Oh no, that's not it - by leave-one-year-out, I simply mean I'm testing the algo on data that it was not trained on. This period could be at the beginning, middle, or end of the available dataset, but obviously it makes sense to chose the most recent period, if you plan on implementing it into the future.
Metrics for quantifying an intraday strategy?
- Thread starter Punta
- Start date
Oh no, that's not it - by leave-one-year-out, I simply mean I'm testing the algo on data that it was not trained on. This period could be at the beginning, middle, or end of the available dataset, but obviously it makes sense to chose the most recent period, if you plan on implementing it into the future.
Wysiwyg
Everyone wants money
- Joined
- 8 August 2006
- Posts
- 8,428
- Reactions
- 284
The ASX has been predominantly down daily so that graph is another sign of b.s..
The graph was actually plotted from a "long-only" strategy.
Doesn't mean it's B.S if all the down moves happen over night and during the day prices move to fill the overnight gaps - improbable but not impossible.
The strategy is based on indicators over the last few days. If 60% of the positions you enter are successful, you will make money (assuming no skew in the up/downside), and like I said (and the graph shows), this strategy actually makes money on 75% of the days it trades.
As people have said, maybe the brokerage fees on lots of small trades will kill this (but then again it is possible that fewer larger positions will work equally well).
Wysiwyg
Everyone wants money
- Joined
- 8 August 2006
- Posts
- 8,428
- Reactions
- 284
You're right mate.
- Joined
- 1 May 2007
- Posts
- 2,904
- Reactions
- 52
Think you have it the other way around. Liquidity at the close is massive. Open is pretty crap. You can definitely use the close price, but treat the open with a grain of salt.
You might want to put in some liquidity filters. Eg, you can only do 10% of a day's volume. So that'll filter out small stocks with large moves ticked on a few shares.
