Hypothesis: Yesterday's high/low is (un)important
|November 16th, 2009, 03:32 PM||#11 (permalink)|
Just starting out...
Futures Experience: Intermediate
Favorite Futures: ES
Posts: 22 since Nov 2009
Thanks: 4 given, 12 received
I have been working on implementing a strategy to trade gap fills and tested one similar to the 'rules' posted by verge. Basically, go long if the open is lower than the previous day low; go short if the open (9.30 am) is higher than the previous day high; close when the gap fills (i.e. price = prev high/low), or at EOD (4 pm), or according to a stop loss.
Here are some results for ES 15 min bars, trading on a $30K account using position sizing of 100 max DD...
First, the equity curve:
The strategy was profitable overall, but had a protracted and long drawdown early on.
Second, the performance results:
Only 6/11 years were profitable. The final acct equity was ~$33,400, which is >100% return on the starting equity ($30K). However, this was over 11 years and had a SD of ~$15,000. Average annual DD was 16.5%.
Third, the Monte Carlo simulation results:
The more realiztic estimate of Max DD was 58.9% at a 95% confidence level, which is substantial but not devastating. Estimated total return on equity was 174.6% or $52,380 -- as this was over 11 years, this equals 15.9% annualized ROI...not bad, but not the holy grain by any means.
Fading the gap is a valid strategy, but one that is not reliable or stable. The large potential drawdown means that (1) the minimum adequate capitilization required to avoid a blowout is high and (2) trading this strategy will require superhuman strength to stick with the strategy during a large drawdown. The fact that this strategy CAN be profitable is worth exploring, but would only make sense as a 3rd or 4th line option once better strategies become limited e.g. due to liquidity issues, or when a strategy is needed that relies only on price action and is totally uncorrelated to price trends*.
Anyone else have some results/insights on fading the gap?
*This is yet to be tested so might not be true.