Issues calculating rolling max value

Overall, what I am trying to accomplish is to calculate the drawdown percentage from the 52 week high (ATHs might work too). It was not obvious to me if there was a factor to do this already so I attempted to create a custom factor. I am having issues calculating the trailing max value, which I would use as the 52 week high. I am still getting the hang of quantopian so apologies if I am missing something obvious.

I am receiving the below error. I believe this is more of a numpy error but I am curious if anyone has advice for either fixing this issue (I'm still looking into it) or demonstrating an alternative way of calculating the drawdown percent.

ValueError: could not broadcast input array from shape (252,8474) into shape (8474)

The error ValueError: could not broadcast input array from shape (252,8474) into shape (8474) stems from the following code

out[:] = (t_max - values) / t_max

The issue is (t_max - values) / t_max is an array with 252 rows (one for each day set by the window_length parameter) and 8474 columns (one for each asset). The out variable expects a single value for each asset so a 1 dimensioned array with length 8474 is expected. That is why the error is saying it cannot 'broadcast' one shape array into another shape.

There is a built in factor called MaxDrawdown (https://www.quantopian.com/docs/api-reference/pipeline-api-reference#quantopian.pipeline.factors.MaxDrawdown) however, that doesn't seem to be what you are looking for? If I understand correctly, the goal is to find the percent change from the 52 week high to the current close price? Something like this

(last_close - high_52_week) / high_52_week

This could be done in a custom factor like this

class DrawdownFactor(CustomFactor):  
    # Input the high and the close prices  
    inputs = [USEquityPricing.high, USEquityPricing.close]

    # Look back 1 year for the 52 week high  
    window_length = 252

    def compute(self, today, asset_ids, out, highs, close_price):  
        # Get the highest high for each asset  
        # Specify axis=0 since days are the rows and asset values are in each column  
        # The result will be a 1 dimensioned array  
        highest_high = np.nanmax(highs, axis=0)

        # Get the last row of close_prices which will be a 1 dimensioned array of the latest close price  
        latest_close = close_price[-1]

        # Output the percent change. The result will be a 1 dimension array with a value for each asset  
        out[:] = (latest_close - highest_high) / highest_high

You could also incorporate the number of days since the high. Here is a post which shows a factor for that which may be helpful
https://www.quantopian.com/posts/days-since-52-week-hi-slash-lo

Hope that helps.