I'm trying to implement a QC version of QTradableStocksUS that simply simulates it's universe selection model, but I'm keep getting this error during the backtesting:
Runtime Error: Algorithm took longer than 10 minutes on a single time loop.
The algorithm screens about 1200 stocks every month, and then it took 200-day price data from History api for each stock, I figured out it's the part that caused slowness.
Is there a performance issue for QC/Lean? does it possbile to fix?
Please find the code below:
import numpy as np
import pandas as pd
### <summary>
### Basic algorithm simply simulates quantopian's QTradableStocksUS universe selection model
### </summary>
class QTradableStocksUSUniverseAlgorithm(QCAlgorithm):
def Initialize(self):
self.assets = []
self.universe_changes = []
self.rebalancing = True
self.max_positions = 20
self.total_shares = {}
self.primary_share_class_ids = {}
self.rolling_data = RollingData()
self.SetStartDate(2013,10, 7) #Set Start Date
self.SetEndDate(2017,10,11) #Set End Date
self.SetCash(100000) #Set Strategy Cash
# Find more symbols here: http://quantconnect.com/data
self.AddEquity("SPY", Resolution.Daily)
self.UniverseSettings.Resolution = Resolution.Daily
self.AddUniverse(self.coarse_selection, self.fine_selection)
self.Schedule.On(self.DateRules.MonthStart("SPY"), self.TimeRules.AfterMarketOpen("SPY", 10), Action(self.start_rebalance))
def OnData(self, data):
if self.universe_changes is None or self.universe_changes.Count < 2:
self.Debug("{}: No universe changed, skip rebalancing...".format(self.Time))
return
if not self.rebalancing:
return
if not self.assets:
return
self.rebalancing = False
self.rebalance()
#self.Debug("{}: Data changed.".format(self.Time))
def OnSecuritiesChanged(self, changes):
#self.Debug("{}: OnSecuritiesChanged fired: {}".format(self.Time, changes))
self.universe_changes = changes
def start_rebalance(self):
self.universe_changes = None
self.rebalancing = True
#self.Debug("{}: starting rebalancing...".format(self.Time))
def rebalance(self):
assets = self.assets#[:20]
self.Debug("{}: start rebalancing: assets = {}".format(self.Time, assets))
for i in self.Portfolio.Values:
if (i.Invested) and (i.Symbol not in assets):
self.Liquidate(i.Symbol)
for symbol in assets:
self.SetHoldings(symbol, 1.0/len(assets))
def coarse_selection(self, coarse):
if not self.rebalancing:
return []
#self.Debug("{}: No. of coarse symbols: {}".format(self.Time, len(coarse)))
filtered = [x for x in coarse if x.HasFundamentalData
#and x.Volume > 0.0
#and x.Price > 0.0]
and x.DollarVolume > 2.5e6]
filtered = sorted(filtered, key=lambda x: x.DollarVolume, reverse=True)
self.rolling_data.update(self, filtered)
return [ x.Symbol for x in filtered ]
def fine_selection(self, fine):
if not self.rebalancing:
return []
filtered = [x for x in fine \
if True #(x.CompanyReference.CountryId == "USA")
and (not x.CompanyReference.IsLimitedPartnership) # not limited partnership
and (not x.CompanyReference.IsREIT) # not REITs
and (not x.SecurityReference.IsDepositaryReceipt) # not ADR
and (x.CompanyReference.PrimaryShareClassID)
and (x.SecurityReference.ShareClassStatus == "A") # active trading
and (x.SecurityReference.CurrencyId == "USD")
and (x.SecurityReference.SecurityType == "ST00000001") # common stock
and (x.SecurityReference.ExchangeId != "OTC") # not OTC markets
and ('_WI' not in x.Symbol.Value)
#and x.EarningReports.BasicAverageShares.ThreeMonths * (x.EarningReports.BasicEPS.TwelveMonths * x.ValuationRatios.PERatio) > 5e8
]
self.total_shares = {}
self.primary_share_class_ids = {}
for x in filtered:
self.total_shares[x.Symbol] = x.EarningReports.BasicAverageShares.ThreeMonths
self.primary_share_class_ids[x.Symbol] = x.CompanyReference.PrimaryShareClassID
symbols = [ x.Symbol for x in filtered ]
assets = self.indicator_selection(symbols)
#assets = filtered[:self.max_positions]
self.Debug("{}: No. of fine symbols: {}".format(self.Time, len(assets)))
self.assets = assets
return assets
def indicator_selection(self, final):
symbols = []
data = []
count_new_ipo = 0
count_suspension = 0
count_low_volume = 0
count_micro_size = 0
for symbol in final:
try:
#algo.Debug("Breakpoint 3: %s" % symbol)
prices = self.rolling_data.closes.loc[str(symbol)]
#algo.Debug("Breakpoint 4: %d" % len(prices))
if prices['close'].count() < 180: # new IPOs, trading halts
count_new_ipo += 1
continue
if prices['close'].tail(20).isnull().sum() > 0: # stock suspension
count_suspension += 1
continue
market_cap = float(self.total_shares[symbol]) * prices['close'].tail(20).mean()
if market_cap < 350000000.0: # market cap < 350 Million
count_micro_size += 1
continue
dollar_volumes = prices['close'] * prices['volume']
dollar_volume_median = dollar_volumes.median()
if dollar_volume_median < 2500000.0: # day's volume < 2.5 Million
count_low_volume += 1
#algo.Debug("Symbol %s is low volume at %.2f" % (symbol.Value, dollar_volume_median))
continue
symbols.append(symbol)
data.append([ dollar_volume_median, self.primary_share_class_ids[symbol] ])
except KeyError:
self.Debug("Symbol %s is not exists in history data." % symbol.Value)
# remove duplicate share classes by the same company, select most liquidity share class
merged = pd.DataFrame(data, columns=['dollar_volume', 'primary_share_class_id'], index=symbols)\
.sort_values('dollar_volume', ascending=False)\
.drop_duplicates('primary_share_class_id', keep='first')\
.index.tolist()
self.Debug("Removed %d of duplicated companies." % (len(symbols) - len(merged)))
self.Debug("Removed %d of new IPO companies." % count_new_ipo)
self.Debug("Removed %d of suspended trading companies." % count_suspension)
self.Debug("Removed %d of low volume companies." % count_low_volume)
self.Debug("Removed %d of micro size companies." % count_micro_size)
self.Debug("No. of dynamic symbols: final=%d, merged=%d" % (len(final), len(merged)))
return merged[:self.max_positions]
class RollingData:
window_length = 200
resolution = Resolution.Daily
def __init__(self):
self.closes = pd.DataFrame(data=[], columns=['symbol', 'time', 'close'], index=[]).set_index(['symbol', 'time'])
def update(self, algo, corase):
for item in corase:
if str(item.Symbol) in self.closes.index.levels[0]:
self.closes.loc[(str(item.Symbol), item.EndTime),] = item.Price
else:
history = algo.History([item.Symbol], self.window_length, self.resolution)
prices = history.loc[str(item.Symbol)]
for time, price in prices.iterrows():
self.closes.loc[(str(item.Symbol), time),] = price['close']
self.closes.sort_values('time', ascending=True)
algo.Debug("{}: No. of records in rolling data: {}".format(algo.Time, len(self.closes)))
Jared Broad
We made a similar algorithm here, would this meet your criteria?
History requests are currently fairly slow. We're actively working on this now and have a solution which makes it 5x faster.
The material on this website is provided for informational purposes only and does not constitute an offer to sell, a solicitation to buy, or a recommendation or endorsement for any security or strategy, nor does it constitute an offer to provide investment advisory services by QuantConnect. In addition, the material offers no opinion with respect to the suitability of any security or specific investment. QuantConnect makes no guarantees as to the accuracy or completeness of the views expressed in the website. The views are subject to change, and may have become unreliable for various reasons, including changes in market conditions or economic circumstances. All investments involve risk, including loss of principal. You should consult with an investment professional before making any investment decisions.
John Ma
The material on this website is provided for informational purposes only and does not constitute an offer to sell, a solicitation to buy, or a recommendation or endorsement for any security or strategy, nor does it constitute an offer to provide investment advisory services by QuantConnect. In addition, the material offers no opinion with respect to the suitability of any security or specific investment. QuantConnect makes no guarantees as to the accuracy or completeness of the views expressed in the website. The views are subject to change, and may have become unreliable for various reasons, including changes in market conditions or economic circumstances. All investments involve risk, including loss of principal. You should consult with an investment professional before making any investment decisions.
To unlock posting to the community forums please complete at least 30% of Boot Camp.
You can continue your Boot Camp training progress from the terminal. We hope to see you in the community soon!