BK · The Bank of New York Mellon Corporation
Stock Universe · 全球股票市场Quantitative Strategy Performance Analysis: BK · BK 15-Minute Algorithmic Trading Model
This analytical report evaluates the historical performance of an algorithmic trading strategy developed for BK in the stock market over a 7.02-year testing history from September 16, 2019, to September 23, 2026. Operating on a 15-minute timeframe, the strategy generated a cumulative historical return of +2248.22% across 107 completed trades, achieving a high win rate of 74.77% and a profit factor of 2.50. Despite evaluating price data on a granular intraday interval, the strategy exhibits long holding durations, averaging 420.12 hours (17.5 days) per trade with an average exit spacing of 23.96 days. Risk metrics reveal a maximum historical drawdown of 20.26% and a maximum losing streak of just 2 trades. Profit generation is broadly distributed across positions, with the three largest winning trades accounting for only 12.68% of gross profit. Recent performance since June 1, 2024, comprises 27 trades generating a cumulative return sum of +129.64%, confirming sustained consistency in recent market regimes. Ranking second for the BK ticker with a DevioLab Score of 74.35, this strategy offers a compelling decision logic model for automated trading bots and trading agents tailored for multi-week stock positioning.
分析日期： 23.09.2026. 策略数据单独更新。阅读完整分析
Strategy profile
The quantitative strategy under review operates on BK within the stock market framework, utilizing a 15-minute candle interval for price evaluation. Across its backtested evaluation period from September 16, 2019, to September 23, 2026—spanning 7.02 years—the strategy executed 107 completed trades. Over this sample, it achieved a historical closed-trade sum of +2248.22%, corresponding to an annualized rate of 218.92%, with a profit factor of 2.50. It holds a DevioLab Score of 74.35 and is ranked second among quantitative strategies evaluated for the BK ticker. The model is structured to serve as execution decision logic for automated trading bots, trading robots, and quantitative trading agents. Rather than engaging in micro-level intraday churning, the strategy leverages the high resolution of 15-minute data to refine entry timing while maintaining a multi-week position holding horizon.
Trading rhythm and position duration
Despite operating on a 15-minute chart resolution, the statistical profile of this strategy reflects a patient swing trading model rather than high-frequency execution. The average completed trade holding duration is 420.12 hours (approximately 17.5 days), while the median holding duration stands at 264.00 hours (11.0 days). This positive divergence between average and median holding times indicates that while half of all positions resolve within 11 days, selected strong trends are held significantly longer, stretching the overall mean. Execution frequency is sparse: the strategy averages 15.24 completed trades per year, or approximately 1.65 trades per active month. The temporal spacing between trade exits is similarly stretched, with an average of 23.96 days and a median of 16.51 days between exits. For automated trading agents and deployment engines running this strategy, long multi-week intervals between signals are common, requiring operational patience during extended market exposure.
Quality of historical results
The strategy demonstrates high payoff consistency and statistical balance across its 107 completed historical trades. Out of these, 80 trades were profitable while 27 resulted in losses, producing a win rate of 74.77%. The overall profit factor of 2.50 indicates that total gross profits were two and a half times larger than total gross losses. Notably, the mathematical expectations for individual trade returns show exceptional alignment: the average trade return is +3.37%, while the median trade return is +3.49%. In historical backtests, when median trade expectations closely track or slightly exceed average return metrics, it indicates that total returns are driven by structural consistency across trades rather than being skewed by a few extreme statistical anomalies. Furthermore, the top three winning trades combined generated only 12.68% of the strategy's total gross profit. The best single trade yielded +23.10%, while the worst single trade incurred a loss of -13.99%. This limited profit concentration confirms that equity accumulation was broad-based across the majority of the 80 winning trades.
Risk, drawdown and losing behavior
Risk characteristics for the strategy reveal a controlled drawdown structure relative to its cumulative yield. The maximum peak-to-trough historical drawdown recorded across the 7.02-year dataset was 20.26%. Comparing this maximum drawdown to the worst historical trade loss of -13.99% reveals that severe equity pullbacks were caused either by a combination of consecutive losses or open position equity pullbacks prior to trade exit. However, losing streaks were historically brief: the strategy's longest losing streak did not exceed 2 consecutive trades. In contrast, the longest winning streak reached 16 consecutive profitable trades. The asymmetric relationship between a maximum losing streak of 2 trades and a maximum winning streak of 16 trades underscores strong trend retention. Nevertheless, the worst trade loss of -13.99% demonstrates that tail losses can be substantially larger than the average trade gain of +3.37%, meaning position sizing and stop enforcement remain critical parameters when embedding this algorithm into automated trading robots.
Behavior through time and yearly stability
An examination of the yearly breakdown from 2019 through 2026 reveals exceptional chronological stability, with positive net trade returns recorded in every single calendar year of the dataset. In 2019, across a partial year sample of 2 trades, the strategy returned +9.08% with a 50.00% win rate. The strategy experienced its highest volume of activity in 2020, executing 23 trades with a 73.91% win rate and delivering a net return sum of +101.17%. In 2021, 18 trades yielded +68.29% with an 83.33% win rate. During 2022 and 2023, return sums moderated to +17.06% (18 trades, 66.67% win rate) and +26.16% (16 trades, 62.50% win rate), respectively, reflecting shift patterns in market behavior while maintaining positive edge. Performance re-accelerated in 2024, yielding +45.57% across 6 trades (83.33% win rate), followed by a perfect 100.00% win rate in 2025 across 10 trades for a sum of +52.62%. In 2026, 14 trades produced +40.76% with a 71.43% win rate. Win rates remained above 60% in every full year, showing structural resilience across changing market regimes.
Recent period since 2024-06-01 versus full history
Analyzing recent data since June 1, 2024, offers insight into the current validity of the strategy logic. Across this specific evaluation window, 27 trades were closed, generating a cumulative return sum of +129.64%. These 27 recent trades represent 25.23% of all 107 historical trades executed since September 2019, reflecting an increase in exit frequency during the most recent two-year span compared to the multi-year average. The performance during this period aligns strongly with the long-term track record, demonstrating that the underlying statistical edge on BK has not degraded in recent market conditions. The substantial recent return sum of +129.64% indicates that the algorithm continued to capture major price trends effectively, validating its relevance for modern automated trading agents.
Strengths and limitations
The primary strength of this quantitative model lies in its high win rate of 74.77%, low profit concentration (12.68% in top 3 winners), and consistent yearly profitability across all seven recorded years. The maximum losing streak of just 2 trades provides psychological and structural durability during drawdowns. The median trade return (+3.49%) exceeding the average trade return (+3.37%) highlights an exceptionally reliable distribution of winning trades. On the limitation side, the strategy features a low total sample size of 107 trades over 7.02 years, averaging only 15.24 trades per year. This low transaction volume requires statistical caution, as small samples can increase uncertainty regarding future distribution parameters. Additionally, holding positions for an average of 420.12 hours exposes capital to overnight gap risk and multi-week market volatility, while the worst single trade (-13.99%) indicates that losing exits can be relatively severe when stop boundaries are hit.
DevioLab analytical conclusion
With a DevioLab Score of 74.35 and a rank of second for the BK asset, this 15-minute algorithmic strategy displays a robust historical profile characterized by steady equity accumulation, low trade frequency, and strong return symmetry. The strategy effectively bridges short-term chart monitoring with medium-term trend capture, keeping trades open for an average of 17.5 days while maintaining a profit factor of 2.50 and a maximum drawdown of 20.26%. The absence of profit concentration in extreme outliers combined with sustained profitability in recent sample windows (2024–2026) reinforces its quality. For quantitative developers and traders seeking a low-frequency, high-win-rate logic framework to power automated trading bots or trading robots, the strategy presents a highly stable historical foundation.
Data scope and methodology
The statistical findings in this report are based strictly on simulated historical closed trades for BK between September 16, 2019, and September 23, 2026, using a 15-minute candle evaluation timeframe. All metrics, including the total historical profit of +2248.22%, average trade return of +3.37%, profit factor of 2.50, and recent since-2024 metrics, are derived exclusively from completed closed trades present in the source dataset. The recent window since June 1, 2024, includes all trades whose sell timestamp occurred on or after that date. Statistics do not account for live order execution friction, exchange fees, slippage, borrowing costs, or capital leverage unless explicitly noted. Historical backtested performance is strictly analytical and does not guarantee future results.