The indicator did not start this morning

The stock closed at ₹640 yesterday and opens at ₹648 this morning. At 9.20 the first five-minute candle completes, and on the chart it has closed above the 20-period exponential moving average, which is sitting at ₹641.20. That is a crossover, the rule says take crossovers, and the entry goes in at 9.21. By 9.55 the position is stopped out on a stop sized from the fourteen-period ATR. Reviewing it in the evening, two things are worth noticing about the two numbers that produced that trade, and neither of them is visible on the chart. The average that was crossed was ninety per cent yesterday. And the volatility figure that set the stop was, at that moment, mostly a measurement of a gap that happened while the market was shut.

The previous lesson was about where the bars get cut. This one is about what happens at the join. The chart draws no line at 9.15. A running indicator does not know that a session ended and another began; it simply receives the next bar and updates.

Think of it like this

The pot that never gets washed

A dhaba keeps one pot of gravy going. Each morning the cook adds today’s onions and tomatoes to what is already in it. At eight in the morning the pot is yesterday’s gravy with a spoon of today’s in it, and by evening it is mostly today. Nothing has gone wrong; that is how the dish is made. It only becomes a problem when somebody tastes it at eight and reports it as today’s cooking.

In the market

A running average is that pot. At 9.20 the pot contains one five-minute bar of today and nineteen of yesterday, and the indicator reports a single number with no note attached about the proportions. By about half past ten the mixture has genuinely turned over. What people do is taste it at 9.20.

Which readings restart at the bell, and which do not

Two families of number on the same screen

Session-anchored — these genuinely start at 9.15

The day’s high and the day’s low
The day’s traded volume
The session [[VWAP]], which by definition begins accumulating at the open
The [[Opening range]], and anything measured from the open
The percentage change shown against the previous close

Running — these carry yesterday across the join

Every moving average, simple or exponential
RSI, MACD, the stochastic and every oscillator built on them
ATR and every stop sized from it
Bollinger Bands, which are an average plus a running deviation
Donchian channels and any lookback expressed as “the last N bars”

Putting a number on how much of today is in there

An [[Exponential moving average]] gives the newest bar a fixed weight and everything before it the remainder, so the arithmetic is straightforward.

α = 2 ÷ (N + 1), and the share of the value coming from the last k bars = 1 − (1 − α)^k

N: The period setting on the indicator — 20, in the example above
α: The weight given to the newest bar. For N = 20 it is 2 ÷ 21, about 9.5%
k: How many bars of today have closed so far

Example: One bar into the session, 9.5% of the average is today and 90.5% is yesterday. Nothing about the chart says so.

Worked example

How long before a 20-period EMA is mostly today?

A 20-period EMA on a five-minute chart, where a session holds 75 bars

At 9.20 — one bar closedThe crossover in the opening paragraph happened against a number that was nine-tenths yesterday afternoon9.5% of the value is today

At 9.40 — five bars closed1 − (1 − 0.095)⁵. Still a minorityAbout 39%

At about 10.15 — twelve bars closedAn hour of trading is what it takes for the average to be predominantly about todayAbout 70%

At about 10.55 — twenty bars closedAnd the remaining 13.5% is still yesterday, because an exponential average never fully forgets anythingAbout 86.5%

The same setting on a daily chartThere the newest 9.5% is a complete day of collective behaviour, which is why nobody ever notices this problem on daily barsOne bar is one whole session

And a simple average behaves differently againA 20-period simple average is exactly 100% today from the twentieth bar of the session onward, and exactly 0% today before the first — a cliff rather than a fade. Neither is better; they are different failures at the same joinIt drops the oldest bar completely

The number on the chart at 9.20 is not wrong, and no platform is misreporting anything. It is answering a question — what has the average of the last twenty five-minute bars been — that simply is not the question the trader thought he was asking, which was whether today has turned. The gap between those two questions is widest in the first half hour of every session and closes by about half past ten, which is precisely the window in which most intraday signals are taken.

Loading interactive demo…

Move the period and watch how much of the line is being set by old data. The same effect, at the session join, is what this lesson is about.

The gap that lands inside one bar

The second half of the problem is worse, because it concentrates rather than fades. True range is defined so that an overnight move is not lost: it is the largest of the bar’s own high-to-low range, the distance from the high to the previous close, and the distance from the low to the previous close. On a daily chart that is exactly right and it is why the measure exists. On an intraday chart it means the whole overnight gap is deposited into the true range of one five-minute bar.

Worked example

One bar carrying eight rupees of a gap

The share above: previous close ₹640, opens at ₹648

The 9.15 – 9.20 candle itselfIts own range is ₹3, which is an ordinary opening five minutes for this stockHigh ₹649, low ₹646

Its true rangeMeasured from the previous close, because that is what the definition says. Three times the bar’s own range₹649 − ₹640 = ₹9

What a 14-period ATR on this chart normally spansFourteen five-minute bars. On a daily chart the same setting spans about three weeksAbout seventy minutes of trading

So at 9.25, the ATR window isA single bar carrying an overnight move now dominates a window of quiet afternoon barsTwo bars of today and twelve of yesterday afternoon — with the ₹9 bar among them

And with Wilder’s smoothing it does not drop out cleanlyWilder’s ATR is a running average with a weight of 1 ÷ 14 on each new bar, so roughly a third of the reading is always older than the stated period. The gap bar fades; it never leavesOnly about 65% of the value comes from the last fourteen bars

What that does to a stop sized as “1.5 × ATR”It is too wide for the morning’s actual five-minute movement, and it was computed from a number the trader believes describes intraday volatilityA stop set by an event that happened while the market was shut

The same brick sits inside anything computed across the join. An RSI on five-minute bars takes the change from yesterday’s last close to this morning’s first close as one observation, so a 1.25% overnight gap enters the calculation as though it were a single five-minute move — and on a fourteen-period Wilder [[Smoothing|smoothing]] it will still be influencing the reading well after lunch. None of this is an argument that the gap is unreal. It is an argument that a gap is a daily-chart event, and putting it inside an intraday window measures it on the wrong scale.

Three honest responses

1
Take nothing from a running indicator until it is mostly today
The simplest fix and the one that costs nothing. For a period of N bars, roughly N bars of the session have to close before an exponential average has shed most of yesterday. A twenty-period setting on five-minute bars means waiting until the twentieth bar closes, at about five to eleven — which happens to remove the most expensive part of the day as a side effect.
2
Or restart the indicator at the session open, and accept what that costs
Some platforms and most self-built systems can compute an indicator from 9.15 only. It removes yesterday entirely — and it means the first N bars of every session have no valid reading at all, and that the indicator is a different one from the one everyone else on the stock is watching. That is a real trade-off, not a free improvement.
3
Or use the readings that are session-anchored by construction
Session VWAP, the opening range, the day’s high and low and the distance from the previous close all begin at the bell. If you want a morning reference that is genuinely about this morning, it should come from that column rather than from a running average with a period setting.
4
Size stops from a session-aware volatility figure
If a stop is meant to survive ordinary intraday movement, measure ordinary intraday movement: the average of the bars’ own high-to-low ranges within the session, rather than a true range that includes the overnight move. Say which of the two you used, because they are different numbers with similar names.
5
Check what your platform already does
Some feeds compute intraday indicators continuously across days, some restart them, and a few offer the choice. Whichever yours does is now part of your rule, so find out and write it down next to the anchor from the previous lesson.

◆ Your call

A crossover at 9.20 on a stock that gapped up 1.5%

The 20-period EMA on the five-minute chart sits below the first candle’s close. Your written system is a moving-average crossover with an ATR-based stop, tested on this stock over two years of five-minute data.

Check yourself

On a five-minute chart, which of these readings genuinely contains no data from yesterday at 9.20 this morning?

Simple bhasha mein

Nal ka pehla mug kal ka paani hai

Subah geyser chalu karke turant nal kholo toh pehla mug thanda aata hai — woh raat bhar pipe mein pada paani hai, geyser ka aaj ka kaam nahi. Do minute baad paani aaj ka ho jaata hai. Chart pe 9.20 baje ka average bilkul wahi pehla mug hai. 20-period EMA mein sabse nayi bar ka wazan 2 ÷ 21 hota hai, yaani 9.5% — matlab 90.5% abhi kal shaam ka hai. Paanch bar baad (9.40) sirf 39%, baarah bar baad (10.15) lagbhag 70%, aur bees bar baad (10.55) 86.5%. Doosri taraf ATR ulta dhokha deta hai: share kal ₹640 pe band hua aur aaj ₹648 pe khula, toh 9.15–9.20 waali candle ki apni range sirf ₹3 thi (649 se 646), par uska true range pichhle close se naapa jaata hai — ₹649 − ₹640 = ₹9, teen guna. Yaani subah aapko trend ka number milta hai jo kal ka hai, aur volatility ka number jo raat ki chhalaang ka hai. Session VWAP, din ka high-low aur opening range — yeh teen hi ghanti se shuru hote hain.

What to remember

A chart draws no line at 9.15, and a running indicator does not know a new session has begun.
For an exponential average, one bar contributes 2 ÷ (N + 1) — so a 20-period EMA is 9.5% today at 9.20 and only predominantly today after about an hour.
True range measures the gap against the previous close, so the entire overnight move lands inside the first bar of the day.
Wilder’s smoothing keeps roughly a third of an ATR reading older than the stated period, so a gap bar fades but never leaves.
Session VWAP, the day range and the opening range are the only common readings that genuinely start at the bell.

Finished this lesson?

Mark it done to track your progress through the curriculum.