Skip to content

python-runtime-comparison.md

Latest commit

 

History

History
134 lines (95 loc) · 5.55 KB

python-runtime-comparison.md

File metadata and controls

134 lines (95 loc) · 5.55 KB
Raw

Python 3.13 vs 3.14 Runtime Comparison Report

I checked whether the TestX7 speed result was mostly coming from Python 3.14, since the maintainer is still on Python 3.13.

In this local live-style analyze benchmark, Python 3.14 was not the reason for the large speedup. TestX7 also passed the same 80-pair speed gate on Python 3.13.

Quick Takeaway

  • TestX7 was tested on Python 3.13.13 and Python 3.14.4.
  • Both runtimes used Freqtrade 2026.4 and the same major dependency versions.
  • TestX7 passed 40/40 80-pair live-style analyze loops under 5 seconds on Python 3.13.
  • Original X7 stayed around 206-209 seconds for one 80-pair live-style analyze loop on both runtimes.
  • The large TestX7 speedup looks strategy-side, not Python-version-side.
  • This is not a backtest-speed result and not a live production recommendation.

Test Setup

The current default Freqtrade Docker image in this repo uses:

freqtradeorg/freqtrade:stable
Freqtrade 2026.4
Python 3.14.3

For a cleaner runtime comparison, two local PyPI-based images were built from the same Dockerfile. This avoids comparing the official stable image against a custom Python 3.13 image with slightly different dependency patch versions.

Runtime Image Freqtrade Important packages
Python 3.13 nfi-x7-freqtrade:py313-2026.4 2026.4 pandas 3.0.2, numpy 2.4.4, pyarrow 24.0.0, TA-Lib 0.6.8, ft-pandas-ta 0.3.16, technical 1.6.0, ccxt 4.5.52
Python 3.14 nfi-x7-freqtrade:py314-pypi-2026.4 2026.4 pandas 3.0.2, numpy 2.4.4, pyarrow 24.0.0, TA-Lib 0.6.8, ft-pandas-ta 0.3.16, technical 1.6.0, ccxt 4.5.52

Common benchmark conditions:

Host CPUs: 10
Docker CPUs: 10
Workers: 9
Pairs: 80
Rows per dataframe: 1200
Runmode: dry_run
Data source: local Binance candle data mounted read-only from freqtrade-nfi

TestX7 Live-Style Analyze Result

Runtime Loops Average Min Max Over 5s Gate
Python 3.13.13 40 2.946143s 2.782652s 3.378742s 0 PASS
Python 3.14.4 40 2.969508s 2.761902s 3.489651s 0 PASS

What I read from this:

  • TestX7 works on Python 3.13 and still stays under the 5-second 80-pair gate.
  • In this run, Python 3.14 was not faster than Python 3.13 in a meaningful way.
  • The difference is small enough that it should be treated as benchmark noise unless repeated on more machines.

Original X7 One-Loop Reference

Original X7 is much slower, so only one live-style analyze loop was measured per runtime.

Runtime Strategy Loops Time
Python 3.13.13 NostalgiaForInfinityX7 1 206.224397s
Python 3.14.4 NostalgiaForInfinityX7 1 208.610913s

What I read from this:

  • The original strategy remained around 206-209 seconds for one 80-pair live-style loop.
  • Python runtime version does not explain the gap between original X7 and TestX7.
  • The large speedup is from the TestX7 strategy-side changes, not from moving from Python 3.13 to 3.14.

What This Supports

This benchmark supports one narrow claim:

TestX7 still shows the live-style analyze speedup on Python 3.13.

It does not support a claim that Python 3.14 is faster for this strategy workload. In this measurement, Python 3.13 and Python 3.14 were effectively the same, with Python 3.13 slightly ahead by a small margin.

What This Does Not Claim

  • It does not prove Python 3.13 is generally faster than Python 3.14.
  • It does not prove the same timing on every VPS or Docker host.
  • It does not prove backtesting is faster. Backtest speed still needs its own benchmark.
  • It does not prove production live safety. Process lifecycle, fallback, worker limits, and long-running dry-run/live tests still matter.

Evidence Files

File What it contains
user_data/backtest_results/test_x7_python313_80pair_gate/testx7-live-loop-80pairs-40loops-workers9-20260512-001929.txt TestX7, Python 3.13, 80-pair, 40-loop gate
user_data/backtest_results/test_x7_python314_pypi_80pair_gate/testx7-live-loop-80pairs-40loops-workers9-20260512-003515.txt TestX7, Python 3.14 PyPI image, 80-pair, 40-loop gate
user_data/backtest_results/runtime_compare_20260512/original-x7-python313-80pairs-1loop.txt Original X7, Python 3.13, 80-pair, 1 loop
user_data/backtest_results/runtime_compare_20260512/original-x7-python314-pypi-80pairs-1loop.txt Original X7, Python 3.14 PyPI image, 80-pair, 1 loop

Reproduce

Build the Python 3.13 runtime:

docker compose -f docker-compose.yml -f docker-compose.python313.yml build freqtrade

Run the Python 3.13 TestX7 gate:

TEST_X7_GATE_DATA_DIR_HOST=/path/to/user_data/data/binance \
  tools/test_x7/run_80pair_5s_gate_python313.sh

Build the matching PyPI-based Python 3.14 runtime:

docker compose -f docker-compose.yml -f docker-compose.python314-pypi.yml build freqtrade

Run the matching Python 3.14 TestX7 gate:

TEST_X7_COMPOSE_FILES=docker-compose.yml:docker-compose.python314-pypi.yml \
TEST_X7_GATE_DATA_DIR_HOST=/path/to/user_data/data/binance \
TEST_X7_GATE_OUTPUT_DIR=user_data/backtest_results/test_x7_python314_pypi_80pair_gate \
  tools/test_x7/run_80pair_5s_gate.sh

Caveats

  • This is a local WSL/Docker benchmark, not a live deployment test.
  • Original X7 was measured with one loop only because each loop takes several minutes.
  • The result should be repeated before making a general claim about Python 3.13 vs 3.14.
  • Backtest speed is a separate question and still needs a dedicated benchmark.