Best times for trading the DOW

Before I get into the subject matter, I should really comment on the previous post: what a load of tosh that turned out to be. Instead of seeing the bears run rampant, we saw new all-time highs on DAX and DOW, although the first week of Sept started very bearish.   NASD also made a good gain, after a heavy start, getting back to test the ATH supply, formed back in July.  It just goes to show that you can’t always rely on history repeating itself nor can you rely solely on statistics to accurately predict the future of price action.
​
The subject of statistics is the point of this blog post, once again – I am rather fond of them, despite the September-effect being irrelevant this year.  I’ve always remembered the formula for Spearman’s Rank Coefficient, for some odd reason.  I memorised if for my Statistics ‘A’ level exam, back in the days when Hugh Cornwell was the front man for my all-time favourite band, and was very disappointed to see it presented at the top of the exam paper. There is little else that I remember from my mathematical learning, other than the benefits of statistical analysis I’ve yet to use Spearman’s formula for trading, although have seen reference to it.  Instead, I’ve simply focused on win/loss ratios using the EA that’s been in forward-testing since June.  If, like me, you trade the DOW, from when the NYSE bell rings, you might find these stats to be of interest, or even relate to them from your own trading experiences.

As per previous stats blog post, the EA basically trades Fibonacci retraces, with confluence.  There is seemingly an infinite number of permutations and combinations for the setups.  As each week passes, I find more and code the extra variables, seeking to improve the profits and reduce the drawdown. All the trade history is posted to a CSV file, so - with the help of some pivot tables - I can see what works well and what should be avoided; that applies to the setups and, more recently as I’ve discovered, times. I was hoping to have this EA running nicely without the ‘tweakamisations’ by now but the markets keep throwing their curve balls, so am resigned to the work continuing until year end, if not into next year. 

Anyway, enough waffle and onto the stats. The data below is for a little over one year of EA-trading DOW – mostly back-tested results, although back testing is very close to forward testing, give or take some spread variations that can occur at random times.
  
Best months – bearing in mind that I have some overlap with August and September history: 

January and July top the list with a 32%-win rate.  April was a super month for profit but that had one trade with a 50% gain, so is a tad misleading.  August and September, as we’ve seen from our live trading, have been very tricky months. We had 17 losing trades in a row in Sept: no fun at all.   I’ll need another 4 years of forward-testing to get some real meaningful monthly stats – remind me to share the results in 2028.
​
Best days of the week to trade DOW: -

Thursdays are, without doubt, the best day of the week, followed by Tuesdays.  You might as well take Monday off, although even with a win rate of just 22.11% there was still a profit to be had, from 190 trades.  A return of 257R – or 257% (non-compound) gain if you risk 1% per trade – from just trading on a Thursday is impressive.

Best hour on any day of the week – this is broker time (GMT+2) with the NYSE bell going ding ding at 16:30.  Note that I’ve previously found that 20:00 has a very low win rate, so the EA only takes a trade on the hour, not during it: there have been two such trades.

Apart from 20:00, the lowest chance of success is in the hour after the NYSE opened.  The first half hour generates the best returns but with just a 1 in 4 chance of success; that’s perfectly understandable given the often-seen whipsaw followed by the long duration for which the trade might stay open. The best chance of success is during 21:00 but the returns aren’t so great, due to the limited time left for the trade to stay open: the EA closes any open trades before the markets close and, it’s worth pointing out, will only take one trade during the first 17 minutes after the bell rings.

​If we look at every hour in each day of the week, with focus on there being a reasonable number of trades taken, the stats show that the first half hour of Thursdays is the time to trade, if you can only spare that amount of time in a week. Thursday 19:00 also generates good returns with a 47% win rate, but there’s only a sample of 17 trades for that time. Friday 16:00 is crap, although … law of sods ... this Friday, with that time switched off for the live account EA but not the demo account EA, it did this trade, for over 5R; would have been 7R by sticking with the TP and not activating the trailing trade management.

I’ll not go into all the stats now, but I’ve also gathered additional data for success rates: the day before US holidays (bad) and the actual US holiday (surprisingly good); the days before FOMC (bad) and FOMC day after 18:00 (good); Triple Witching days aren’t so good but better if you wait until after 19:00 (although I only have 5 of those days in my stats); GDP news day – bad when it’s on a Wednesday but good on a Thursday ... and so on ...

These stats are from the EA trading but I believe the times discussed should be relevant for many DOW traders who buy the dips and sell the rallies (basic price action trading, in a nutshell). It might be a different experience for the squiggly line or breakout traders (shudder at the thought).  I’ve found all this to be interesting and the data certainly improves the results for the robot. As with all stats though, they’ll need to be revisited as we get new data. Please drop me an email if your own experiences are completely different.