Indicators lag because every one of them is calculated from prices that have already printed. An average of the last twenty bars cannot know about the twenty-first. Lag is not a defect in the settings or the platform — it is the arithmetic. The fix is not a faster indicator; it is using indicators for the job they can actually do.
This one sentence resolves most of the confusion around technical tools. People spend months hunting for the setting that is early instead of late, when the delay is baked into the definition of the calculation. Once you accept that, the useful question changes from "which indicator predicts?" to "what is an average of the recent past actually good for?"
The arithmetic, in one line
A simple moving average gives equal weight to each of its N bars. The centre of mass of that data sits (N − 1) ÷ 2 bars back. So a 20-period SMA is describing conditions centred roughly 9.5 bars ago; a 50-period is centred about 24.5 bars ago.
Put that in time and it becomes concrete:
| Setting | Approx. centre of data | On a 5-min chart | On a 1-hour chart |
|---|---|---|---|
| 9 SMA | 4 bars back | ~20 minutes | ~4 hours |
| 20 SMA | 9.5 bars back | ~48 minutes | ~9.5 hours |
| 50 SMA | 24.5 bars back | ~2 hours | ~1 day |
| 200 SMA | 99.5 bars back | ~8 hours | ~4 days |
An exponential moving average weights recent bars more heavily and therefore turns sooner, which is why EMAs are the default for intraday work. It still never reaches zero. Weighting the past differently is not the same as not using the past. Everything in the moving averages guide is subject to this, and so is every indicator built on top of one — MACD is a difference of two EMAs, and inherits the lag of both.
The trade you are actually making
Shortening the lookback does not remove lag. It exchanges lag for false signals, one for one. A 9 EMA turns faster than a 50 EMA and also turns on moves that go nowhere. There is no setting on the dial that gives you speed without noise, because both properties come from the same place: the indicator has no information other than what already happened.
So the question is not "which is better" but "which failure can I live with." A slower setting gives fewer, later, more reliable signals — you miss the first part of moves. A faster setting gives more, earlier, less reliable ones — you get chopped up in ranges. Both are survivable with proper risk sizing. Neither is survivable if you keep switching between them after every losing trade.
Why optimising the settings makes it worse
The natural next move is to test settings on historical data until you find the ones that would have worked. This is the single most seductive mistake in technical trading, and there is a well-known body of research about exactly it.
Sullivan, Timmermann and White assembled a universe of 7,846 technical trading rules — drawn from five common families including moving-average and filter rules — and applied them to 100 years of daily Dow Jones Industrial Average data, using a bootstrap "reality check" to measure how much of the best rule's apparent performance was simply the product of having searched through thousands of candidates (Sullivan, Timmermann & White, "Data-Snooping, Technical Trading Rule Performance, and the Bootstrap", The Journal of Finance 54(5), 1999). The point of the exercise was not that technical rules are worthless; it was that if you test enough of them, the best one will look excellent whether or not it has any edge at all.
That has a direct, practical consequence for anyone tuning a MACD. Trying twelve settings on your own chart history is a small version of the same search. The winner will look convincing for the same statistical reason, and it will be no more likely to work tomorrow. Time spent on settings is almost always time that would pay better spent on sizing from risk, which is not a search problem at all.
Four uses that survive the lag
- Trend context, not entries. "Price is above the 50 EMA on the hourly" is a fair description of the current regime and a bad reason to buy. Let the indicator tell you which direction you are willing to trade, and let a level tell you where.
- A dynamic level to react at. Widely-watched averages attract orders because many people are watching them, so price often reacts around them. Treat that reaction like any other level test — wait for the close.
- Divergence as a flag, not a trigger. Price making a higher high while momentum does not is worth noticing. It is a reason to tighten up or take partials, not a reason to short into strength.
- Cutting indicators down. Six indicators computed from one price series are one opinion stated six times. Two is usually plenty, and the redundancy you remove is confirmation bias you stop paying for.
Notice what all four have in common: the indicator supplies context, and price supplies the decision. That ordering is the whole answer to lag. You cannot make an average of the past arrive on time, but you can stop asking it to.
Frequently Asked Questions
How much does a moving average actually lag?
For a simple moving average, the centre of mass of the data sits at (N minus 1) divided by 2 bars back, so a 20-period SMA is describing conditions centred about 9.5 bars ago and a 50-period SMA about 24.5 bars ago. On a 5-minute chart that is roughly 48 minutes and just over two hours respectively. An exponential moving average weights recent bars more heavily and responds faster, but it never reaches zero delay, because it is still an average of prices that have already printed.
Can you remove indicator lag by using shorter settings?
No — you can only trade lag for noise. A shorter lookback reacts sooner, but it also reacts to moves that turn out to be nothing, so you replace late signals with frequent false ones. This is a genuine trade-off with no setting that escapes it, because both problems come from the same source: the indicator has no information except prices that already happened. The practical response is to stop trying to win the trade-off and instead choose which failure mode you can manage.
Are leading indicators actually leading?
Not in the sense the name implies. Oscillators such as RSI and stochastics are often called leading because they can reach an extreme before price turns, but they are still calculated entirely from past prices. What they lead is not the market — it is the moving average sitting next to them. An oscillator can flag that momentum is slowing, which is genuinely useful context, but it cannot tell you that a reversal has begun, which is why oscillator extremes stay extreme for long stretches during a strong trend.
Should beginners use indicators at all?
Use one or two, and use them for context rather than triggers. The failure mode for new traders is not using indicators — it is stacking six of them, all computed from the same price series, and mistaking their agreement for confirmation. Three moving averages agreeing is one piece of information reported three times. A single moving average for trend direction plus price levels for entries covers most of what a beginner needs, and it keeps the decision anchored to what price is doing now rather than to what an average of it was doing an hour ago.
Bottom line
Lag is the price of smoothing, and smoothing is the only thing an indicator does. A 20-period average is centred nearly ten bars in the past by definition, a shorter setting buys speed with noise, and hunting for the setting that would have worked is a search problem the data-snooping literature already answered. Use indicators to describe the regime, use price and levels to decide the trade, and put the effort you were spending on settings into risk instead. For the tools themselves, read moving averages explained and what MACD actually tells you; for the confirmation rule that replaces a crossover trigger, read break and hold; and for the wider picture, start with technical analysis basics.