End of week systems
- Thread starter newbunch
- Start date
Which is why I use EOW (saves a lot of Excel rows and CPU usage). I've tested my system and EOW performs just as well as EOD (obviously, it's a long term system, I haven't traded in 7 weeks).Quote from kut2k2:
Shouldn't be too hard to adapt an EOD system to EOW. Just a matter of tossing away ~80% of the data points.
1. Larger compared to what? All long-term systems have "larger" drawdowns. Of course, drawdown also depend on how much risk you take. But the real problem, imo, is not the drawdowns, but the time to recovery. Right now, my joint (bond and stock) system is having it's worst six month period since 1999. No, I'm not losing a lot at all, but the waiting to recover is hard.
2. Of course I miss moves. But if I traded EOD, I'd still miss intraday moves. And I don't catch all the weekly moves either. In fact, my backtested system makes money 62% of weeks. So I obviously miss about 38% of the weekly moves in addition to any intraweek/intraday moves.
3. My system only trades the S&P and 10-year bond. I have built other profitable systems, but have found that they do not improve my return vs risk even when accounting for diversification. Right now, I'm short both the S&P and 10-year bonds (getting hit this week). I've been short stocks since 11/25/05 (S&P was at 1268) and short bonds since 1/27/06 (TNX was at 4.34%). In reality, I am also trading cash since each system determines whether S&P/bond will earn more or less than cash.
I've been building my system for the past 5 years. I haven't made any significant changes in 7 months because I haven't been able to find any new ideas that work (despite reading countless books and magazines on the subject). So now I'm asking on ET, though I'm not expecting much since I've already tried the most common stuff and less common stuff and stuff I invented myself.
Many of the common ideas in trading is totally wrong. The whole idea of max-drawdown or VaR is fraught with errors. I'll give an example: is 90% drawdown twice as bad as losing 45%? If you have a million dollars and lost 45%, you are left with $550K. If you lost 90%, you have $100K left. So you lost twice as much, but you have only 18% as much money, so you'd be 5.5 time better off if you only lost 45%. To be even more extreme, is losing 150% three times as bad as losing 50%. On the million dollars, in one case you still have $500K and the other case you are $500K in the hole.
So the real question is how to measure risk? Max drawdown? Stdev? Exposure?
In reality, I use more than one measure for risk and I have tweaked them to fix problems such as the above example.
Thanks for your well wishes. On days/weeks like this one, I need them.
That is an interesting way to look at drawdowns - I had not given much thought to it that way. Have you come up with an alternative measure or are you pointing out the problem?
Before answering, let me say that I firmly believe that you should figure this stuff out for yourself. There's a quote in Education of a Speculator: "The player who knows how will usually draw, the player who knows why will usually win." Tom Wiswell (World Checkers Champion, 1951-76) By solving these problems yourself, you'll have a better grasp of what you are using and will be able to further improve upon it. If somebody gives you the answer, you don't know if the answer is good, don't understand it, and can't improve upon it (at least not as much).
Now, to answer: don't look at the drawdown, look at the return required to recover that loss. A 50% drawdown requires a 100% recovery. A 75% drawdown requires a 300% recovery (so a 75% drawdown is 3 times worse than a 50% drawdown). A 100% drawdown would require an infinte recovery (and is infinitely worse than any other loss since you cannot recover from losing everything). I use the recovery rates as my risk instead of the drawdown. It's like a geometric drawdown instead of a arithmetic one. By doubling your "risk" as measured by drawdown, you are more than doubling the required return to recover your loss.