I hope you didn't mean that I gave you ideas to improve the LQD:TLT system. That system might look good over the last several months, but a longer historical study of that system (which is dead simple to do) will show you it's a nice fruit while on the tree but will eventually drop and rot.
I have a general rule that an indicator should not be used until you know how to calculate it and understand it fully. Otherwise, they are misused. To learn how to calculate most indicators from scratch, you can find good descriptions at Stockcharts. It's also a good idea to learn how the inventors of the indicators intended they be used -- most message board posters misuse indicators by using their default values for all situations.
In it's early days ATR was used to set stop losses in commodities trading. To find out how ATR is used for setting equity stop losses, you can search the internet and get lots of hits. I find most of those useless because they multiply ATR by integers, such as 2 X ATR, to calculate the offset. In my mind that only works if you have a very, very low volatility equity trade, or you have a huge profit.
As shown in the system I exposed, I only use fractional values of ATR to set stop losses, and the system only uses stop losses to get me out of a trade that otherwise shouldn't have been entered in the first place. Those bad entries will happen when the moving averages are pinched together and there is no clear direction. Once the direction has been clarified by the separation and proper ordering of the moving averages, the stop loss is removed (or maybe a better choice is to leave it but widen it significantly so that a good trade isn't accidentally dropped because of a one-day news event that will dissipate).
There is another way to use ATR which I use in another system which works better than the one I showed. It is used all the time in that system and not as a stop-loss. The system has no stop-loss, ever, because the system is always in a trade, either long or short. Instead, a fraction of ATR is used to offset the trigger line in the direction that will require a larger price move in order to trigger the trade in the first place. The penalty for doing that is an overall drop in trading efficiency which is small per trade but accumulates over many trades.
-----
Allow me to give you new idea which I have started studying a couple of weeks ago. It places a narrow band on the moving average (or line, whatever creates it) which is being used to trigger a system Buy/Sell trade. The width of the band is a fraction of that equity's or index's ATR, and it is symmetrical about the line. A trigger is immediately given if price leaps beyond the band. A trigger is not given if price lands within the band, even if price crosses the line within the band. When this occurs, an additional breakout test is initiated which I call modified-NR5. (NR5 stands for Narrow Range 5, and is a five-candle pattern. It's the same as NR7, only shorter. You can find information on that by doing a search. If you need help, I can do that when I have more time.) Once NR5 is invoked, a trigger is inevitable and will be in the direction of the breakout from the NR5 pattern. An exception to that is if there is a breakout of the band (during the NR5 process) plus an additional trigger caused by a different test (for instance, the crossing of a different line). You can also look at using an NR4 or NR3, but I think the NR3 is too short to be useful.
