strategy-genome
A typed genetic algorithm (GA) and genetic programming (GP) optimizer for Python, with an Optuna sampler so any Optuna study can use it.
- You bring the fitness function.
evolve(fitness, space)callsfitness(genome)and expects a float, or a tuple of floats for multi-objective runs. Nothing about trading is built in. - Mixed search spaces: floats (optionally on a step grid), integers, log-scale floats, categoricals and booleans.
- Real GA machinery: elitism, tournament selection, SBX / arithmetic / uniform / one- and two-point crossover, gaussian / polynomial / creep / boundary mutation, adaptive mutation rate and adaptive operator choice, NSGA-II for multiple objectives, an island model, stagnation restarts, and a hall of fame.
- Reproducible: the same seed and the same fitness give the same run.
- Light: the GA core is pure standard-library Python. numpy is used by the GP submodule and the trading examples. Optuna is an optional extra.
Carved out of srfm-lab
(ml/genetic), MIT licensed. See provenance for what was
changed and which upstream bugs were fixed on the way out.
Install
pip install strategy-genome # GA core + GP (numpy)
pip install "strategy-genome[optuna]" # adds the Optuna sampler
Python 3.10 or newer.
Quickstart
from strategy_genome import evolve, space as sp
def fitness(g):
# g["name"] reads one parameter; g.params is the whole dict
return (g["x"] - 3.0) ** 2 + (g["y"] + 1.0) ** 2
result = evolve(
fitness,
{"x": sp.Float(-10, 10), "y": sp.Float(-10, 10)},
population=40,
generations=80,
seed=7,
direction="minimize",
)
result.best_params # {'x': 3.00..., 'y': -0.99...}
result.best_value # what your function returned for the best genome
result.hall_of_fame # best unique genomes seen, best first
result.diversity_history # mean pairwise parameter distance per generation
Search spaces
space = {
"fast": sp.Int(2, 50),
"slow": sp.Int(10, 200, step=5),
"stop": sp.Float(0.005, 0.10),
"lr": sp.LogFloat(1e-5, 1e-1),
"kind": sp.Categorical(["ema", "sma", "wma"]),
"short": sp.Bool(),
}
Shorthand also works: (0, 10) is an integer range, (0.0, 1.0) a float
range, and a list is a categorical.
Several objectives
Return a tuple and give one direction per objective. The run switches to
NSGA-II and result.pareto_front holds the non-dominated genomes.
def fitness(g):
returns = backtest(g.params)
return sharpe(returns), max_drawdown(returns)
result = evolve(fitness, space, population=60, generations=100, seed=1,
directions=["maximize", "minimize"])
for g in result.pareto_front:
print(g.values, g.params)
Failures
If fitness raises or returns NaN, that genome scores -inf and the run
continues. result.n_failed_evaluations counts them. Pass
raise_errors=True to stop on the first error instead.
Use it from Optuna (and freqtrade)
import optuna
from strategy_genome.optuna import GeneticStrategySampler
study = optuna.create_study(sampler=GeneticStrategySampler(population_size=30, seed=1))
study.optimize(objective, n_trials=500)
freqtrade's hyperopt runs on Optuna and lets a strategy return any
optuna.samplers.BaseSampler from HyperOpt.generate_estimator, so the
sampler drops straight in. See Optuna and freqtrade,
including a benchmark against Random, NSGA-II and TPE.
More
- API reference
- Optuna and freqtrade
- Genetic programming (
strategy_genome.gp) - Trading examples (
strategy_genome.trading) - Provenance and upstream fixes
- Ecosystem plan (OptunaHub)
- Examples:
examples/
License
MIT. Copyright (c) 2026 Matthew Busel. See LICENSE and NOTICE.
Metadata
Release files for strategy-genome 0.1.0
For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.
Source distribution (sdist)
| File | Size | Uploaded | Provenance | Actions |
|---|---|---|---|---|
| strategy_genome-0.1.0.tar.gz | 113.5 kB | With provenance | Details for strategy_genome-0.1.0.tar.gz |
Built distribution (wheel)
| File | Interpreter | ABI | Platform | Actions Reset |
|---|---|---|---|---|
| strategy_genome-0.1.0-py3-none-any.whl | Python 3 | none | any | Details for strategy_genome-0.1.0-py3-none-any.whl |
Total release size: 212.8 kB
Release files / strategy_genome-0.1.0.tar.gz
| Download URL | strategy_genome-0.1.0.tar.gz |
|---|---|
| Size | 113.5 kB |
| Tags | Source |
|
SHA-256 checksum How to use checksums |
fdac375e1e3cce00f6b99adf303b86b9be3640b57bdcf432953a0915fdd92fe3
|
|
BLAKE2b-256 checksum How to use checksums |
74865978c620f4b5d83cd053fc55fbb7a559959538b70a63a8ef3735242e36f8
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
Yes |
| Uploaded via |
twine/7.0.0 CPython/3.13.16
|
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
PyPI Publish Attestation
PyPI verified that this artifact, at this checksum, originated from the publisher listed below.
Signed by GitHub Actions, verified by PyPI on Nov 15, 2026.
Transparency logRelease files / strategy_genome-0.1.0-py3-none-any.whl
| Download URL | strategy_genome-0.1.0-py3-none-any.whl |
|---|---|
| Size | 99.3 kB |
| Tags | Python 3 |
|
SHA-256 checksum How to use checksums |
7ed86c8aa176849c07e25dad20c3e8a46d95affaceb416f78411ecb0e58931ab
|
|
BLAKE2b-256 checksum How to use checksums |
a2a87475cda83c1be12f39a56210b710c74833693ca986b7d5e889b0839c85f5
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
Yes |
| Uploaded via |
twine/7.0.0 CPython/3.13.16
|
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
PyPI Publish Attestation
PyPI verified that this artifact, at this checksum, originated from the publisher listed below.
Signed by GitHub Actions, verified by PyPI on Nov 15, 2026.
Transparency logSecurity for strategy-genome 0.1.0
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
Full provenance
Both files in this release have been uploaded with provenance metadata.
Attestation for strategy_genome-0.1.0.tar.gz
PyPI Publish Attestation
PyPI verified that this artifact, at this checksum, originated from the publisher listed below.
Signed by GitHub Actions, verified by PyPI on Nov 15, 2026.
Transparency logA note on package security
PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.
To learn more about protecting yourself from malicious actors, visit our package security documentation.