class MinusOneRisk(RiskManagementModel):
    def __init__(self, maximumDrawdownPercent = 0.1):
        self.maximumDrawdownPercent = -abs(maximumDrawdownPercent)
        self.liquidated = set()

    def ManageRisk(self, algorithm, targets):
        targets = []
        algorithm.Log(len(self.liquidated))
        for s in algorithm.Securities:
            security = s.Value
            
            if not security.Invested:
                continue

            pnl = security.Holdings.UnrealizedProfitPercent
            if pnl < self.maximumDrawdownPercent or security.Symbol in self.liquidated:
                # liquidate
                targets.append(PortfolioTarget(security.Symbol, 0))
                if algorithm.Securities[security.Symbol].Invested:
                    self.liquidated.add(security.Symbol)
                    algorithm.Log(f"Liquidating {security.Symbol} to avoid drawdown")

        return targets

class MinusOneAlphaModel(AlphaModel):
    # Direction score close to 50 is good? https://www.quantconnect.com/forum/discussion/7082/alpha-score-vs-performance/p1
    def __init__(self, algorithm):
        self.leverage = 1
        self.algo = algorithm
        
    def Update(self, algorithm, data):
        # This is where insights are returned, which are then passed to the
        # Portfolio Construction, Risk, and Execution models.
        
        # The following Insight properties MUST be set before returning
        #   - Symbol        -- Secuirty Symbol
        #   - Duration      -- Time duration that the Insight is in effect 
        #   - Direction     -- Direction of predicted price movement 
        #   - Weight        -- Proportion of self.algo capital to be allocated to this Insight

        # onData gets called after onSecuritiesChanged here because we filtered the entire data first
        insights = []
        for symbolData in self.algo.symbolDataDict.values():
            symbol = symbolData.symbol
            symbolData.marketCap = self.algo.symbolToMarketCap.get(symbol,-1)
            # you need to update this in case some of the stocks persisted into next selection period, which won't be warmed up in onSecuritiesChanged
            security = self.algo.Securities[symbol]
            self.algo.symbolDataDict[symbol].update_closes(security.Close)
            self.algo.symbolDataDict[symbol].update_highs(security.High)
            self.algo.symbolDataDict[symbol].update_lows(security.Low)
        if not self.algo.selection_flag:
            return insights
        self.algo.selection_flag = False
    
        symbolDataList = list(self.algo.symbolDataDict.values())
        # sort by market cap
        sorted_by_market_cap = [s.symbol for s in sorted(symbolDataList, key=lambda s:s.marketCap, reverse=True)]
        top = sorted_by_market_cap[:self.algo.top_by_market_cap_count//2]
        bottom = sorted_by_market_cap[-self.algo.top_by_market_cap_count//2:]
        # sort performance (from worst) for top market cap and performance (from best) for bottom market cap
        top_performances = {symbol : self.algo.symbolDataDict[symbol].period_return() for symbol in top if self.algo.symbolDataDict[symbol].is_ready()}    
        bottom_performances = {symbol : self.algo.symbolDataDict[symbol].period_return() for symbol in bottom if self.algo.symbolDataDict[symbol].is_ready()}
        sorted_top_performances = [x[0] for x in sorted(top_performances.items(), key=lambda item: item[1])]
        sorted_bottom_performances = [x[0] for x in sorted(bottom_performances.items(), key=lambda item: item[1], reverse=True)]
        
        # select securities to short and long (50-50)
        self.algo.long = sorted_top_performances[:self.algo.stock_selection//2]
        self.algo.short = sorted_bottom_performances[:self.algo.stock_selection//2]
            
        invested = [x.Key for x in self.algo.Portfolio if x.Value.Invested]
        for symbol in invested: # if they are not to be selected again, then they are liquidated
            if symbol not in self.algo.long + self.algo.short:
                #self.algo.Liquidate(symbol)
                insights.append(Insight(symbol, timedelta(30), InsightType.Price, InsightDirection.Flat, None, None))
        # reduce the insight valid period         
        insight_period = timedelta(hours=6)
        insights += [Insight.Price(s, insight_period, InsightDirection.Up, None, None, None, self.leverage/len(self.algo.long)) for s in self.algo.long]
        insights += [Insight.Price(s, insight_period, InsightDirection.Down, None, None, None, self.leverage/len(self.algo.short)) for s in self.algo.short]
                
        self.algo.long.clear()
        self.algo.short.clear()
        return insights

    def OnSecuritiesChanged(self, algorithm, changes):
        #self.Debug("change")
        for removed in changes.RemovedSecurities:
            self.algo.symbolDataDict.pop(removed.Symbol, None)
        for security in changes.AddedSecurities:
            security.SetLeverage(self.leverage)
            symbol = security.Symbol
            daily_history = self.algo.History(symbol, self.algo.period+1, Resolution.Daily)
            if daily_history.empty:
                algorithm.Log(f"Not enough data for {symbol} yet")
                continue
            if symbol not in self.algo.symbolDataDict.keys():
                symbolData = SymbolData(symbol, self.algo.period, self.algo)
                self.algo.symbolDataDict[symbol] = symbolData
                symbolData.warmup(daily_history)
                
class SymbolData():
    def __init__(self, symbol, period, algo):
        self.symbol = symbol
        self.marketCap = None
        self.algo = algo
        self.closes = RollingWindow[float](period+1)
        self.highs = RollingWindow[float](period+1)
        self.lows = RollingWindow[float](period+1)
        self.period = period
        
    def update_closes(self, close):
        self.closes.Add(close)
    
    def update_highs(self, high):
        self.highs.Add(high)
    
    def update_lows(self, low):
        self.lows.Add(low)
        
    def period_return(self):
        if self.closes[self.period] == 0:
            self.algo.Debug(self.symbol)
            return None
        return self.closes[0] / self.closes[self.period] - 1
        
    def is_ready(self) -> bool:
        return self.closes.IsReady and self.highs.IsReady and self.lows.IsReady and self.closes[self.period] != 0
        
    def warmup(self, daily_history):
        if daily_history.empty:
            return
        closes = daily_history.loc[self.symbol].close
        for time, c in closes.iteritems():
            self.update_closes(c)
        highs = daily_history.loc[self.symbol].high
        for time, h in highs.iteritems():
            self.update_highs(h)
        lows = daily_history.loc[self.symbol].low
        for time, l in lows.iteritems():
            self.update_lows(l)

from alpha import MinusOneAlphaModel
from risk import MinusOneRisk

class MinusOneAlphaAlgorithm(QCAlgorithm):
    def Initialize(self):
        
        # The blocked section of code below is to remain UNCHANGED for the weekly competitions. 
        # 
        # Insight-weighting portfolio construction model:
        # - You can change the rebalancing date rules or portfolio bias
        # - For more info see https://github.com/QuantConnect/Lean/blob/master/Algorithm.Framework/Portfolio/InsightWeightingPortfolioConstructionModel.py
        # 
        # Use the Alpha Streams Brokerage Model: 
        # - Developed in conjunction with funds to model their actual fees, costs, etc. Please do not modify other models.
        ###############################################################################################################################
        self.SetStartDate(2015, 3, 1)   # 5 years up to the submission date
        self.SetCash(1000000)           # Set $1m Strategy Cash to trade significant AUM
        self.SetBenchmark('SPY')        # SPY Benchmark
        self.SetBrokerageModel(AlphaStreamsBrokerageModel())  
        self.SetExecution(ImmediateExecutionModel()) 
        self.SetPortfolioConstruction(InsightWeightingPortfolioConstructionModel())
        ###############################################################################################################################
        # Do not change the code above 
        
        self.symbol = self.AddEquity('SPY', Resolution.Daily).Symbol
        
        self.coarse_count = 500
        self.stock_selection = 10 # 5
        self.top_by_market_cap_count = 100
        
        self.period = 2
        self.SetWarmUp(timedelta(self.period))
        
        self.long = []
        self.short = []
        
        # symbolData at each moment
        self.symbolDataDict = {}
        self.symbolToMarketCap = {}
        self.SetBenchmark("SPY")
        self.day = 1
        self.selection_flag = False
        self.UniverseSettings.Resolution = Resolution.Daily
        self.AddUniverse(self.CoarseSelectionFunction, self.FineSelectionFunction)
        self.Schedule.On(self.DateRules.EveryDay("SPY"), self.TimeRules.BeforeMarketClose("SPY"), self.Selection)
        #self.AddRiskManagement(MaximumDrawdownPercentPortfolio(0.1,True))
        self.AddRiskManagement(MaximumDrawdownPercentPerSecurity(maximumDrawdownPercent = 0.1))
        # Add the alpha model and anything else you want below
        self.AddAlpha(MinusOneAlphaModel(self))
        
        # Add a universe selection model
        self.AddUniverse(self.CoarseSelectionFunction, self.FineSelectionFunction)
        #
    def CoarseSelectionFunction(self, coarse):
        #self.Debug("coarse")
        if not self.selection_flag: # self.selection_flag is only true when day is 5 or it is a Friday.
            return Universe.Unchanged
    
        selected = sorted([x for x in coarse if x.HasFundamentalData and x.Market == 'usa' and x.Price > 1],
            key=lambda x: x.DollarVolume, reverse=True)
        selected = [x.Symbol for x in selected][:self.coarse_count]
    
        return selected
        
    def FineSelectionFunction(self, fine): # the long and short lists are updated daily
        #self.Debug("Fine")
        if not self.selection_flag: # self.selection_flag is only true when day is 5 or it is a Friday.
            return Universe.Unchanged
        fine = [x for x in fine if x.MarketCap != 0]
            
        sorted_by_market_cap = sorted(fine, key = lambda x:x.MarketCap, reverse = True)
        top_by_market_cap = {
            "top":[x.Symbol for x in sorted_by_market_cap[:self.top_by_market_cap_count//2]],
            "bottom":[x.Symbol for x in sorted_by_market_cap[-self.top_by_market_cap_count//2:]]
        }
        
        # get the top market cap and bottom market cap (50-50)
        filtered = top_by_market_cap["top"]+top_by_market_cap["bottom"]
            
        filteredFine = sorted_by_market_cap[:self.top_by_market_cap_count//2]+sorted_by_market_cap[-self.top_by_market_cap_count//2:]
        for f in filteredFine:
            self.symbolToMarketCap[f.Symbol] = f.MarketCap
                
        return filtered
    def Selection(self):
        if self.day == self.period:
            self.selection_flag = True
            
        self.day += 1
        if self.day > self.period:
            self.day = 1