<![CDATA[Multi-pass Backtesting]]>Hello !

I'd like to rewrite this code for Multi-pass Backtesting to pass the competition filter.
What should I do ?

def pair_trading(data):
    pair = ['NAS:FWLT', 'NAS:PCAR']
    close = data.sel(field='close')
    price = close.sel(asset=pair).transpose('time', 'asset').fillna(0) #欠損値処理
    modelCoint = VECM(price.values, k_ar_diff=0, deterministic='na')
    res = modelCoint.fit()
    pairs_value = res.beta[0]*price.sel(asset=pair[0])+ res.beta[1]*price.sel(asset=pair[1]) #ポートフォリオ価値の作成
    diff = pairs_value.shift(time=-1) - pairs_value
    threshold = diff.std(dim='time')*2 #閾値の設定
    signal = np.sign(diff)
    
    weights = xr.zeros_like(close) #ポートフォリオウェイトの初期化

    for t in data.time:
        if (weights.sel(time=t)==[0,0]).any() and pairs_value.sel(time=t) >= threshold:
            weights.loc[dict(time=t)] = [1,-1]
            
        if (weights.sel(time=t)==[0,0]).any() and pairs_value.sel(time=t) <= -threshold:
            weights.loc[dict(time=t)] = [-1,1]
    
        if (weights.sel(time=t)==[1,-1]).any() and signal.sel(time=t) < 0:
            weights.loc[dict(time=t)] = [0,0]
            
        if (weights.sel(time=t)==[-1,1]).any() and signal.sel(time=t) >= 0:
            weights.loc[dict(time=t)] = [0,0]

    return weights

Best regards

]]>http://quantiacs.com/community/topic/415/multi-pass-backtestingRSS for NodeMon, 13 Jul 2026 18:42:36 GMTWed, 09 Aug 2023 05:55:27 GMT60<![CDATA[Reply to Multi-pass Backtesting on Thu, 10 Aug 2023 06:51:03 GMT]]>@eddiee

Hello.

This code looks to the future.
It is needed to train the model.
Pay attention to the name of the variable.

]]>http://quantiacs.com/community/post/1193http://quantiacs.com/community/post/1193Thu, 10 Aug 2023 06:51:03 GMT<![CDATA[Reply to Multi-pass Backtesting on Wed, 09 Aug 2023 15:18:29 GMT]]>@support is ".shift(time=-1)" really forward looking?
In the Quantiacs Ridge Classifier example ".shift(time=-1)" is used, so I guess this operation is not forward looking. 🙂

def get_target_classes(data):
    # define categorical targets
    price_current = data.sel(field="close").dropna("time") # rm NaN
    price_future = price_current.shift(time=-1).dropna("time")

    class_positive = 1
    class_negative = 0

    target_is_price_up = xr.where(price_future > price_current, class_positive, class_negative)
    return target_is_price_up.to_pandas()
]]>http://quantiacs.com/community/post/1192http://quantiacs.com/community/post/1192Wed, 09 Aug 2023 15:18:29 GMT<![CDATA[Reply to Multi-pass Backtesting on Wed, 09 Aug 2023 13:17:15 GMT]]>@cyan-gloom Hi, before that, are you sure that your code is not looking forward? There is a shift with a negative value "-1" for the variable "diff", which in turns is used via "signal" for defining "weights". Naively, we would say that the code is looking into the future, correct?

]]>http://quantiacs.com/community/post/1191http://quantiacs.com/community/post/1191Wed, 09 Aug 2023 13:17:15 GMT