def fitness(self, genes):
'''
Cost function in the optimization process
_________________________________________
Parameters
genes: list(3)
ordered parameters to optimize
_________________________________________
Return
score: float(1)
Error of the cost function ran with this solution
'''
# No evoluation in manager (Constant) configuration so reads it statically from file
# We want here to optimize the algorithm parameters
engine = Simulation()
try:
engine.configure(bt_cfg=self.bt_cfg,
a_cfg={'long_window': genes[0], 'ma_rate': float(genes[1] / 10.0), 'threshold': genes[2]},
m_cfg=None)
results = engine.run_backtest()
risk_metrics = engine.overall_metrics()
except:
pdb.set_trace()
log.error('Exception caught while running cost function')
return 1
return self.evaluate(results, risk_metrics)
results = engine.run_backtest()
'''--------------------------------------------------------------------------------------------- Results ----'''
log.info('Portfolio returns: {}'.format(results.portfolio_value[-1]))
if args['live'] or results.portfolio_value[-1] == 100000:
# Currently tests don't last more than 20min, analysis is not relevant, neither backtest without orders
sys.exit(0)
#TODO Implement in datafeed a generic save method (which could call the correct database save method)
#NOTE Could do a generic save client method (retrieve the correct model, with correct fields)
perf_series = engine.rolling_performances(timestamp='one_month', save=True, db_id=args['database'])
#TODO save returns not ready yet, don't try to save
#TODO more benchmarks choice (zipline modification)
returns_df = engine.get_returns(benchmark='^GSPC', save=False)
risk_metrics = engine.overall_metrics(save=True, db_id=args['database'])
#FIXME irrelevant results if no transactions were made
log.info('\n\nReturns: {}% / {}%\nVolatility:\t{}\nSharpe:\t\t{}\nMax drawdown:\t{}\n\n'.format(
risk_metrics['Returns'] * 100.0,
risk_metrics['Benchmark.Returns'] * 100.0,
risk_metrics['Volatility'],
risk_metrics['Sharpe.Ratio'],
risk_metrics['Max.Drawdown']))
# If we work in local, draw a quick summary plot
if not args['remote']:
data = returns_df.drop(['Returns', 'Benchmark.Returns'], axis=1)
data.plot()
plt.show()
results = engine.run_backtest()
'''--------------------------------------------------------------------------------------------- Results ----'''
log.info('Portfolio returns: {}'.format(results.portfolio_value[-1]))
if args['live'] or results.portfolio_value[-1] == 100000:
# Currently tests don't last more than 20min, analysis is not relevant, neither backtest without orders
sys.exit(0)
#TODO Implement in datafeed a generic save method (which could call the correct database save method)
#NOTE Could do a generic save client method (retrieve the correct model, with correct fields)
perf_series = engine.rolling_performances(timestamp='one_month', save=False, db_id='test')
#TODO save returns not ready yet, don't try to save
#TODO more benchmarks choice (zipline modification)
returns_df = engine.get_returns(benchmark='SP500', save=False)
risk_metrics = engine.overall_metrics(save=True, db_id='test')
log.info('\n\nReturns: {}% / {}%\nVolatility:\t{}\nSharpe:\t\t{}\nMax drawdown:\t{}\n\n'.format(
risk_metrics['Returns'] * 100.0,
risk_metrics['Benchmark.Returns'] * 100.0,
risk_metrics['Volatility'],
risk_metrics['Sharpe.Ratio'],
risk_metrics['Max.Drawdown']))
# If we work in local, draw a quick summary plot
if not args['remote']:
data = returns_df.drop(['Returns', 'Benchmark.Returns'], axis=1)
data.plot()
plt.show()
# R statistical analysis
results = engine.run_backtest()
'''--------------------------------------------------------------------------------------------- Results ----'''
log.info('Portfolio returns: {}'.format(results.portfolio_value[-1]))
if args['live'] or results.portfolio_value[-1] == args['cash']:
# Currently tests don't last more than 20min, analysis is not relevant, neither backtest without orders
sys.exit(0)
#TODO Implement in datafeed a generic save method (which could call the correct database save method)
#NOTE Could do a generic save client method (retrieve the correct model, with correct fields)
perf_series = engine.rolling_performances(timestamp='one_month', save=True, db_id=args['database'])
#TODO save returns not ready yet, don't try to save
#TODO more benchmarks choice (zipline modification)
returns_df = engine.get_returns(benchmark='cac', save=False)
risk_metrics = engine.overall_metrics(metrics=perf_series, save=True, db_id=args['database'])
#FIXME irrelevant results if no transactions were made
log.info('\n\nReturns: {}% / {}%\nVolatility:\t{}\nSharpe:\t\t{}\nMax drawdown:\t{}\n\n'.format(
risk_metrics['Returns'] * 100.0,
risk_metrics['Benchmark.Returns'] * 100.0,
risk_metrics['Volatility'],
risk_metrics['Sharpe.Ratio'],
risk_metrics['Max.Drawdown']))
# If we work in local, draw a quick summary plot
if not args['remote']:
data = returns_df.drop(['Returns', 'Benchmark.Returns'], axis=1)
data.plot()
plt.show()
import ipdb; ipdb.set_trace()
