More and more people are moving to 15 and 30 minute charts for day trades, since it helps them avoid overtrading, and avoid getting chopped up in fast chart noise. It’s also just a lot more relaxing to watch! But, is that the best way to achieve these goals? In this article, I’ll describe an alternate approach to trading off slow X-minute charts.
A Motivating Example
I’ve been looking at examples of so-called Dummy Trades recently. I’m no expert, but it seems a major element is locating a narrow range 30 minute bar, and then trading with the overall trend. For long plays, the entry will be a break of the narrow bar’s high, and the initial stop will be below the bar’s low. Other than being a convenient way to identify a tight stop, I think the narrow bar represents a mini-consolidation period, after which a breakout on volume is likely to propel the price enough to make a good risk:reward trade.
But, this got me thinking… on a 30 minute chart, you’ll see consolidation from 10:00 to 10:30, no problem. But what if the consolidation is between 10:20 and 10:50, with bigger moves on either side? There’ll be no narrow bars on the chart, but isn’t this 30 minutes of consolidation just as good for trading?
(I’m using dummy trades as an example because I’ve just started reading about them, and they are fresh on my mind. But, I believe this article is applicable to any trading system.)
How 30 Minute Charts Can Hide Trading Opportunities
Slower charts like 30-minute or 15-minute candles can lead to better decisions, because they don’t bombard the trader with lots of noisy information. But, they do this by breaking the day into a few large-grain chunks, and a lot of useful information is hidden inside.
I’ll continue the dummy-trading type of example. Here’s 2 hours of (fake) 1-minute chart data (click to enlarge). There’s no time or scale on it, but let’s say this chart goes from 10:00 to 12:00, so it’s easy to talk about:
See how there’s a period of very narrow price action right before the drop at the end? Here’s the 30 minute candle version of this chart:
Where’s my consolidation? Where’s my low-risk entry? It’s invisible… Further, I’ll claim that all intraday bar and candle charts put a false significance on “open” and “close” values. I know what it means for a stock to close at $11 on Tuesday. I’m not sure I understand what difference it makes where a stock “closes” at 10:30, and 11:00, etc, other than you can sometimes see the effect of lots of traders looking for specific candle shapes.
An Alternative Approach
There’s more than one way to de-noise data, and I’m going to suggest that breaking my day into 30-minute chunks may not be the best way. What I’d like to do is retain the slow-changing, summary-level properties of 30-minute chart data, but get back the precision and responsiveness of a 1-minute chart. Clearly I’m going to have to compromise, somewhere.
In my past life as a computer scientist doing signal analysis algorithms, one way I would make just such a compromise is by applying sliding window techniques. As the name suggests, this method is like cutting a hole into a piece of paper, and laying that over a chart. You can only see the data in the “window.” So, you do your analysis on that window of data, and then slide the window over a little and repeat your analysis. A moving average is an example of a sliding window calculation, as are many other standard indicators. If it’s not clear yet, hopefully the next paragraph will shed some light.
The most important thing about sliding window techniques is that the windows overlap. Why is this important? Well, look at the 30 minute chart above. That’s an example of non-overlapping 30 minute windows. All we’ve got is 10:00 to 10:30, and then 10:30 to 11:00, and some useful data is invisible to us. But, what if we had all the overlapping candlesticks for 10:01 to 10:31, and 10:02 to 10:32, and so on? If we did, a few of those candles would show us the consolidation area we can spot in the 1-minute charts, even though we’re still always looking at 30-minutes of data at a time. That is how sliding windows bridge the gap between summary data and noisy precision.
Finishing the Example
So, if we’re looking for 30 minutes of narrow-range activity, we could just chart the 30-period high and 30-period low from our 1-minute data. This is an example of a sliding window calculation, since the first data point for the 30-period high represents the high from 10:00 to 10:30, and the second point is the high from 10:01 to 10:31, and so on.
A chart of the 30-period high/low data is below. Notice that we aren’t superimposing the data on top of the one-minute candles themselves–that’d reintroduce the noise and confusion that we’re trying to avoid. We’ll just look at the slower, derived data:
You can see, it’s a little noisier than the 30-minute bars, but nowhere near as distracting as 1-minute candles. That’s our compromise, and what we get for it is the ability to see that period where the range contracts. Now we can spot any half-hour period of consolidation, and can take our entry.
What, no Bars? No Candles?
A lot of us take for granted that when we’re trading we should look at the open-high-low-close price data, usually in candle or bar form. But, why do we think this? Candles and bars are just summarizing indicators, no different than moving averages or bollinger bands. We tend to think of them as “raw” market data, but one look at the 30-minute chart above shows that this is clearly not true.
So, what should we be looking at? I’d say you should consider devising a set of indicators that are designed to show you exactly what your system looks for, and little else. If you need to see a half hour of narrow range, the price channels above are a good way to go (for extra credit, plot them as % of the price, so that a 2%-wide channel looks identical for every stock you watch). If you want to know the direction of the overall trend, put a single +, -, or 0 on your screen that corresponds to slope of an appropriate moving average.
I realize some ideas like this would require custom programming, but the closer you can get to these ideals, the better. How many good trades have I passed up because I noticed something random on my charts that made me hesitate? How many bad trades have I gotten into because I noticed something random on my charts that I liked? All that extraneous information just begs a trader to step outside their system. Try to get rid of it!
Summary
We all have a goal of eliminating distracting and unwanted information from our screens. Fast charts can make us overtrade, or choose entries that leave us chopped up. As shown here, though, slow charts come with their own set of problems. By using summarizing indicators, and sliding window techniques, you can eliminate fast chart noise while retaining fast chart precision. Think about what conditions trigger your trading system, and then figure out which indicators will point out these conditions in the most obvious and precise way. Consider leaving bars and candles off your screen unless you really need to see the data they’re presenting. If you do watch intra-day candles, keep in mind that a small time-shift of the time-and-sales can drastically change the look of the chart even though the price action is identical.
August 9th, 2006 at 12:24 pm
Richard,
Nice article. I totally agree with your analysis. I was thinking about something similar in the past few days. I noitced that OBV is quite different when looking at different intraday timeframes. While smaller the timeframe more accurate the results are, it does come with some degree of noice. Your article gives me another line of thought. Thanks for your blog.
Srini
August 9th, 2006 at 12:50 pm
@Srini: thanks! For all indicators, really, it boils down to three choices: You can use fine-grained data and accept the noisiness of it. You can eliminate some of the noise by using coarse-grained data, and accept the inaccuracies of it. Or, you can try to smooth out the noisy data some other way.
This article attempts to give a couple examples of what “some other way” might be. For fidgety OBV, you could try to watch an exponential moving average of the OBV line, for instance. There are a number of ways to smooth out noisy signals.
August 9th, 2006 at 2:25 pm
good point - I trade 30-min candles, but I will often look at 15-min candle charts for two narrow candles next to each other (e.g. 10:45 & 11) that would not show up on the 30-min candles.
Another thing to note is that round number times seem to be important. I don’t have proof, but psychologically people are drawn to round number times (10,11,12:30). Do you ever ask someone to meet you at 12:17 for lunch? Because the market is just made of the people in it, it seems that the market would also be drawn to these round number times. Anyway, it is something to think about.
August 9th, 2006 at 2:27 pm
Yeah, I meant to mention that at some point in the post. Certainly in terms of volume, there are little pops at round number times.
August 10th, 2006 at 12:31 pm
Richard,
Very useful and well thought-out article. I did notice something familiar about your modified chart–
I mostly daytrade, but a little while back I was doing a trade on weekly bars. I got very frustrated with having to wait for the close on Fridays to have a complete bar to work with. Also I discovered, as you have, that important events take place within longer bars and are often obscured.
Long story short, I began using rolling 5-day highs and lows instead, and VOILA! Better information while still using the longer timeframe. It was only afterwards that I realized I had just re-created a version of Richard Donchian’s work.
Rather than having to keep running calculations on intraday bars in real-time (stressss!), or program a feed into Excel to do it (inconvenient, to say the least), perhaps consider applying 3-period Donchian channels to 10-minute bars (or 6-period channels on 5-minute bars) and see if that won’t give you similar improvements over straight 30-minute bars as the method you have discussed.
Great work, and best of luck w/ your trading.
August 10th, 2006 at 2:59 pm
@Will: Thanks, I’m glad you liked the article. And, I’m glad to hear you’ve arrived at similar conclusions in your own experience.