When are you optimizing too much?
- Thread starter earlyexit
- Start date
Quote from earlyexit:
I've recently started using Ninja Trader.
If you optimize a 1-min strategy over a year.. Do you wonder if that is too much of a sample? I mean, things change so much over a years time. Should you really sample for that long of a period (be easy on me, I'm a self admitted automated newbie).
My initial thought was, you should really be focusing on the not so distant past. Like the last 3-6 months for a 1-min strategy. Perhaps I was incorrect assuming such.
For an example. I've been trading for several years now. Strategies (while not automated, but in my head) came and went. I wouldn't back test them. I just used them until I noted they stopped working and I moved on.
Thanks
You have good thinking.
IMO and this is *IMO*
Last 3-6 months is fluke territory
6-12 months will most likely provide a better overall makeup of what has been going on
Now updating / refining your code once you have this is a different story in these timeframes.
There is a liquidity provider called TradeWorx that never has the same strat used more than 1 time. Everyday it is different. Or they have said that at least. Doesn't have to mean it is the truth.
Anyway, everyone has an opinion. So there is no "rule". I think if you want a real sample though it should be over 500+ then for 1 half year / 6 months or at least 300-500 trade sample size.
But once you have this IMO it should try to be refined 1 time every few weeks / 1 time a month. This can change though if you get into something like earlier this year where VIX goes crazy.
So I guess like everyone that does something like you and I are trying to accomplish has their own "methods". So yeah, once you have been doing this for a great length of time perhaps you can come back to this thread and tell us whats up.
Not if you have a sample size over 4000 winning 85% of the time.
That is why PF is just total shit indicator. Everyone loves it though bc it is the easiest to manipulate.
What do you mean? PF is one of the few non-subjective indicators. It must be > 1 regardless of what your system does or involves, no matter what. Just look at PF, nothing else. If it is above 1.6 then your system may have a chance. If it is below 1.6, do not look at anything else. You will be wasting your time. Most random systems have a PF around 1.3 - 1.4. If you want to beat a monkey in his own game, you have to do at least PF = 1.6.
Many fund managers look at the PF of candidate advisors. There you need at least a year's performance with PF > 3 to have a chance.
Why?
I can lose 3 : 1 amount when I win over 3/4 times.
Are you sure you aren't thinking Sharpe?
EDIT:
You aren't going to find a lot of med-high freq programs that have high profit factors. Not in all my research at least.
You don't understand PF. Sorry.
The way PF works is that you add up the dollar value of all your wins in the sample, and divide by the dollar value of all your losses in the sample. If your PF is less than 1, it means the method in question LOST money over the sample.
While not all PFs greater than 1 are interesting methods, all methods with a PF less than 1 are uninteresting (or should be looked at in reverse).
Profit Factor = Avg Win / Avg Loss does it not?
How much experience do you have with systems that trade in multiples of 1000 times per year?