Indicators & Oscillators
Moving averages and bands
A moving average is the least clever indicator ever invented: add up the last N closes and divide by N. It has no predictive theory behind it and it is always late. It is also, by a wide margin, the most useful line most traders will ever draw — because its one job is to answer a question you cannot answer by staring: what is the average price the crowd has been paying lately, and is that number rising or falling?
Every complaint about moving averages reduces to "it turns after the price does". Of course it does — it is an average of the past, and that is precisely what makes it useful. Raw price is mostly noise: a 3% down day inside an uptrend contains almost no information, and reacting to it costs you the trend. The average deliberately throws that away. You are trading the smoothing, not despite it. The only real design decision is HOW MUCH you throw away, and that is what the period controls: short periods hug price and generate signals constantly, long periods ignore almost everything and turn late but rarely lie about direction.
- As a DIRECTION filter: the slope of a longer average is your permission slip. Rising means you may take long setups only; falling means shorts only. This single rule removes most of the trades that lose money.
- As a DYNAMIC support/resistance: in a healthy trend price pulls back to the average and resumes. That gives you a place to enter with a nearby stop instead of chasing the breakout.
- As a CROSSOVER system: a fast average crossing a slow one (the "golden cross" up, "death cross" down). Simple, mechanical, and only profitable in markets that actually trend.
- Whichever you use, apply it to the CLOSE and act on closes. Intrabar crosses un-cross constantly, and a system that trades them trades noise at full commission.
A simple moving average weights every bar in the window equally, which means a huge move drops out abruptly the day it leaves the window — producing a turn in your indicator caused by old data disappearing rather than by anything happening now. An exponential average weights recent bars more and fades the old ones smoothly, so it responds faster and has no drop-out artefact. Elder prefers the EMA for that reason. In live testing the difference is real but small; it is dwarfed by the choice of period and utterly dwarfed by whether the market is trending at all. Pick one, write it down, and stop revisiting it.
Take a 20-period average and draw two lines around it at two standard deviations of the last 20 closes. Because the deviation is computed from recent data, the bands automatically widen in violent markets and contract in quiet ones — they measure the price's current normal range rather than a fixed percentage. Two readings follow directly. A SQUEEZE, when the bands narrow to an unusual degree, says volatility has collapsed and historically tends to precede an expansion, though it says nothing at all about direction. And in a strong trend price WALKS the upper band for many bars — which is the exact situation in which "price touched the upper band, so sell" bankrupts people.
It is trivially easy to test 200 combinations of fast and slow periods and find the one that would have been best last year. It is also close to worthless: you have found the setting that best fit last year's specific sequence of noise, and there is no reason for that sequence to repeat. The classic values — 50 and 200 for position trading, 13 and 26 on shorter frames — have no magic in them either, but they have one real advantage: enough traders watch them that they carry some self-fulfilling weight, and you did not obtain them by torturing your data.
This is not a minor caveat, it is the main event. A moving average system in a trending market catches the move and pays for a year of small losses. The same system in a sideways market generates a cross every few bars, each one entering near the top of the range and exiting near the bottom, and the account bleeds out in commissions and small stops. Since markets range more often than they trend, an unfiltered crossover system loses money on most instruments most of the time. The lab above lets you watch this happen; the fix is module 5's, and it is to ask a longer timeframe for permission before taking any signal.
A moving average does not predict; it summarises. Its slope is the most honest one-word answer to "which way is this market going", and its distance from price is a decent proxy for how stretched the move has become. Everything else people build on it — crossovers, ribbons, envelopes, bands — is a way of turning those two facts into a decision. When a moving average system fails, it is almost never because the average was wrong about the average. It is because there was no trend to summarise.
In the lab, switching from Uptrend to Range with the same 10/30 pair multiplies the number of crossover signals and most of them are whipsaws. Explain why in terms of what the two averages are measuring. (In a trend the two averages measure the same rising quantity at different lags, so they only cross when the trend genuinely turns. In a range both are measuring the same flat mean plus noise, so their difference oscillates around zero and crosses it constantly. The signals multiply precisely because there is no longer any real difference between the two numbers.)
Use the slope of a longer moving average as a permission filter before anything else — it is the cheapest edge in this course. Treat the average as dynamic support in a trend and enter on pullbacks to it rather than chasing. Prefer the EMA if you want fewer artefacts, but do not spend a week on the choice. Read Bollinger squeezes as volatility news with no directional content, and remember that price walking a band is strength, not an invitation to fade it. And never run a bare crossover system without a higher-timeframe filter.
Put one long moving average on every chart you follow — 50 on a daily is a fine start — and add one rule to your written plan: "I take long setups only when the 50 is rising, short setups only when it is falling." Change nothing else about how you trade for a month, then compare that month to the previous one. Most traders find the filter removed more losers than winners, which is the whole reason it exists.
Indicators & Oscillators