def calc(self,sdata,dates,tbegin,evthreshold,**kwargs):
buy_func = fcustom(self.buy_func,**kwargs)
sell_func = fcustom(self.sell_func,**kwargs)
m = Mediator(buy_func,sell_func)
name = m.name()
trades = m.calc_matched(sdata,dates,begin=tbegin)
ev = self.evaluate_func(trades,self.datemap,**kwargs)
if(not evthreshold(ev)):
ev.matchedtrades,ev.balances = [],[] #相当于先删除。为保证ev的一致性而都赋为[]。否则str(ev)中的zip(...)会出错
#self.ev_result[name] = ev #移入到jduge中，因为可能值不是ev
#print 'null list:',get_null_obj_number(list),',null tuple:',get_null_obj_number(tuple)
return name,ev
def calc(self, sdata, dates, tbegin, evthreshold, **kwargs):
buy_func = fcustom(self.buy_func, **kwargs)
sell_func = fcustom(self.sell_func, **kwargs)
m = Mediator(buy_func, sell_func)
name = m.name()
trades = m.calc_matched(sdata, dates, begin=tbegin)
ev = self.evaluate_func(trades, self.datemap, **kwargs)
if (not evthreshold(ev)):
ev.matchedtrades, ev.balances = [], [
] #相当于先删除。为保证ev的一致性而都赋为[]。否则str(ev)中的zip(...)会出错
#self.ev_result[name] = ev #移入到jduge中，因为可能值不是ev
#print 'null list:',get_null_obj_number(list),',null tuple:',get_null_obj_number(tuple)
return name, ev
def nmediator_factory(trade_signal_maker=make_trade_signal_advanced,
trade_strategy=B1S0,
pricer=cl_pricer):
return fcustom(NMediator,
trade_signal_maker=trade_signal_maker,
trade_strategy=trade_strategy,
pricer=pricer)
def __init__(self,buy_signal_maker,sell_signal_maker
,taxrate=125
,trade_signal_maker=make_trade_signal
,matcher = match_trades
,trade_strategy=B1S1
,pricer = default_pricer,extra_func=atr_extra):
self.buy_signal_maker = buy_signal_maker
self.sell_signal_maker = sell_signal_maker
self.trade_signal_maker = trade_signal_maker
self.matcher = matcher
self.trade_strategy = trade_strategy
self.buy_pricer = pricer[0]
self.sell_pricer = pricer[1]
self.extra_func = extra_func
self.taxrate = taxrate
self.make_trades = fcustom(make_trades,taxrate=taxrate,extra_func=extra_func)
self.last_trade = fcustom(last_trade,taxrate=taxrate,extra_func=extra_func)
def atr_seller_factory(stop_times=3 * BASE / 2,
trace_times=2 * BASE,
covered=10,
up_sector=HIGH):
return fcustom(atr_seller,
stop_times=stop_times,
trace_times=trace_times,
covered=covered,
up_sector=up_sector)
def limit_adjust_deprecated(
source_signal, limit_signal, trans_signal, covered=10
): # covered不能大于127否则会溢出, np.sign(bool array)返回的是int8数组
""" 根据停板信号limit_signal和交易日信号trans_signal调整原始信号，使原始信号避开停板到开板日
可能因covered的原因导致连续非停板日中间出现停板日后,信号多发. 但这个可由makke_trade之类的函数处理掉
只有covered=2时,不会出现这个情况
covered=10，最多10个停板，超过则可能导致跌停情况下卖出信号被忽略
"""
adjuster = fcustom(_limit_adjuster_deprecated, covered=covered)
return smooth(trans_signal, source_signal, limit_signal, sfunc=adjuster)
def limit_adjust_deprecated(
source_signal,
limit_signal,
trans_signal,
covered=10): #covered不能大于127否则会溢出, np.sign(bool array)返回的是int8数组
''' 根据停板信号limit_signal和交易日信号trans_signal调整原始信号，使原始信号避开停板到开板日
可能因covered的原因导致连续非停板日中间出现停板日后,信号多发. 但这个可由makke_trade之类的函数处理掉
只有covered=2时,不会出现这个情况
covered=10，最多10个停板，超过则可能导致跌停情况下卖出信号被忽略
'''
adjuster = fcustom(_limit_adjuster_deprecated, covered=covered)
return smooth(trans_signal, source_signal, limit_signal, sfunc=adjuster)
def __init__(self,
buy_signal_maker,
sell_signal_maker,
taxrate=125,
trade_signal_maker=make_trade_signal,
matcher=match_trades,
trade_strategy=B1S1,
pricer=default_pricer,
extra_func=atr_extra):
self.buy_signal_maker = buy_signal_maker
self.sell_signal_maker = sell_signal_maker
self.trade_signal_maker = trade_signal_maker
self.matcher = matcher
self.trade_strategy = trade_strategy
self.buy_pricer = pricer[0]
self.sell_pricer = pricer[1]
self.extra_func = extra_func
self.taxrate = taxrate
self.make_trades = fcustom(make_trades,
taxrate=taxrate,
extra_func=extra_func)
self.last_trade = fcustom(last_trade,
taxrate=taxrate,
extra_func=extra_func)
def atr_xseller_factory(stop_times=3 * BASE / 2, trace_times=2 * BASE, covered=10, up_sector=HIGH):
""" 用于评估的seller_factory
将倒数第二个信号位置位，使得所有开仓合约平仓
"""
inner_seller = fcustom(
atr_seller, stop_times=stop_times, trace_times=trace_times, covered=covered, up_sector=up_sector
)
def seller(stock, buy_signal, **kwargs):
ss = inner_seller(stock, buy_signal, *kwargs)
twhere = np.where(stock.transaction[VOLUME] > 0)[0]
if len(twhere) and twhere[-1] - 1 >= 0:
ss[twhere[-1] - 1] = 1 # 最后交易日之前的那一天置位，以便最后卖出平仓。
return ss
return seller
def atr_xseller_factory(stop_times=3 * BASE / 2,
trace_times=2 * BASE,
covered=10,
up_sector=HIGH):
''' 用于评估的seller_factory
将倒数第二个信号位置位，使得所有开仓合约平仓
'''
inner_seller = fcustom(atr_seller,
stop_times=stop_times,
trace_times=trace_times,
covered=covered,
up_sector=up_sector)
def seller(stock, buy_signal, **kwargs):
ss = inner_seller(stock, buy_signal, *kwargs)
twhere = np.where(stock.transaction[VOLUME] > 0)[0]
if len(twhere) and twhere[-1] - 1 >= 0:
ss[twhere[-1] - 1] = 1 #最后交易日之前的那一天置位，以便最后卖出平仓。
return ss
return seller
'''
tsx1 = fpeak(source)
sx1 = np.select([tsx1 != 0], [source], 0)
if order <= 1:
return sx1
icursx = np.nonzero(tsx1)[0]
for i in xrange(1, order):
sxx = source[icursx]
tsxx = fpeak(sxx)
icursx = icursx[np.nonzero(tsxx)[0]]
osx = np.zeros_like(source)
osx[icursx] = source[icursx]
return osx
zhpeak = fcustom(zpeak, fpeak=fhigh)
zlpeak = fcustom(zpeak, fpeak=flow)
def zpeaki(source, order=1, fpeak=fhigh):
'''
寻找n阶高/低点
返回值为高点数据序列，以及该高点最大跨度的坐标(即计算该高/低点所需用到的最远的未来数据的坐标)
order默认为1,小于1当作1
返回值中第一个是高/低点非0,其余为0的序列 sh
第二个是该高低点的最远未来数据的坐标序列 si
其中 sh[np.nonzero(sh)]为高点序列, si[np.nonzero(sh)]为坐标序列,sif.time[si[np.nonzero(sh)]]为坐标的影响时间序列
'''
tsx1 = fpeak(source)
sx1 = np.select([tsx1 != 0], [source], 0)
icovered = rollx(np.arange(len(source)), -1)
def atr_seller_factory(stop_times=3 * BASE / 2, trace_times=2 * BASE, covered=10, up_sector=HIGH):
return fcustom(atr_seller, stop_times=stop_times, trace_times=trace_times, covered=covered, up_sector=up_sector)
index,
attr_name_from,
attr_name_to=None):
if attr_name_from in stock.__dict__:
if not attr_name_to:
attr_name_to = attr_name_from
trade.__dict__[attr_name_to] = int(
stock.__dict__[attr_name_from]
[index]) #将atr转换为平凡的int类型，而非numpy.int32，以便后续处理
else:
logger.warn('append attribute error:%s do not have attribute %s',
stock.code, attr_name_from)
return trade
atr_extra = fcustom(append_attribute_extra,
attr_name_from='atr') #atr_extra的一般定义方法
def make_trades(stock,
signal,
tdate,
tpositive,
tnegative,
begin=0,
taxrate=125,
trade_strategy=buy_first,
extra_func=default_extra):
''' stock为stock
ssingal为买卖信号,对于次日买卖的信号，输入前需要将signal roll(1)
tpositive,tnegative为信号值为正和负时的选择价格
taxrate为税率，默认为千分之八
def nmediator_factory(trade_signal_maker=make_trade_signal_advanced,trade_strategy=B1S0,pricer = cl_pricer):
return fcustom(NMediator,trade_signal_maker = trade_signal_maker,trade_strategy = trade_strategy,pricer=pricer)
order默认为1,小于1当作1
'''
tsx1 = fpeak(source)
sx1 = np.select([tsx1!=0],[source],0)
if order <= 1:
return sx1
icursx = np.nonzero(tsx1)[0]
for i in xrange(1,order):
sxx = source[icursx]
tsxx = fpeak(sxx)
icursx = icursx[np.nonzero(tsxx)[0]]
osx = np.zeros_like(source)
osx[icursx] = source[icursx]
return osx
zhpeak = fcustom(zpeak,fpeak=fhigh)
zlpeak = fcustom(zpeak,fpeak=flow)
def zpeaki(source,order=1,fpeak=fhigh):
'''
寻找n阶高/低点
返回值为高点数据序列，以及该高点最大跨度的坐标(即计算该高/低点所需用到的最远的未来数据的坐标)
order默认为1,小于1当作1
返回值中第一个是高/低点非0,其余为0的序列 sh
第二个是该高低点的最远未来数据的坐标序列 si
其中 sh[np.nonzero(sh)]为高点序列, si[np.nonzero(sh)]为坐标序列,sif.time[si[np.nonzero(sh)]]为坐标的影响时间序列
'''
tsx1 = fpeak(source)
sx1 = np.select([tsx1!=0],[source],0)
icovered = rollx(np.arange(len(source)),-1)
if order <= 1:
#收盘价买入，下限突破价卖出，必须有下限突破线
cl_pricer = (lambda s: s.transaction[CLOSE], lambda s: s.down_limit)
#开盘价买入，下限突破价卖出，必须有下限突破线
ol_pricer = (lambda s: s.transaction[OPEN], lambda s: s.down_limit)
#开盘价买入，开盘价卖出
oo_pricer = (lambda s: s.transaction[OPEN], lambda s: s.transaction[OPEN])
#收盘价买入，开盘价卖出
co_pricer = (lambda s: s.transaction[CLOSE], lambda s: s.transaction[OPEN])
#可定制自己的pricer，如
#my_pricer = (lambda s : s.buyprice,lambda s : s.sellprice)
#定制的Mediator
#一次买入一次买出，买入信号次日有效(开盘买入)，卖出信号当日起效
Mediator10 = fcustom(Mediator, trade_strategy=B1S0, pricer=ol_pricer)
#允许连续买入一次卖出，买入信号次日有效(开盘买入)，卖出信号当日起效
CMediator10 = fcustom(Mediator,
trade_signal_maker=make_trade_signal_advanced,
trade_strategy=B1S0,
pricer=ol_pricer)
OMediator10 = fcustom(Mediator,
trade_signal_maker=make_trade_signal_advanced,
trade_strategy=B1S0,
pricer=ol_pricer)
def mediator_factory(trade_signal_maker=make_trade_signal_advanced,
trade_strategy=B1S0,
pricer=cl_pricer):
return fcustom(Mediator,
return 0
trade.set_volume(wanted_size)
tolds.append(trade)
return trade.calc()
#平均损失函数，返回的是千分比表示的平均损失
def ev_lost(trade):
return trade.parent.evaluation.lostavg
def atr_lost(trade, times=1000):
return trade.atr * times / trade.tprice
atr_lost_2000 = fcustom(atr_lost, times=2000)
atr_lost_1200 = fcustom(atr_lost, times=1200)
def RPR(xt, y): #净值评估函数,xt为日期维x,y为相应净值
'''#根据海龟交易法则
计算方法来自http://www.scipy.org/Cookbook/LinearRegression
'''
(ar, br) = np.polyfit(xt, y, 1) #一阶拟合
xr = np.polyval([ar, br], xt)
err = sqrt(sum((xr - y)**2) / len(xt)) #标准差
#(a_s,b_s,r,tt,stderr)=stats.linregress(xt,y) #len(xt)必须>2，否则会有问题. 即[begin,end)包含的实际日期数必须大于2
#year_inc_rate = int(a_s * 365 * POS_BASE/b_s)
year_inc_rate = int(ar * 365 * POS_BASE / br)
#logger.debug('rar:year_inc_rate=%s,a=%s,b=%s,k=a/b=%s,stderr=%s,err=%s',year_inc_rate,a_s,b_s,a_s/b_s,stderr,err)
logger.debug('rar:year_inc_rate=%s,a=%s,b=%s,k=a/b=%s,err=%s',
logger.debug('second wanted volume is too small : %s %s',trade.tstock,trade.tdate)
trade.set_volume(0)
return 0
trade.set_volume(wanted_size)
tolds.append(trade)
return trade.calc()
#平均损失函数，返回的是千分比表示的平均损失
def ev_lost(trade):
return trade.parent.evaluation.lostavg
def atr_lost(trade,times=1000):
return trade.atr * times / trade.tprice
atr_lost_2000 = fcustom(atr_lost,times=2000)
atr_lost_1200 = fcustom(atr_lost,times=1200)
def RPR(xt,y): #净值评估函数,xt为日期维x,y为相应净值
'''#根据海龟交易法则
计算方法来自http://www.scipy.org/Cookbook/LinearRegression
'''
(ar,br)=np.polyfit(xt,y,1) #一阶拟合
xr = np.polyval([ar,br],xt)
err=sqrt(sum((xr-y)**2)/len(xt)) #标准差
#(a_s,b_s,r,tt,stderr)=stats.linregress(xt,y) #len(xt)必须>2，否则会有问题. 即[begin,end)包含的实际日期数必须大于2
#year_inc_rate = int(a_s * 365 * POS_BASE/b_s)
year_inc_rate = int(ar * 365 * POS_BASE/br)
#logger.debug('rar:year_inc_rate=%s,a=%s,b=%s,k=a/b=%s,stderr=%s,err=%s',year_inc_rate,a_s,b_s,a_s/b_s,stderr,err)
logger.debug('rar:year_inc_rate=%s,a=%s,b=%s,k=a/b=%s,err=%s',year_inc_rate,ar,br,ar/br,err)
#logger.debug('rar:ar:%s,br:%s',ar,br)
#收盘价买入，下限突破价卖出，必须有下限突破线
cl_pricer = (lambda s : s.transaction[CLOSE],lambda s : s.down_limit)
#开盘价买入，下限突破价卖出，必须有下限突破线
ol_pricer = (lambda s : s.transaction[OPEN],lambda s : s.down_limit)
#开盘价买入，开盘价卖出
oo_pricer = (lambda s : s.transaction[OPEN],lambda s : s.transaction[OPEN])
#收盘价买入，开盘价卖出
co_pricer = (lambda s : s.transaction[CLOSE],lambda s : s.transaction[OPEN])
#可定制自己的pricer，如
#my_pricer = (lambda s : s.buyprice,lambda s : s.sellprice)
#定制的Mediator
#一次买入一次买出，买入信号次日有效(开盘买入)，卖出信号当日起效
Mediator10 = fcustom(Mediator,trade_strategy=B1S0,pricer = ol_pricer)
#允许连续买入一次卖出，买入信号次日有效(开盘买入)，卖出信号当日起效
CMediator10 = fcustom(Mediator,trade_signal_maker=make_trade_signal_advanced
,trade_strategy=B1S0,pricer = ol_pricer)
OMediator10 = fcustom(Mediator,trade_signal_maker=make_trade_signal_advanced
,trade_strategy=B1S0,pricer = ol_pricer)
def mediator_factory(trade_signal_maker=make_trade_signal_advanced,trade_strategy=B1S0,pricer = cl_pricer):
return fcustom(Mediator,trade_signal_maker = trade_signal_maker,trade_strategy = trade_strategy,pricer=pricer)
dummy_trades = [Trade(-1,0,0,1),Trade(-1,0,0,-1)]
class MM_Mediator(Mediator):
''' 只用于计算mm_ratio的mediator
'''
def _calc(self,tmaker,sdata,dates,begin=0,**kwargs):
raise NotImplementedError(u'MM_Mediator不能调用_calc')
return trade
def atr_extra_custom(trade,stock,index): #将atr值赋给trade，专门定制，已经在后面被一般方法覆盖
trade.atr = int(stock.atr[index]) #将atr转换为平凡的int类型，而非numpy.int32，以便后续处理
return trade
def append_attribute_extra(trade,stock,index,attr_name_from,attr_name_to=None):
if attr_name_from in stock.__dict__:
if not attr_name_to:
attr_name_to = attr_name_from
trade.__dict__[attr_name_to] = int(stock.__dict__[attr_name_from][index]) #将atr转换为平凡的int类型，而非numpy.int32，以便后续处理
else:
logger.warn('append attribute error:%s do not have attribute %s',stock.code,attr_name_from)
return trade
atr_extra = fcustom(append_attribute_extra,attr_name_from='atr') #atr_extra的一般定义方法
def make_trades(stock,signal,tdate,tpositive,tnegative
,begin=0,taxrate=125,trade_strategy=buy_first,extra_func=default_extra):
''' stock为stock
ssingal为买卖信号,对于次日买卖的信号，输入前需要将signal roll(1)
tpositive,tnegative为信号值为正和负时的选择价格
taxrate为税率，默认为千分之八
begin为起始交易日
trade_strategy为交易方式，先买后卖，先卖后买，还是均可
以买入开始计算
'''
assert len(tpositive) == len(tnegative) == len(signal)
#print signal,'=signal'
sis = signal.nonzero()[0] #非0信号的index
slen = len(sis)
