equity returns occurs overnight." (Summary)
Implication: "Intraday minute-bar edges" themselves are often relatively thin. Therefore, intraday time-of-day PFs tend to have higher reproducibility when conditioned by event days / market conditions / stock attributes.
3) Individuals/organizations that continuously publish "similar analyses" (site examples)
Market Chameleon: Visualization of performance by stock and time of day (above). You can check "which time of day is effective" for minute bars using a publicly available UI. Ideal as a first step for verification.
Seasonax: Systematically visualizes seasonality by day of the week, month, and specific date. While not pinpointed to minute bars, it is strong at grasping tailwinds/headwinds of higher cycles.
-
Quantpedia (Research Survey): Overnight vs. Intraday and seasonality-related strategies are continuously organized with summaries and literature links. Useful for understanding the background of minute-bar edges.
“Trading strategies related to intraday and overnight effects…” (Summary)
-
Academic Community (arXiv, etc.): Time-of-day structures of high-frequency data correlation and volatility and intraday seasonality are continuously published in the latest papers. Provides theoretical and statistical backing for quantitative verification of minute bars.
“Opening hours dominated by idiosyncratic risk…” (Time-of-day structure of high-frequency correlation)
“Time-of-day effects aid forecasting.” (Machine learning x Volatility forecasting)
Note: Personal blogs that continuously share trading edge biases at the pure “second-to-minute” level using publicly available long-term backtest results are in the minority. The reasons are:
Alpha degrades quickly (easily disappears due to publication/arbitrage),
Risk of being offset by slippage and commissions,
Exchange terms and restrictions on secondary data distribution.
Therefore, “public verification” in the form of commercial tools (UI disclosure) or academic papers (aggregation/statistical disclosure) is the mainstream.
4) How to utilize your “bot time-of-day PF ranking” (Implementation hints)
Cross-referencing with external benchmarks
Compare the average behavior of SPY/futures during the same time period using Market Chameleon to determine **whether the PF maximum/minimum zones of your own bot are 'derived from market conditions or unique to the logic'**.Two-stage optimization of higher-level seasonality × minute bars
First, select higher-level seasonality such as day of the week, end of the month, or around employment statistics using tools like Seasonax, then re-aggregate minute-bar PF only under those conditions (= conditional time-of-day PF).Model splitting for the first/second half of the market session
As research findings suggest, first half = strong individual factors / second half = strong market factors, so switch features and thresholds by time period (size adjustment, exit rules, reversal filters, etc.).Normalize risk by 'volatility × slippage'
To determine if the high minute-bar PF is a temporary spike, adjust it using the realized spread and average slippage at the same time to evaluate the true PF. Note that high-frequency research assumes time-of-day bias in intraday volatility.Generalization test: cross-asset expansion / recent rollout
Academic papers also suggest that **'time-of-day effects can be general-purpose features'**. Verify whether the PF is maintained in unlearned assets or recent periods through sequential rollout.
5) Common misconceptions and pitfalls (how to read public verification)
-
“High minute-bar PF does not equal arbitrageability.”
Trading fees, execution latency, and order book depth often cause it to vanish instantly. Always verify backtest tick-data fidelity. In academic terms, consideration for asynchronous trading and microstructure noise is a prerequisite.“Asynchronous trading and microstructure noise… may severely underestimate correlations.” (Abstract)
Do not ignore the decomposition from overnight
In market phases where much of the long-term excess return is explained by overnight performance, “low-margin, high-turnover trading with strict cost management” is essential for intraday minute-bar edges.Mixing event days and normal days
Friday closes, month-ends, index rebalancing, and employment statistics are examples where singular days distort the average. Use tools to incorporate event-day flags for re-aggregation.
6) Summary
“Minute/second-level” time-of-day habits are supported by classical U-shaped pattern research and the latest high-frequency/machine learning studies.
In practice, it is pragmatic to quickly identify average behavior by time of day using public UIs like Market Chameleon, and overlay higher-level seasonality using tools like Seasonax.
For your bot time-of-day PF ranking, the shortcut to solidifying reproducibility (out-of-sample) is to follow this flow: (1) External benchmark matching -> (2) Conditional re-aggregation of seasonality x minute bars -> (3) Model splitting into first/second halves -> (4) Cost adjustment -> (5) Generalization testing.
7) List
Ernie Chan (QTS Capital)
A leading expert in systematic trading. His blog and books cover 'intraday seasonality' and 'preprocessing of high-frequency data (triple barrier, etc.)'. He has deep ties with Hudson & Thames (MlFinLab) and provides extensive explanations on implementation.
Scribd
Robot Wealth (Kris Longmore)
A community for individual quantitative traders. They verify intraday patterns and demonstrate how to find edges using real data. Many articles deal with 'distortions' around market open and specific time zones.
VM155-207-203-236
Quantified Strategies (Oddmund Grotte)
Continuously publishes time-dependent statistics, focusing on SPY, such as 'times of day when highs and lows are likely to occur' and 'expected values for the first/last hour'. The methods can be applied directly to US100/US30 and spread trading.
Quantified Strategies
Quantpedia
A catalog that summarizes and implements edges from academia and blogs. They also publish workflows for automatically exploring new edges, which are easily applicable to time-based rules.
Quantpedia
QuantInsti (Blog/Blueshift)
Provides utilities and explanations for verifying data while removing time-of-day bias, such as 'intraday_seasonality_func' for deseasonalizing intraday seasonality.
QuantInsti Blog
Hudson & Thames (MlFinLab)
Emphasizes 'edge preservation' in machine learning preprocessing, feature engineering, and labeling (CUSUM, triple barrier). Introduces verification procedures robust to the non-stationarity of time zones.
Hudson Thames
Alpha Architect (& Academic Introductions)
Breaks down and introduces intraday momentum research, such as 'the first 30 minutes predict the last 30 minutes'. Provides hints for testing correlations between time zones using US indices.
Alpha Architect
Academic Open Access
Example: Intraday seasonality and non-stationarity of DJIA stocks (where statistics change by time of day). It suggests the premise that 'distributions shift by time of day' even in spreads and index futures.
PLOS
