Missed a winning trade. But I'm not upset. Really. OK, maybe a little.

If you've been following the blog much lately you'll know that I currently get a lot fewer trading signals because of some strict filters I've set up. So when a trading signal does come along, it's a big deal. And yesterday I got one to go short the EUR/USD for 22 pips, and in the ranging price action that followed it turns out it would have been a winning trade. Which would make it the 6th winning trade in a row, possibly setting a new record if I kept track of these things. But I couldn't make it, for a couple of reasons.

First, I was out hiking with some friends and then out having a beer at a favorite venue in San Francisco. So I missed the ideal entry point for my trade, which was at 5:00 pm, just as I was wandering around the forests south of here. When I got back, 10 pips profit had already been shaved off the trade, so it would've been only a 12 pip trade.

All of which is entirely my fault and if the trade had been enough of a priority I clearly would've been sitting right here glued to my laptop. And I wasn't, and while I'm a little annoyed with myself I still had a fun afternoon. And fun has a cost, which in this case can be measured in pips.

The second reason is that FX Engines, my trading platform, had a server go down and wasn't working at all. This is the first time they've gone down during trading hours and it came as a bit of a nasty surprise. Who knows, maybe if I'd showed up on time the site still would've been up and I could've made the trade, but there's not much point in wondering about it now.

Anyway, the upshot of all this was: no 22 pips, and a break in a nice winning streak. The larger lesson, and one I think I'll revisit in a more comprehensive post on the subject, is that there are a lot of unforeseen logistical risks to your trading lurking out there and you'd better be ready to fix the ones you can (like being on time for a trade, if that's important to you) and accept the ones you can't (like your broker's server going down). And if the same problems keep cropping up, it might be time to make some changes to your trading setup. But more about all that in a future post...

Risk/Reward Ain't Always What It Seems

It's very easy to get caught in the illusion that a tight stop-loss and a generous take-profit/limit order are going to earn you profits in that ratio. You know, setting your stop at 20 pips, your take-profit at 50 pips, and waiting for those consistent profits to start rolling in. This is a newbie error I made plenty of times before learning my lesson. What I've found since then is that a very generous stop-loss and a conservative take-profit strategy is often more consistently profitable.

A case in point: I'm currently running a GBP/USD trading system that features timed exits of around a week, take-profit at 100 pips, and stop-losses set at 300 pips. Yes, 300 pips. Sounds a bit weird, I know, but the key here is that those stop-losses almost never get hit. What's far more likely is that the price will advance 100 pips in that week, or the trade will time out if it doesn't (and sometimes with a profit). In fact, I could probably run the system without a stop-loss at all and the results would be similar, thanks to those timed exits - I'll have to backtest that notion soon, though the idea of trading without any stop-losses makes me nervous.

In general I've found that stop-losses work best for me in extreme emergencies, when the market is acting wilder than usual and hence could lose me more than usual. So for my trading style it makes sense to set them at the outer limits of the likely trading range, rather than squarely in the middle of the range where they're likely to get hit by a whipsaw or retracement.

Timed Exits

In an effort to keep tightening up my trades to cut short losses and lock in profits quickly, I've been playing around with some timed exit strategies for the past week or so (in case you were wondering why it's been so quiet around here).

In my regular daily trading I already have a very basic timed exit in place - after 24 hours in a trade, I'll either close it out or roll it over to the next day. But this is more a side-effect of the price data I used to design my trading system - I get the data in 24 hour units, so by default that's my minimum trading period.

However, I've been trying more aggressive trade-timing strategies via FX Engines, and so far have managed to crank up the backtested performance of my best GBP/USD engine by almost 1000 pips. I did this by using the time-out feature in their engine design tool, specifying that the engine automatically close out a trade after a certain number of days. The underlying idea is that a trade is likely to be profitable within a certain time window, and after that the chances of success drop off dramatically. This was already quite evident in the backtested results before I added a timed exit: successful trades averaged about 2.5 days in length, while unsuccessful ones tended to drag out for close to a week. So I figured, if the longest trades are most likely to fail, why not start cutting them short and see what happens?

After testing out that theory with a variety of timed exits, it turned out that there was indeed an optimal time period for my GBP/USD engine, after which the most profitable strategy was to just close out the trade.

This same principle might apply to your own trading. To find out, you could try timing every trade and then see how long a successful trade lasts, on average, versus how long an unsuccessful one takes. You could also monitor trades that closed quickly to see if they might have been profitable if they'd continued for a longer period.

Based on the data you gather, you may begin to notice patterns: perhaps you let your trades run on too long, hoping they'll turn around, when you should really be closing them quickly according to a strict timetable. Or on the other hand, maybe you're closing a lot of trades too soon because of tight stop-losses, or impatiently taking small profits instead of waiting for big ones. Whatever the answer, you'll be on your way to discovering the most profitable timeframe for your style of trading. Hope you find it!

Using time-of-day market analysis to improve your trading odds and profitability

I've periodically wondered if there are certain hours of the day whose trading activity acts as a microcosm of trends for that day. Is it possible to predict the market trend for a 24 hour period based on the price behavior from just one hour of the day? Obviously this could be a big help in placing trades and saving time, since you'd be able to place day-long trades after observing just one hour of the day. Since a major goal of mine is to reduce the amount of time I need to look at forex charts and data to a minimum, I decided to take a look at whether market activity during certain hours of the day could yield any clues as to how I should be trading that day - and I discovered some interesting correlations.

All my initial studies were of the EUR/USD pair over a period spanning six months, followed by further testing via FX Engines over a 2 year period. I then tested the EUR/USD results on the GBP/USD, and found the hourly correlations translated to that pair as well. The basic question I asked was: if the market is up in hour X, will it be up 23 hours later, at the end of the 24-hour period commencing with hour X? Or will it be down? And if the market is down at the end of hour X, will it be down 23 hours later, or up? You get the idea.

Most hours of the day proved to be uncorrelated to any particular market direction. However, a handful really jumped out at me for being closely correlated to the day's trend, either directly or inversely. Here they are - keep in mind the correlation isn't 100%, but a tendency in one direction or the other:

Direct correlations

8 am - 9 am, US EST: hourly uptrend correlates with an uptrend for the day (that is, the following 23 hours). Hourly downtrend correlates with a downtrend for the day.

8 pm-9 pm, US EST: Hourly uptrend correlates with an uptrend for the day. Hourly downtrend correlates with a downtrend for the day.

12 midnight - 1:00 am, US EST: Hourly uptrend correlates with an uptrend for the day. Hourly downtrend correlates with a downtrend for the day.

Inverse correlations

6 pm-7 pm, US EST: Hourly uptrend correlates with a downtrend for the day. Hourly downtrend correlates with an uptrend for the day.

2 am-3 am, US EST: Hourly uptrend correlates with a downtrend for the day. Hourly downtrend correlates with an uptrend for the day.

4 am-5 am, US EST: Hourly uptrend correlates with a downtrend for the day. Hourly downtrend correlates with an uptrend for the day.

I'm not sure why these correlations exist, or seem to exist, but my guess is it involves market openings in Asia, Europe or the US and the times during which trends are established and then reacted to or corrected for. Perhaps a strong Asian opening then leads to a correction in the 6 pm period, which then reverses back to the original direction - hence the inverse correlation.

The most important practical effect of this research is that it's allowed me to create my most profitable FX Engines yet - and not just for the EUR/USD pair, but for the GBP/USD as well. For example, I created a long GBP/USD engine that is triggered by an uptrend in the 8 am-9 am period, and in backtesting it surpassed the results of my next-best GBP/USD engine by over 500 pips. I also created a EUR/USD engine that places long trades based on downtrends in the inversely-correlated 2 am-3 am period, which surpassed my best EUR/USD engine by close to 1000 pips.

Of course, it's impossible to say whether these engines will continue to be profitable in the future, but initial results based on these time-of-day studies are extremely promising. Hope you find them helpful as well!

Market Activity on Different Days of the Week

I'd always wondered what effect, if any, the day of the week had on forex trading activity. Based on anecdotal observations, it seemed like there might be some differences between, say, early in the week vs. Friday. So I recently decided to test this theory and did some historical research on the EUR/USD pair going back almost four years to September, 2002. It turned out there were some remarkable differences in market behavior depending on the day of the week.

Before I get into what happened on each day, here's a little context on what the overall EUR/USD market did in that time. Between September 4, 2002 and Wednesday, June 7, 2006, the Euro increased in value by 2833 pips. So the long-term trend for this pair was long, and on 52.57% of the trading days in that time, the Euro increased in value vs. the dollar. This data alone can be an interesting benchmark for comparing your own trading results - for instance, if you've ever wondered whether you'd beat someone who just went long and held for that whole period.

A few important caveats about this study: the data I used was originally from this study of correlations between Dow Jones Industrial Average activity and EUR/USD trends. As such, it excludes Sundays and US holidays, so it's not a compete picture of all forex trading activity in that time. The reasons I had to use this data are fairly tedious and have to do with MS Excel formatting of dates. Which is too bad, since I've always thought Sunday trading was kind of weird and mysterious, and would've liked to be able to quantify it. At some point I may manually add data for Sundays and holidays, but right now the prospect of that is more boring than assembing a blank jigsaw puzzle. So anyway, apologies for the missing data, and take what you read here with that grain of salt.

Now, on to the data. The most dramatic activity I found was on Tuesdays:

The EUR/USD pair was long on 55.1% of Tuesdays, the highest percentage of any day. If you'd just placed long trades on Tuesdays since September, 2002, you'd have made 1499 pips, minus spread costs of 588 pips (assuming a 3 pip spread). So spread costs are a big part of the picture here.

Since this activity occurred in a market that was long overall, I'm curious whether Tuesdays would have had the most dramatically short activity in a bearish EUR/USD environment. To put it another way - are Tuesdays intrinsically long, or are they the day that most strongly encapsulates the overall market trend, long or short? My suspicion is that it's the latter, and in a short EUR/USD market, Tuesdays would have the most dramatic short activity.

The days with the next-highest pip totals were Thursdays and Fridays. Percentage-wise, the two days were fairly similar, with 52.63% of EUR/USD trades long on Fridays and 52.38% of trades long on Thursdays - note how close these numbers are to the overall long percentage of 52.57% for this period. However, Thursday's pip total was much higher, leading me to conclude that trades are likely to be more profitable on Thursdays. If you'd just placed long trades on Thursdays, you would've made almost twice as much as on Fridays: 1225 pips vs. 615 pips. In fact, on Fridays spread costs would've eaten up almost all your gains. Perhaps this reflects the fact that Friday is a shorter trading day, and some traders are leaving early for the weekend (?).

Mondays and Wednesdays had the least impressive totals, and if you'd traded long on these days, you would've come out in the negative after subtracting spread costs. Monday actually had only 48.31% long activity and an insignificant 39 pip gain for long trades, pre-spread, in the testing period. Quite a contrast with Tuesday's numbers.

While Wednesday had a relatively high percentage of long activity - 53.85% - this would've only yielded 220 pips on long trades, which would have ended up in the negative by a few hundred pips after spread costs.

Here's the overall summary of activity by day of the week between September 4, 2002 and June 7, 2006 - note that total number of days varies because I had to exclude holidays:

Mondays:
48.31% long days, 50.57% short days, 1.12% no change, 39 pip gain (long), 178 total days in the study.

Tuesdays:
55.1% long days, 44.9% short days, 1499 pip gain (long), 196 total days in the study.

Wednesdays
53.85% long days, 46.15% short days, 220 pip gain (long), 195 total days in the study.

Thursdays
52.38% long days, 46.56% short days, 1.06% no change, 1225 pip gain (long), 189 total days in the study.

Fridays
52.63% long days, 46.84% short days, .53% no change, 615 pip gain (long), 190 total days in the study.

Whether there's a larger significance to this daily data I can't really say. Maybe if you're thinking of cutting back your trading activity (for instance, if you think you've been overtrading) this data can help you focus your trading on the most active, most profitable days so you can go fishing the rest of the week. Or maybe it's just some interesting forex trivia that's only marginally relevant to the real world of trading.

More to the point, will I be using it to place a long trade this coming Tuesday? Well...no, I don't think so.