@Tony,

Try this:

# Tony B momentum by Vladimir v2  
import pandas as pd  
# ---------------------------------------------------------------------------------------------------  
assets = symbols('TLT','TLH','IEF'); bond = symbol('IEI');  M = 21; MOM = [M*1, M*2, M*3, M*4, M*5];  
signals = pd.DataFrame(0., index = MOM, columns = assets); LEV = 1.0;  
# ---------------------------------------------------------------------------------------------------  
def initialize(context):  
    schedule_function(trade,date_rules.month_end(), time_rules.market_open(minutes = 65))     

def trade(context,data):  
    price = data.history(assets,'close', M*5 + 1,'1d')  

    for m in MOM:  
        mom = price.pct_change(m).iloc[-1]  
        for sec in assets:  
            if mom[sec] >= 0: signals[sec][m] = 1  
            if mom[sec] < 0: signals[sec][m] = 0 

    wt = LEV*signals.sum() / len(MOM) / len(assets)  
    wt[bond] = LEV - wt.sum()  

    for sec, weight in wt.items():  
        order_target_percent(sec, weight); record(**{sec.symbol: weight})  

    record(leverage = context.account.leverage)

Here is an approach which builds on what was started. Use the numpy where method to convert positive and negative returns into 1s and 0s respectively. Then add the resulting values. The result will be a numpy array. Probably create a series by adding an index to associate the values with the assets. One can also multiply by the desired 5% weighting.

def trade(context,data):  
    price = data.history(ETFS,'close',M*12,'1d')

    mom1 = price.pct_change(M).iloc[-1]  
    mom2 = price.pct_change(M*2).iloc[-1]  
    mom3 = price.pct_change(M*3).iloc[-1]  
    mom4 = price.pct_change(M*4).iloc[-1]  
    mom5 = price.pct_change(M*5).iloc[-1]

    mom1_positive = np.where(mom1>0, 1, 0)  
    mom2_positive = np.where(mom2>0, 1, 0)  
    mom3_positive = np.where(mom3>0, 1, 0)  
    mom4_positive = np.where(mom4>0, 1, 0)  
    mom5_positive = np.where(mom5>0, 1, 0)

    positive_counts = mom1_positive + mom2_positive + mom3_positive + mom4_positive + mom5_positive

    weights = pd.Series(positive_counts * .05, index=price.columns)

However, if one wants actual month to month momentum (rather than assuming 21 days in a month), then the pandas resample method works great.

def trade(context, data):  
    price = data.history(ETFS, 'close', M*12, '1d')

    # resample to just get the month end prices  
    # only look at the last 5 months (6 including current month)  
    last_5_month_end_prices = price.resample('M').last().tail(6)

    # momentum will be positive if the current price > than the month end prices  
    positive_momentum_to_date = last_5_month_end_prices.iloc[-1] > last_5_month_end_prices

    # sum the positive momentums ie the True values  
    positive_momentum_counts = positive_momentum_to_date.sum()

    weights = positive_momentum_counts * .05

The attached algo implements this latter approach.

Good luck

Hi @Vladimir, @Dan,

Thank you so much for your help and your advice