You're assuming "monthly data = huge drawdowns". Post proof, don't just assume. I have backtested weekly and monthly data drawdowns vs. daily EOD bars and you're wrong.
its common sense, no? the larger space between data points the greater potential range--- im not as kind as mr. king--- i think that study is extremely weak overall fundamentally flawed, particullarly with the monthly data useage. what are those guys living in the stone age?? chart by hand??
No it's not common sense. It's your personal assumption, nothing else. Are monthly drawdowns smaller if you trade tick data compared to EOD daily bars? Are monthly drawdowns smaller if you trade if you trade daily bars as compared to weekly bars? Are monthly drawdowns smaller if you trade weekly bars over monthly bars? Surf, you will never know. You don't backtest, you don't trade. You don't validate any of your ridiculous assumptions.
A very simple system trading monthly bars. Annualized ROR% ~39% End Of Month Max DD ~21% End of Week Max DD ~28% End of Day Max DD ~34% You're wrong, again. Case closed. Surf, why do you keep making assumptions without verifying them first. You don't trade do you?
It's trivially true that a series that's sampled less frequently will show shallower drawdowns. What's less clear is if there's a real effect in Mr Makloda's series, beyond the statistical artifact. But without more color on the underlying process (for all we know, it could be random numbers) it's hard to guess.
By definition a max drawdown on tick by tick measurement cannot be lower than a max drawdown on a higher timeframe, and will at least sometimes be higher. Therefore extending the timeframe will falsely minimise the apparent max drawdown. The right approach is to use the max tick DD for all time series. Any system which uses a stop-loss on any timeframe beyond tick data is taking potentially infinite risk on shorts and 100% risk on unleveraged longs, this risk increases with the length between stop loss/exit signals. Imagine a system with a 100 year timeframe between signals, are you sure this would have a lower max drawdown than one with a 1 day exit timeframe? When markets are in a mean-reverting state, the longer timeframe will reduce whipsaws, thus improving risk/reward. But when markets are in momentum mode the longer timeframe will penalize the system.
Sorry but those data points seem to refute your position more than marketsurfer's. I think you've misunderstood what he was saying (assuming I understand what he meant). N.B. all trading systems must be tested on tick data, not EOD data (let alone longer). Max DD must be measured on tick data. If the max DD on EOD was 34% then the max DD on EOM was also 34%. A drawdown does not disappear just because you decide not to measure it.
Since drawdowns occur in real-time, and market values are not suspended for 30 days each month, a less frequently sampled series can't have shallower drawdowns. If you disagree, ask a margin clerk. Furthermore, because of the 1 month delay between signals, the less frequent the trade signal, the more the risk. There is a chance of a correlated move across the system's positions over 1 month that would go way further than the stop-loss maximum on a system that took signals on tick data in real-time. Using anything beyond tick data for your exits is basically placing an artificial mean reversion bet in disguise. Like all mean reversions, most times it will work but occasionally it will go spectacularly wrong. For example, what if you are 100% long across 3 positions, then some news occurs which causes all 3 markets to get 99% correlated and fall 90% in the next 30 days? Using a 1 month signal you will lose 90%. Using a tick by tick exit you will lose far less than that.