MC,
Someone here got hourly data back to 1999. I put it in a spreadsheet and calculated the total return for every single combination of rsi under, over and stop loss. Stop loss increment was 0.1% and over under numbers where whole numbers (e.g., 30/50). I ran every combo (it took hours to run) including the one Noc uses. There was no combination that turned out to be as profitable as buy and hold. That's quite a statement remember since there was a crash.
Since 2005 however, there are hundreds of combinations that do better than b&h. The best combination was not the same as the one Noc used and I published it several times. Actually, I had two combinations--one for buys and a different one for sells because I optimized each separately to increase the return. Noc thought it was interesting but did not use the optimized numbers.
By the way, I got the same returns using Noc's numbers that he got and published. So, I think the results are correct. Also, I always matched the daily rsi with vto when I ran a daily model, so I am sure the formula I use is correct.
I don't want to go back and revisit the rsi hourly. My time is limited and I have other ideas. It is working fine now, but who knows how long it will continue. I think Noc added something to make it work when back-testing, but I don't recall what it was.
News 
Market Data 
Discover 
