Regrettably I’m not as gifted as you obviously are. I enjoy the work immensely, even if I can’t follow all of it. But if you follow the above system long 100% when the slopes are all rising, intuitively it seems you would likewise follow the system short 100% when the slopes are all rising, because the system is working. It appears the shorts go 100% only if all the slopes are declining and go 50% short when the 50 day is rising. What am I overlooking?

Hi larry, thanks for the compliment! Actually its very simple: the 3 year slope ultimately determines whether the bias should be long or short. If the 3 year slope is up the system can only be considered for trading it long—-from here the 50 day and 252 day slope dictate position size. If the 1 year slope is negative we go to cash regardless of where the 50day is. If the 3 and 1 year are in alignment, we go 100% long if the 50 day slope is positive, and only a half bet size if the 50 day slope is negative. This is to prevent a major drawdown.For shorts we do that exact opposite—and when i mean shorting—i mean we are effectively reversing the rules of the system. One way to simplify this to one system is to simply invert the rules as discussed and simply apply the long rules to keep it less confusing. So if for example in this case we bought after up days and shorted after down days, if the 3 year slope became negative we would now buy on down days and short on up days.

Is the assumption in the underlying system that you enter your position at the close of the market or on the close of the following day? I assume for the former?

One more question – how are you calculating the slope of the equity curve – linear regression, or just taking the value at the beginning of the period and the value at the end of the period to calculate a line?

David, (polite cough) what are the data sources for your x and y values in this function? Obviously one is the (running?) equity curve but I am unsure of the other (the close?).

It appears that managing the equity curve efficiency is inversely proportional to how well the underlying strategy is working. In this example, daily follow through works well in the 1982-2000 bull market and the method lags. As the market transitions to mean reverting and the original strategy falls apart then the method shines.

Perhaps applying this method to strategies with t-stat < -1.6 (i.e. declining equity curves) would be optimum?

hi kevin, those are all good points. previously i had showed the impact of using the t-stat in the Adaptive Time Machine article. this is a brute but effective method. bear in mind that one issue is the 50 day slope—as a strategy performs really well, exiting under the 50 day slope for a 50% position is very costly. one method to rectify the disparity you pointed out is to trade the relative performance between the timed equity curve and the buy and hold equity curve as long as the 3 year slope dictates the strategy is in favor.other methods to improve results would be to sell at extreme performances and buy on pullbacks when the strategy is really doing well. Of course none of what i demonstrated is optimal, and what is used behind the scenes is obviously more designed to capture the “best” method.

Thanks for the fast response on ways to improve the method. I was thinking that to purposely look for a system with a negative equity slope and apply the method ‘as is’ would be a winner (extrapolating from this one example!). Perhaps due to ‘gaining back’ the 50 day slope cost you mentioned.

Very cool DV.

thanks wood!!

dv

Regrettably I’m not as gifted as you obviously are. I enjoy the work immensely, even if I can’t follow all of it. But if you follow the above system long 100% when the slopes are all rising, intuitively it seems you would likewise follow the system short 100% when the slopes are all rising, because the system is working. It appears the shorts go 100% only if all the slopes are declining and go 50% short when the 50 day is rising. What am I overlooking?

Hi larry, thanks for the compliment! Actually its very simple: the 3 year slope ultimately determines whether the bias should be long or short. If the 3 year slope is up the system can only be considered for trading it long—-from here the 50 day and 252 day slope dictate position size. If the 1 year slope is negative we go to cash regardless of where the 50day is. If the 3 and 1 year are in alignment, we go 100% long if the 50 day slope is positive, and only a half bet size if the 50 day slope is negative. This is to prevent a major drawdown.For shorts we do that exact opposite—and when i mean shorting—i mean we are effectively reversing the rules of the system. One way to simplify this to one system is to simply invert the rules as discussed and simply apply the long rules to keep it less confusing. So if for example in this case we bought after up days and shorted after down days, if the 3 year slope became negative we would now buy on down days and short on up days.

hope that helps

cheers

dv

Thanks for the clarification and your patience.

Is the assumption in the underlying system that you enter your position at the close of the market or on the close of the following day? I assume for the former?

hi damian……entry is on the close same day and exit same day on close as the signal.

cheers

dv

One more question – how are you calculating the slope of the equity curve – linear regression, or just taking the value at the beginning of the period and the value at the end of the period to calculate a line?

its the slope of the line in excel (“SLOPE” function) which is the x/y fit for the regression.

cheers

dv

David, (polite cough) what are the data sources for your x and y values in this function? Obviously one is the (running?) equity curve but I am unsure of the other (the close?).

Thanks

John

hi john we use the excel “slope” function where the date and curve value are considered for y and x values.

best

dv

Hi David,

It appears that managing the equity curve efficiency is inversely proportional to how well the underlying strategy is working. In this example, daily follow through works well in the 1982-2000 bull market and the method lags. As the market transitions to mean reverting and the original strategy falls apart then the method shines.

Perhaps applying this method to strategies with t-stat < -1.6 (i.e. declining equity curves) would be optimum?

Kevin

hi kevin, those are all good points. previously i had showed the impact of using the t-stat in the Adaptive Time Machine article. this is a brute but effective method. bear in mind that one issue is the 50 day slope—as a strategy performs really well, exiting under the 50 day slope for a 50% position is very costly. one method to rectify the disparity you pointed out is to trade the relative performance between the timed equity curve and the buy and hold equity curve as long as the 3 year slope dictates the strategy is in favor.other methods to improve results would be to sell at extreme performances and buy on pullbacks when the strategy is really doing well. Of course none of what i demonstrated is optimal, and what is used behind the scenes is obviously more designed to capture the “best” method.

cheers

dv

Hi David,

Thanks for the fast response on ways to improve the method. I was thinking that to purposely look for a system with a negative equity slope and apply the method ‘as is’ would be a winner (extrapolating from this one example!). Perhaps due to ‘gaining back’ the 50 day slope cost you mentioned.

Kevin