def cal_seri_indicator(seri):
df = pd.DataFrame()
df['value'] = seri
median = seri.median()
df['median'] = median
leastsq = math_util.const_leastsq([median], seri.index, seri.values)
df['leastsq'] = leastsq
return df
def run(self):
self.target_sec.load_data()
self.screen_print('---{0} {1}---'.format(self.target_sec.code,
self.target_sec.name))
self.screen_print('[info]')
yesterday_price = self.target_sec.close_price_seri[-1]
self.screen_print('yesterday price:{0}'.format(
self.target_sec.close_price_seri[-1]))
result = pd.DataFrame(columns=['buy_profit', 'shape'])
# cal three line
df = pd.DataFrame()
df['value'] = self.target_sec.close_price_seri
for period in self.period_list:
seri = self.target_sec.close_price_seri[-period:]
median = seri.median()
leastsq = math_util.const_leastsq([median], seri.index,
seri.values)
df['{0}'.format(period)] = leastsq
buy_profit = (leastsq - yesterday_price) / yesterday_price * 100
self.screen_print('{0} price:{1}, buy_profit:{2:.2f}%'.format(
period, leastsq, buy_profit))
result.loc[period, 'buy_profit'] = buy_profit
if self.show:
plt.figure()
df.plot()
plt.show()
# cal shape
self.screen_print('[shape]')
for period in self.period_list:
seri = self.target_sec.close_price_seri[-period:]
shape_ins = shape.Shape(0.05)
(shape_id, shape_name) = shape_ins.what_shape(seri)
self.screen_print("{0}:{1}".format(period, shape_name))
result.loc[period, 'shape'] = shape_name
# analysis
self.screen_print('[analysis]')
def adjust_profit(row):
if row['shape'] in ('up_line', 'down_line'):
row['buy_profit'] = 0
return row
result = result.apply(adjust_profit, axis=1)
max_id = result['buy_profit'].idxmax()
min_id = result['buy_profit'].idxmin()
latest_id = min(result.index)
max_row = result.loc[max_id]
min_row = result.loc[min_id]
latest_row = result.loc[latest_id]
self.screen_print('latest: shape({0}, day({1}))'.format(
latest_row['shape'], latest_id))
self.screen_print(
'max profit: profit({0:.2f}%) shape({1}) day({2})'.format(
max_row['buy_profit'], max_row['shape'], max_id))
self.screen_print(
'min profit: profit({0:.2f}%) shape({1}) day({2})'.format(
min_row['buy_profit'], min_row['shape'], min_id))
return (latest_row, max_row, min_row)
def run(self):
self.target_sec.load_data()
seri = self.target_sec.close_price_seri
# basic info
self.screen_print('---{0} {1}---'.format(self.target_sec.code,
self.target_sec.name))
self.screen_print('[info]')
yesterday_price = self.target_sec.close_price_seri[-1]
self.screen_print('yesterday price:{0}'.format(
self.target_sec.close_price_seri[-1]))
# dynamic cal tp
df = pd.DataFrame()
df['price'] = self.target_sec.close_price_seri
result = pd.DataFrame(columns=['profit'])
tp_df = pd.DataFrame()
tp_seri = pd.Series()
regime_tp_list = []
# temp var
last_tp = None
candidate_tp = None
regime_tp_num = 0
max_regime_tp = None
min_regime_tp = None
tp_win_list = []
for i, (rq, price) in enumerate(seri.iteritems()):
point = {'idx': i, 'rq': rq, 'price': price, 'pct': 0}
if last_tp is None:
last_tp = point
regime_tp_list.append(point)
continue
percent_to_turn = (point['price'] -
last_tp['price']) / last_tp['price']
point['pct'] = percent_to_turn
if candidate_tp is None:
# find the first candidate
if percent_to_turn >= self.change_percent or percent_to_turn <= -self.change_percent:
candidate_tp = point
continue
is_up_trend = (candidate_tp['pct'] >= 0
and percent_to_turn > candidate_tp['pct'])
is_down_trend = (candidate_tp['pct'] <= 0
and percent_to_turn < candidate_tp['pct'])
if is_up_trend or is_down_trend:
candidate_tp = point
else:
percent_to_candidate = (point['price'] - candidate_tp['price']
) / candidate_tp['price']
if percent_to_candidate < self.change_percent and percent_to_candidate > -self.change_percent:
continue
# regime shift
if max_regime_tp is not None:
percent_to_max = abs(
(point['price'] - max_regime_tp['price']) /
max_regime_tp['price'])
else:
percent_to_max = 0
if min_regime_tp is not None:
percent_to_min = abs(
(point['price'] - min_regime_tp['price']) /
min_regime_tp['price'])
else:
percent_to_min = 0
if percent_to_max > self.regime_percent or percent_to_min > self.regime_percent:
# cal profit
head_tp = regime_tp_list[-1]
tail_tp = last_tp
if head_tp['idx'] != tail_tp['idx']:
win_seri = seri[head_tp['idx']:tail_tp['idx']]
# if window seri change between regime_percent
if (regime_tp_num >= self.regime_tp_num):
median = win_seri.median()
leastsq = math_util.const_leastsq([median],
win_seri.index,
win_seri.values)
days = (tail_tp['rq'] - head_tp['rq']).days
period = '{0}<-{2}->{1}'.format(
head_tp['rq'], tail_tp['rq'], days)
df[period] = leastsq
df[period + '_'] = win_seri
tp_win_list.append(win_seri)
profit = (leastsq -
yesterday_price) / yesterday_price * 100
self.screen_print(
'{0}: price({1}), profit({2:.2f}%)'.format(
period, leastsq, profit))
result.loc[period, 'profit'] = profit
# print('rtl append {0}'.format(point))
regime_tp_list.append(point)
# clear regime tp
# print('--------- regime end-------')
regime_tp_num = 0
max_regime_tp = None
min_regime_tp = None
# new tp
# filter non regime tp
if len(regime_tp_list
) == 0 or regime_tp_list[-1]['idx'] <= last_tp['idx']:
# print('regime tp {0} {1} {2}'.format(last_tp['rq'], last_tp['price'], regime_tp_num))
regime_tp_num += 1
if max_regime_tp is None or last_tp[
'price'] > max_regime_tp['price']:
max_regime_tp = last_tp
if min_regime_tp is None or last_tp[
'price'] < min_regime_tp['price']:
min_regime_tp = last_tp
tp_seri.loc[last_tp['rq']] = last_tp['price']
last_tp = candidate_tp
point['pct'] = percent_to_candidate
candidate_tp = point
# process final tp
tp_seri.loc[last_tp['rq']] = last_tp['price']
if candidate_tp is not None:
tp_seri.loc[candidate_tp['rq']] = candidate_tp['price']
# draw result
if self.show:
df.plot()
# draw tp
tp_df['price'] = tp_seri
for x, y in enumerate(tp_win_list):
tp_df[str(x)] = y
if self.show:
tp_df.plot()
# analysis
self.screen_print('[analysis]')
max_id = result['profit'].idxmax()
min_id = result['profit'].idxmin()
max_row = result.loc[max_id]
min_row = result.loc[min_id]
self.screen_print('max profit: profit({0:.2f}%) day({1})'.format(
max_row['profit'], max_id))
self.screen_print('min profit: profit({0:.2f}%) day({1})'.format(
min_row['profit'], min_id))
# draw at the end
if self.show:
plt.show()
return (max_row, min_row)