I've just pushed a major upgrade to psystemtrade. Read: this. Now you can: get data from quandl.com, store in mongodb.com and github.com/manahl/arctic, and do your own back-adjustment of futures prices.
Oh no, not another python backtester...
- Thread starter globalarbtrader
- Start date
Hi GAT,
I'm wondering about 2 issues in bootstrapping forecast weights using chapter15.futuresestimateconfig.yaml
1. carry weight seems to be below 0.1, not sure if this is too low
2. when pooling instruments, weights for the same trading rule for each instrument is different
forecast_weight_estimate:
method: bootstrap
pool_instruments: True
CORN
ewmac64_256 0.195
ewmac16_64 0.521
carry 0.087
ewmac32_128 0.198
EDOLLAR
ewmac64_256 0.329
carry 0.132
ewmac32_128 0.539
EUROSTX
ewmac64_256 0.274
ewmac16_64 0.450
carry 0.116
ewmac32_128 0.160
MXP
ewmac64_256 0.227
ewmac16_64 0.536
carry 0.090
ewmac32_128 0.147
US10
ewmac64_256 0.206
ewmac16_64 0.520
carry 0.121
ewmac32_128 0.152
V2X
ewmac64_256 0.360
carry 0.119
ewmac32_128 0.520
I'm wondering about 2 issues in bootstrapping forecast weights using chapter15.futuresestimateconfig.yaml
1. carry weight seems to be below 0.1, not sure if this is too low
2. when pooling instruments, weights for the same trading rule for each instrument is different
forecast_weight_estimate:
method: bootstrap
pool_instruments: True
CORN
ewmac64_256 0.195
ewmac16_64 0.521
carry 0.087
ewmac32_128 0.198
EDOLLAR
ewmac64_256 0.329
carry 0.132
ewmac32_128 0.539
EUROSTX
ewmac64_256 0.274
ewmac16_64 0.450
carry 0.116
ewmac32_128 0.160
MXP
ewmac64_256 0.227
ewmac16_64 0.536
carry 0.090
ewmac32_128 0.147
US10
ewmac64_256 0.206
ewmac16_64 0.520
carry 0.121
ewmac32_128 0.152
V2X
ewmac64_256 0.360
carry 0.119
ewmac32_128 0.520
Code:
from systems.provided.futures_chapter15.estimatedsystem import futures_system
from sysdata.configdata import Config
system = futures_system()
print(system.accounts.portfolio().sharpe())
for instr_code in system.get_instrument_list():
rule2weight = system.combForecast.get_forecast_weights(instr_code).iloc[-1].to_dict()
print(instr_code)
for rule in rule2weight:
print(rule + " " + '{0:.3f}'.format(rule2weight[rule]))Hi is there anyone using Pysystemtrade actively?
Will love to create a live chat like discord etc or email so that we can talk about issues and stuff. I'm trying to get it to backtest other asset classes like stock, but I'm not sure if the sharpe ratios look correct and what are the changes necessary. Would love to discuss...
Will love to create a live chat like discord etc or email so that we can talk about issues and stuff. I'm trying to get it to backtest other asset classes like stock, but I'm not sure if the sharpe ratios look correct and what are the changes necessary. Would love to discuss...
Question for the general audience: I see that Quandl have stopped providing FX prices up to the current day. What are people doing for FX prices, what sources are people going to, and is anyone importing them into this upgraded data component?
I pull all prices, including FX, from IB. But I use my own software, not the tools developed by @globalarbtrader .
I was using the codes configured out of the box in pysystemtrade e.g. "CURRFX/EURUSD". The series seem to stop at 2/28/2018 and there now seems to be a charge for them which is high (higher than GAT would approve of paying). You are right there do seem to be other series available also on quandl e.g. there is a series available from FRED which has all of the currencies except Korean Won e.g. "FRED/DEXUSEU" for Euro. Some of these are the wrong way around, which is not the end of the world. However I see that in spite of being advertised as daily, this one seems a few days out of date e.g. on today 6/27/2018 the latest available price is from 6/22/2018. So, usable for backtesting and research, but not for daily trading.
Hi,
I'm at the point right now where my own system that's doing real-time execution and IB integration is ready for paper-trading (C#+SQL, event-driven), only need to fit the parameters, but something doesn't seem to be working in pysystemtrade..
So just to confirm, should the current version of pysystemtrade work correctly or there are some temporary issues with it at the moment ? (specifically the optimisation\fitting part)?
Because no matter what I do, I'm always getting estimated weight on the carry rule only around 0.1 (same as pointed by traider) on all instruments (out of 4 EWMAC + 1 carry rules), which in theory shouldn't be like that (carry should receive around 0.5 weight).
Also, when I'm printing positions with get_actual_position('CORN') they seem too large for 250k capital (though, I might be misunderstanding something here..).
So if it's expected because of current "work in progress", what earlier version was relatively stable for weights optimisation ? (or maybe I'm just doing something wrong and everything should actually be working, that's always possible of course
)
here's the code I'm using:
from systems.provided.futures_chapter15.estimatedsystem import futures_system
system=futures_system()
system.set_logging_level("on")
system.config.forecast_weight_estimate["pool_instruments"]=True
system.config.forecast_weight_estimate["method"]="bootstrap"
system.config.forecast_weight_estimate["equalise_means"]=False
system.config.forecast_weight_estimate["monte_runs"]=200
system.config.forecast_weight_estimate["bootstrap_length"]=104
system.accounts.portfolio().curve().plot()
#print forecast weights:
for instr in system.get_instrument_list():
print(instr)
fcW = system.combForecast.get_forecast_weights(instr)
print(fcW.tail(1))
fcW.plot(title=instr)
#print positions:
system.portfolio.get_actual_position('CORN').plot(title='Corn Positions')
I'm at the point right now where my own system that's doing real-time execution and IB integration is ready for paper-trading (C#+SQL, event-driven), only need to fit the parameters, but something doesn't seem to be working in pysystemtrade..
So just to confirm, should the current version of pysystemtrade work correctly or there are some temporary issues with it at the moment ? (specifically the optimisation\fitting part)?
Because no matter what I do, I'm always getting estimated weight on the carry rule only around 0.1 (same as pointed by traider) on all instruments (out of 4 EWMAC + 1 carry rules), which in theory shouldn't be like that (carry should receive around 0.5 weight).
Also, when I'm printing positions with get_actual_position('CORN') they seem too large for 250k capital (though, I might be misunderstanding something here..).
So if it's expected because of current "work in progress", what earlier version was relatively stable for weights optimisation ? (or maybe I'm just doing something wrong and everything should actually be working, that's always possible of course
)here's the code I'm using:
from systems.provided.futures_chapter15.estimatedsystem import futures_system
system=futures_system()
system.set_logging_level("on")
system.config.forecast_weight_estimate["pool_instruments"]=True
system.config.forecast_weight_estimate["method"]="bootstrap"
system.config.forecast_weight_estimate["equalise_means"]=False
system.config.forecast_weight_estimate["monte_runs"]=200
system.config.forecast_weight_estimate["bootstrap_length"]=104
system.accounts.portfolio().curve().plot()
#print forecast weights:
for instr in system.get_instrument_list():
print(instr)
fcW = system.combForecast.get_forecast_weights(instr)
print(fcW.tail(1))
fcW.plot(title=instr)
#print positions:
system.portfolio.get_actual_position('CORN').plot(title='Corn Positions')