Selectivity vs turnover

[Euler]

The method above is how I have solved every problem in the market. I then joined up the dots on each to come up with a general description of what was going on which led me to explore and refine specific points on each strategy.

[ruthlessimon]

The method above is how I have solved every problem in the market. I then joined up the dots on each to come up with a general description of what was going on which led me to explore and refine specific points on each strategy.

Would this be along the right lines :

1. Take a basic strategy
2. Look at all the losers & compare it to a variable (i.e. entry time) - calc the correlation
3. Look at all the winners & also compare it to the variable above - calc the correlation

If the correlation differs between steps 2 & 3, the said variable has had an impact on the strategy.

I'm tempted to have a crack at that, but again I do fear the "look-elsewhere effect" when performing step 2 - is it a valid concern?

[ShaunWhite]

I think the key phrase was 'come up with a general description of what's going on' beyond it being a restatement of the parameters.

It must surely be the case that variables can't just showing a correlation to outcomes, they also must have relevance in the general description. Without that 2nd confirmation, they're just a coincidence that might not persist or might even become detrimental. For instance, distance should have no bearing on a strategy that's related to reversion or momentum. If it does then it's inclusion will be a problem in the future when it randomises. The very parameter which pulled out the winning selections is the one that guarantees failure unless it's relevant.

Eliminate the irrelevant, but knowing what's irrelevant might only become obvious once the general description is known?

[ruthlessimon]

For instance, distance should have no bearing on a strategy that's related to reversion or momentum.

Exactly

In fact, I'm tempted to build a spreadsheet around distance & using it to trade pre-race - just as a case study for randomness, & the dangers of curve fitting.

There will be an optimal distance than "predicts" steams/drifts over x time period

& again - this is why psychology is misleading. Mark Douglas won't be able to help if the strategy is curve fit to a random variable

[FrogThimble]

That's another tricky part. Especially if you're the sort of person who won't be beaten just because something's difficult, and when you're told that almost every strategy can be refined by discarding the negatives and improving the positives.

Yes, I'm very stubborn... But even I know that if I beat my head against the wrong brick wall for long enough then my skull will cave in before the wall ever crumbles.

[ShaunWhite]

That's another tricky part. Especially if you're the sort of person who won't be beaten just because something's difficult, and when you're told that almost every strategy can be refined by discarding the negatives and improving the positives.

Yes, I'm very stubborn... But even I know that if I beat my head against the wrong brick wall for long enough then my skull will cave in before the wall ever crumbles.

I'm hoping for an indicator that's not quite so brutal

It would be so much easier and quicker to spot if I had a decent methodology.

[stueytrader]

Great discussion points added to my (rather simplistic) initial question all here.

I have to admit, one of the reasons I initially posted was because I am currently trying to reduce those 'niggly losses' that were mentioned earlier - if that is possible at all. I presumed I was getting quite a few smaller losses because I wasn't being selective enough in my initial approaches.
The tricky part when you become very selective is that of course opportunities as a mass set of trading become limited also, so you have to nail them more clearly....

[shortcut]

Been pondering this specific trade-off (pardon the pun) in my own trading.

From analysing my performance I can see this is something of a double-edged choice. Of course, being much more selective will likely improve performance in each specific race/trade. However, a larger turnover of trading and races will potentially gain more (given the assumption of still doing ok on average with selections).

Just wondering what people think about this operational issue and choice?

Stu.

Have you looked at other software?

I've not used it personally but would be interested to hear opinions of those that did.