EMA cross with CPR: what the two overlays actually tell you

Two overlays ship with the chart in this app, and between them they are most of what gets called a "system" in Indian intraday trading: an 8/20 exponential moving average cross, and the Central Pivot Range. Neither is complicated. Both are routinely traded by people who have never looked at what the numbers are made of.

This post looks. There is one genuinely useful thing hiding in the CPR arithmetic, and it is not what the popular version of the setup says.

The 8/20 cross

An exponential moving average is the previous average pulled part of the way toward the latest close, repeatedly. The 8-period line reacts faster than the 20-period line, so when price turns hard enough for long enough, the fast line crosses the slow one, and that crossing is the signal.

The important property is structural rather than tunable: an average of past closes cannot lead price. By the time two of them cross, the move that crossed them has already happened. That is not a flaw to be optimised away with different periods — it is what a moving average is.

The chart marks each cross with a triangle labelled 8/20. An exact equality between the two lines is treated as a touch and not a cross, so a series that grazes and falls back emits no signal at all. The lines also start short of the left edge, because the first few bars have no average yet, and saying so is different from drawing a zero there.

CPR, and the formula worth reading

The Central Pivot Range is three levels for today, computed from yesterday's high, low and close:

pivot = (high + low + close) / 3
bc    = (high + low) / 2
tc    = 2 * pivot - bc

Three lines, and that is deliberately all — the R1/R2/R3 and S1/S2/S3 family is a different indicator, and nine lines over a candle chart is how a chart stops being read.

One practical note before the interesting part: tc is only the top line on a day whose close sat above the mid-point of its range. On the other half of all days the formula puts it underneath bc. Trusting the names draws the top line below the bottom one, so the levels here are returned ordered instead of named after the formula.

What CPR width actually measures

The whole popular setup rests on CPR width — a narrow range meaning one thing, a wide one meaning another. So work out what the width is:

|tc - bc| = |(high + low + close)/3 - (high + low)/2|
          = |2*close - (high + low)| / 6

Read that carefully. The width depends on the close and the mid-point of yesterday's range, and on nothing else. It is a measure of where yesterday closed relative to its own middle. It is not a measure of how much yesterday moved.

Those come apart constantly:

• A day that swung 400 points and closed dead in the middle of that swing produces a narrow CPR.
• A quiet day that drifted 60 points and closed right on its high produces a wider one.

So "narrow CPR" does not mean "yesterday was quiet". It means "yesterday ended undecided" — buyers and sellers finished the session in balance. And the popular rule, narrow CPR predicts a trending day, is therefore a specific and checkable claim about behaviour: that a session which closes in balance tends to be followed by one that picks a direction and goes.

That is a real hypothesis rather than folklore, which is worth knowing. It is also not the same thing as being true, and there is no published evidence here to hand you. What you can do is check it, which the last section is about.

The setup, stated once

In the direction it is normally traded:

A narrow CPR, with price establishing and holding outside the range, and the 8 crossing the 20 in that same direction on the 5-minute chart — that is the trade. The pivot is the level you are wrong below on a long, and wrong above on a short.

A wide CPR, or price still oscillating inside the range, is the session where the same cross is most likely to be noise. The filter is doing more work than the trigger, which is the opposite of how most people describe it.

Why options punish a lagging trigger harder than the index does

Being late to a move is survivable when you are trading the thing that moved. It is much less survivable through an option.

The cross puts you in after the move that caused it, so you have paid a worse entry price for the same thesis. Meanwhile the premium has been losing time value the entire time you waited for confirmation, and it keeps losing it while you hold. A directional call that is right but slow is a familiar way to be right and still down — the index went nowhere and your option lost money works through why in detail.

Confirmation is not free. On an at-the-money weekly, it is one of the more expensive things you can buy.

Draw the overlays on the right instrument

This is the mistake most specific to using these two indicators for options, and it is easy to make because the overlays will happily compute on whatever series is on the chart.

CPR on an option chart is close to meaningless. The levels would come from that strike's own previous session — a contract whose moneyness, remaining time and expected volatility have all moved since that close was printed. The arithmetic runs and produces three tidy lines. They describe nothing you can trade against.

An 8/20 cross on a premium series is two signals wearing one line. The premium moved because the index moved, and because the day decayed. A cross of averages over that series cannot tell you which.

So the split is: read the signal on the spot chart — NIFTY50, NIFTYBANK, SENSEX — and express it in a contract, marked against that contract's own traded price.

The code already draws this line, as it happens. An index is not tradable at all here; ask for one and it says so, and tells you to trade it through its option contracts. And charting a watchlist row is a separate action from trading it, so having the index on the chart while a strike is in the ticket is the normal case rather than a workaround.

What the chop costs, in rupees

Two moving averages in a rangebound session cross repeatedly. Each cross taken is a round trip, and on Indian F&O a round trip is at least ₹40 of brokerage — ₹20 per executed order, flat, before a single percentage line has been added. More if the order exceeds the exchange freeze quantity for its underlying, because then it reaches the market as two orders and the contract note carries two ₹20 lines per leg.

Six false crosses in one quiet session is ₹240 of entirely certain cost set against an entirely uncertain edge. That is the argument for CPR in this pairing, and it is worth being precise about it: CPR earns its place as a filter on the trigger — a reason to take no trade today — and not as a second trigger that finds you more of them. What a round trip actually costs has the full breakdown.

How to check any of this

Both overlays are off by default and persist per user once you turn them on, so the desk comes back the way you left it.

Look before you trade it. Chart an index at the 15-minute resolution, switch both overlays on, and pan back. Every session drawn carries the pivots that were genuinely real while it traded — yesterday's levels on yesterday's chart, not today's stamped across history — so you can scroll through weeks of sessions and read, for yourself, whether the narrow-CPR days actually trended and whether the crosses on those days went anywhere. That takes an evening and it is the cheapest test available.

Then trade it, with a stop. The ticket pre-fills no stop and no target on purpose; a level sitting in a form field reads as advice, and an empty ticket will place a position with no automatic exit at all. Set one you chose.

Everything defaults to intraday and the 15:10 square-off closes it, so the setup gets measured over a session, which is the horizon it is for.

Judge it on net P&L across thirty trades, not the six you remember. Statistics here compute on net, after charges, and a trade that finished green by less than its own round-trip cost is recorded as a scratch rather than a win — because it was one.

The trigger is not the edge. Two averages crossing is public information available to everybody, instantly, for free. Whatever edge exists in this lives in the days you decline to take the signal, and in whether you kept a record honest enough to tell the difference.