We present FinRL-X, a modular and deployment-consistent trading architecture that unifies data processing, strategy construction, backtesting, and broker execution under a weight-centric interface. While existing open-source platforms are often backtesting- or model-centric, they rarely provide system-level consistency between research evaluation and live deployment. FinRL-X addresses this gap through a composable strategy pipeline that integrates stock selection, portfolio allocation, timing, and portfolio-level risk overlays within a unified protocol. The framework supports both rule-based and AI-driven components, including reinforcement learning allocators and LLM-based sentiment signals, without altering downstream execution semantics. FinRL-X provides an extensible foundation for reproducible, end-to-end quantitative trading research and deployment. The official FinRL-X implementation is available at https://github.com/AI4Finance-Foundation/FinRL-Trading.
Figures & tables
Feature
FinRL-X
Qlib
TradingAgents
Zipline/Backtrader
QuantConnect Lean
Primary Orientation
End-to-End System
ML Research
Agent-Based Trading
Backtesting
End-to-End Platform
Broker Integration
Yes
No
No
No
Yes
Deployment-Consistent Interface
Yes
No
No
No
Partial
Reinforcement Learning Support
Yes
Limited
Yes
No
Partial
Modular Strategy Pipeline
Yes
No
No
No
Partial
Portfolio-Level Risk Overlay
Yes
No
No
No
Partial
Table 1: Comparison of FinRL-X with representative open-source quantitative trading platforms.
Figure 1: FinRL-X Framework: A layered, end-to-end trading architecture that unifies data processing, strategy construction, backtesting, and broker-integrated execution within a consistent pipeline, illustrating the workflow from data ingestion to live execution.
Strategy
Cum. Return
Ann. Return
Ann. Vol
Sharpe
Sortino
Calmar
Max DD
DD Duration
SPY
2.63
0.14
0.17
0.84
0.76
0.60
-0.23
23
QQQ
3.61
0.19
0.20
0.95
0.93
0.60
-0.33
23
KAMA
2.40
0.12
0.21
0.57
0.60
0.43
-0.29
23
MeanVar (No Timing)
2.19
0.11
0.22
0.53
0.56
0.37
-0.31
41
MeanVar (With Timing)
2.59
0.14
0.19
0.74
0.76
0.52
-0.27
38
MinVar (No Timing)
2.49
0.13
0.21
0.63
0.68
0.47
-0.28
25
Table 2: Performance and risk metrics across benchmarks
Figure 2: Ablation study of DRL-based allocation with and without timing adjustment. Incorporating the timing module improves cumulative performance and moderates drawdown relative to both the base DRL strategy and the SPY benchmark.
Metric
Rolling Strategy
Adaptive Rotation
QQQ
SPY
Cumulative Return
5.98
4.80
4.02
2.80
Annualized Return (%)
25.85
22.32
19.56
14.14
Annualized Volatility (%)
27.85
20.30
24.20
19.61
Sharpe Ratio
0.93
1.10
0.81
0.72
Maximum Drawdown (%)
-38.95
-21.46
-35.12
-33.72
Calmar Ratio
0.66
1.04
0.56
0.42
Table 3: Performance comparison of representative use cases and benchmark indices (2018–2025).
Figure 3: Backtest performance comparison across representative strategy configurations under the unified weight-centric protocol (January 7, 2018 – October 24, 2025). Results illustrate cumulative portfolio trajectories relative to benchmark references.
Figure 4: Paper trading performance relative to benchmark indices (October 26, 2025 – March 12, 2026), demonstrating deployment-consistent execution under daily rebalancing.
Metric
Strategy
SPY
QQQ
Cumulative Return
1.20
0.97
0.95
Total Return (%)
19.76
-2.51
-4.79
Annualized Return (%)
62.16
-6.60
-12.32
Annualized Volatility (%)
31.75
11.96
16.79
Sharpe Ratio
1.96
-0.55
-0.73
Maximum Drawdown (%)
-12.22
-5.35
-7.88
Table 4: Performance comparison between paper trading and benchmark indices (Oct 26, 2025–Mar 12, 2026, Daily Turnover).
Figure 5: Portfolio allocation trajectory under the unified weight-based execution framework during paper trading. The figure illustrates time-varying exposure adjustments across asset groups, demonstrating modular allocation outputs that are directly executable without architectural changes.