8.4 KiB
8.4 KiB
celkovy optimalizacni backtest na vetsim oknu 1 - 300 a možná take to udělat jako parametr zkusit CV zobrazit nejak robustnost parametru
In [ ]:
from v2realbot.tools.loadbatch import load_batch from v2realbot.utils.utils import zoneNY import pandas as pd import numpy as np import vectorbtpro as vbt from itables import init_notebook_mode, show import datetime from itertools import product from v2realbot.config import ACCOUNT1_PAPER_API_KEY, ACCOUNT1_PAPER_SECRET_KEY, DATA_DIR init_notebook_mode(all_interactive=True) vbt.settings.set_theme("dark") vbt.settings['plotting']['layout']['width'] = 1280 vbt.settings.plotting.auto_rangebreaks = True vbt.settings.returns.year_freq = "252 days" # Set the option to display with pagination pd.set_option('display.notebook_repr_html', True) pd.set_option('display.max_rows', 10) # Number of rows per page # Define the market open and close times market_open = datetime.time(9, 30) market_close = datetime.time(16, 0) entry_window_opens = 1 entry_window_closes = 370 forced_exit_start = 380 forced_exit_end = 390 #LOAD FROM BATCH # res, df = load_batch(batch_id="f1ac6651", #138170bc 0fb5043a bde6d0be f1ac6651 # space_resolution_evenly=False, # indicators_columns=["Rsi14"], # main_session_only=True, # verbose = False) # if res < 0: # print("Error" + str(res) + str(df)) # df = df["bars"] # basic_data = vbt.Data.from_data(vbt.symbol_dict({"BAC": df}), tz_convert=zoneNY) # #m1_data = basic_data[['Open', 'High', 'Low', 'Close', 'Volume']] # basic_data = basic_data.transform(lambda df: df.between_time('09:30', '16:00')) # #basic_data.info() #LOAD FROM PARQUET #list all files is dir directory with parquet extension dir = DATA_DIR + "/notebooks/" import os files = [f for f in os.listdir(dir) if f.endswith(".parquet")] print('\n'.join(map(str, files))) file_name = "ohlcv_df-SPY-2024-01-01T09:30:00-2024-05-14T16:00:00.parquet" ohlcv_df = pd.read_parquet(dir+file_name,engine='pyarrow') basic_data = vbt.Data.from_data(vbt.symbol_dict({"SPY": ohlcv_df}), tz_convert=zoneNY)
In [ ]:
# basic_data.stats() basic_data.data["SPY"].info()
In [ ]:
vbt.open_api_ref(vbt.base) vbt
In [ ]:
##na toto udelat crosssvalidationu nebo alespon na testovacim ci jinem obdobi #take udelat long leg - tato je shortovaci #8 -0.06 -0.2 0.0028 0.0048 4.156254 #short combination ok for train(4)/test(0.1) (window 1-90, fe 95-100) #2, -0.02, -0.25, 0.0018, 0.0068 #dalsi ok hodnota shortu for train/test 4/1 #70, 8, -0.06, -0.2, 0.0013, 0.0053 #kombinace bez roc_th, train/test 7/-1.5 #70 7 -0.07 0.0033 0.0063 #opet bez roc_th, train(5.77)/test 0.9 - spolus tsl_stop + tsl_th #29 7 -0.09 0.0033 0.0068 #bez roc_th a s trailing sl train/test 8.1/-0.8 #70 2 -0.05 0.0018 0.0068 #SPY - short # entry_window_closes mom_timeperiod mom_th sl_stop tp_stop # 10 6 -0.13 0.0022 0.0057 #5 7 -0.19 0.0037 0.0042 #SPY - long #45 4 0.27 0.0047 0.0067 # TODO: #- vyzkouset zda nejvyhodnejsi kombinace krom train/testu funguje i na nasledujicich dnech po trainu # -zkusit najit v short datasetu neco vyhodneho co funguji i na testu # - dodelat kombinace pro long signaly # - zkusit walk forward # - vytvorit vysledkove totoznou na v2realbot # - podivat se jak detailne funguji tsl_stop a tsl_th #70, 4, -0.07, 0.0048, 0.0068 entry_window_closes, mom_timeperiod, mom_th, sl_stop, tp_stop = 8, 3, 0.07, 0.0028, 0.0033 roc_th = 0 momshort = vbt.indicator("talib:MOM").run(basic_data.close, timeperiod=mom_timeperiod, short_name = "slope_short") rocp = vbt.indicator("talib:ROC").run(basic_data.close, short_name = "rocp") #rate of change + momentum #momshort.plot rocp.real_crossed_below(roc_th) & short_signal = momshort.real_crossed_below(mom_th) long_signal = momshort.real_crossed_above(mom_th) # print("short signal") # print(short_signal.value_counts()) #forced_exit = pd.Series(False, index=close.index) forced_exit = basic_data.symbol_wrapper.fill(False) #entry_window_open = pd.Series(False, index=close.index) entry_window_open= basic_data.symbol_wrapper.fill(False) # Calculate the time difference in minutes from market open for each timestamp elapsed_min_from_open = (forced_exit.index.hour - market_open.hour) * 60 + (forced_exit.index.minute - market_open.minute) entry_window_open[(elapsed_min_from_open >= entry_window_opens) & (elapsed_min_from_open < entry_window_closes)] = True #print(entry_window_open.value_counts()) forced_exit[(elapsed_min_from_open >= forced_exit_start) & (elapsed_min_from_open < forced_exit_end)] = True short_entries = (short_signal & entry_window_open) short_exits = forced_exit entries = (long_signal & entry_window_open) exits = forced_exit #long_entries.info() #number of trues and falses in long_entries # print(short_exits.value_counts()) # print(short_entries.value_counts()) #fig = plot_2y_close([],[momshort, rocp], close) #short_signal.vbt.signals.plot_as_entries(close, fig=fig, add_trace_kwargs=dict(secondary_y=False)) #print(sl_stop) #short_entries=short_entries, short_exits=short_exits, pf = vbt.Portfolio.from_signals(close=basic_data, entries=entries, exits=exits, tsl_stop=sl_stop, tp_stop = tp_stop, fees=0.0167/100, freq="1s") #sl_stop=sl_stop, tp_stop = sl_stop,
In [ ]:
pf_B = pf.resample("30T") pf_B.stats() pf_B.orders.records_readable
In [ ]:
pf.plot()
In [ ]:
pf.get_drawdowns().records_readable
In [ ]:
pf.orders.records_readable
In [ ]:
pf.value.plot().show()
In [ ]: