Standard deviation, read on a chart.
Standard deviation tells you how spread out the selected prices are around their rolling average. A low reading means the values are clustered; a high reading means they are dispersed. It does not tell you whether the next move is up or down.
All prices and outcomes on this page are fictional teaching examples, not performance claims.
One number. One question.
Standard deviation is a rolling description of dispersion. It compares every selected observation with the window's average, makes negative and positive distances comparable by squaring them, averages those squares, then takes the square root to return to price units.
- 1Find the mean
The five closes average 1.1000.
- 2Measure each distance
0, +2, -2, +4 and -4 pips from the mean.
- 3Square and average
(0 + 4 + 4 + 16 + 16) / 5 = 8.
- 4Take the square root
Population standard deviation = 2.8 pips.
sqrt( average( (price - mean)^2 ) )Quiet versus active windows, expansion versus compression, and relative distance when used inside a defined method.
Direction, support, resistance, a reversal point, a stop location or whether returns follow a normal distribution.
A shock can keep the line high until that observation leaves the lookback, even if the latest candles become calm.
Four states, zero directional votes.
Read level and slope separately. The exact number is meaningful only for the same symbol, input, timeframe and settings; use a fixed rolling reference rather than calling a line "high" by eye.
Low + flat
Recent values are close to their mean. A breakout may or may not follow; low dispersion alone supplies no trigger.
Expanding
Dispersion is increasing. Use price structure or a trend rule to determine whether that expansion supports a long, a short or no trade.
High + holding
Values remain widely dispersed. This can be a fast trend or unstable whipsaw; "high" is not synonymous with overbought or due to reverse.
Compressing
Dispersion is declining. Compression can continue for many candles, so wait for a separately defined expansion trigger.
See what makes standard deviation move.
Switch the price behaviour, then change the rolling window. The shaded area is the latest mean plus or minus one population standard deviation.
- Rolling mean
- 1.10260
- Standard deviation
- 1.2 pips
- Window range
- 4.0 pips
The closes are bunched together, so dispersion is low. That describes the recent movement; it does not promise that a breakout is next.
Use dispersion to describe the volatility regime. Add structure or a directional rule for the trade idea, and size the stop from price rather than treating one or two standard deviations as guaranteed turning points.
Turn a reading into testable rules.
These are exact starting hypotheses for demo research, not claims of an edge. Both use a $10,000 USD demo account, a $25 all-in loss cap per trade, completed H1 candles and the stated cost allowance. Keep every rule unchanged for the sample and retain failed and skipped setups.
Use standard deviation as an activity filter; let EMA slope and price location supply direction.
Trend pullback + SD expansion
- Chart
- EUR/USD H1 candles; EMA(50) of close; Standard Deviation(20), SMA method, close, shift 0. In this method, one pip is 0.0001. Evaluate only after an H1 candle closes.
- Context
- Long only when close is above EMA(50) and EMA(50) is above its value five completed bars ago. Short is the exact inverse.
- Pullback
- The first later candle whose low touches or crosses EMA(50) starts a long pullback; for a short, use the first candle whose high touches or crosses it. The signal must occur on that candle or one of the next two. The swing low is the lowest low from the first pullback candle through the long signal; the swing high is the inverse for a short.
- SD filter
- Signal-bar SD(20) is above the median of the previous 20 completed readings, excluding the signal. Sort those values and average the 10th and 11th. The signal value must also be greater than the prior bar's value, which must be greater than the value two bars ago.
- Long / short trigger
- Long: a completed candle closes back above EMA(50); place a buy stop exactly one pip above its high. Short: close back below; sell stop exactly one pip below its low. Cancel after three completed bars.
- Stop / invalidation
- Long stop exactly one pip below the defined pullback swing low; short stop exactly one pip above its swing high. Cancel before entry if a candle closes through that stop price.
- Risk / size
- Account $10,000; maximum all-in loss $25. Reserve $1 for a fictional round-turn commission and $1 for adverse slippage. Divide the remaining $23 by stop distance in pips and the symbol's pip value per lot, then round volume down to the permitted step. The example specification is $10 per pip at 1.00 lot with a 0.01-lot step. Entry and exit prices already include spread.
- Exit
- Set the target two initial stop distances from entry. Exit there or on the first completed close across EMA(50) against the trade, whichever occurs first. Keep the initial stop. Report net R using the actual all-in stop loss after volume rounding as 1R.
- No trade
- Skip if the signal candle's range exceeds 1.5 x ATR(14), spread is above 1.2 pips, or a chosen high-impact event window is active. Before the session, mark every confirmed H1 pivot in the previous 100 bars. A pivot high has a higher high than the two completed bars on each side; a pivot low is the inverse. At the signal, use the nearest marked pivot above a long entry or below a short entry; skip if it is closer than the two-stop-distance target.
The EMA decides the allowed side. The SD condition says activity is expanding; it does not create the long direction. A stopped trade is $22 price loss plus the $2 allowance = $24, or -1R, inside the $25 cap. The assumed 1.0-pip spread is already in the executable fictional prices; overnight financing is excluded.
Define a price box first, then require dispersion to leave a measured low-volatility state.
Compression-to-expansion breakout
- Chart
- GBP/USD H1 candles; Standard Deviation(20), SMA method, close, shift 0. In this method, one pip is 0.0001. Set the box before the five-bar compression check: its high and low are the highest high and lowest low of bars 6 through 25 before the signal candidate.
- Compression context
- For each of the five bars before the signal candidate, SD(20) was below the 25th percentile of the 100 SD readings immediately before that bar. Sort those values low to high and use the 25th value; the tested bar is excluded. Each of those five closes must also stay inside the same fixed box.
- Expansion filter
- On the signal close, SD(20) crosses above the median of its previous 20 completed readings, excluding the signal value, and is higher than on each of the prior two bars. For 20 values, the median is the average of the 10th and 11th values after sorting low to high.
- Long / short trigger
- Long signal closes above the box high; buy stop exactly one pip above its high. Short signal closes below the box low; sell stop exactly one pip below its low. Cancel after two completed bars.
- Stop / invalidation
- Stop exactly one pip beyond the opposite end of the signal candle. Cancel before entry if a completed close returns inside the old box.
- Risk / size
- Account $10,000; maximum all-in loss $25. Reserve $1 for a fictional round-turn commission and $1 for adverse slippage. Divide the remaining $23 by stop distance in pips and pip value per lot, then round volume down to the permitted step. The example specification is $10 per pip at 1.00 lot with a 0.01-lot step. Entry and exit prices already include spread.
- Exit
- Set the target two initial stop distances from entry. Exit there or on the first completed close back inside the old box, whichever occurs first. Keep the original stop and report net R using the actual all-in stop loss after volume rounding as 1R.
- No trade
- Skip if the signal range exceeds twice the median range of the prior 20 bars, the box width exceeds 1.5 x ATR(14), spread is above 1.5 pips, or the data window has missing bars. Before the session, mark every confirmed H1 pivot in the prior 100 bars using Method 01's two-bars-on-each-side rule. At the signal, use the nearest pivot above a long entry or below a short entry; skip if it is closer than the two-stop-distance target.
The box break supplies short direction. SD rising from 3.1 to 5.7 pips meets the expansion filter; it does not guarantee follow-through. A stopped trade is $22 price loss plus the $2 allowance = $24, or -1R, inside the $25 cap. The assumed 1.2-pip spread is already in the executable fictional prices; overnight financing is excluded.
Expansion can be the wrong kind of movement.
Keep these cases in the sample. A filter is useful only if its exclusions and losses are recorded as carefully as its clean examples.
SD expands, but the signal candle is larger than 1.5 x ATR and the next bar erases it. Method 01's no-trade rule rejects the setup before entry.
A wick clears the box, but there is no qualifying close and no SD cross above its reference. Method 02 records no trade, not a missed winner or an invented late entry.
Match the inputs, then compare.
Standard Deviation is built into MT4, MT5 and TradingView. Keep the symbol feed, timeframe and completed-candle timing in your notes; identical labels do not guarantee identical values.
Insert → Indicators → Trend → Standard Deviation
- Open the intended symbol and timeframe.
- Set Period, Shift, MA method and Apply to.
- Use the Data Window to inspect completed-bar values.
Insert → Indicators → Trend → Standard Deviation
- Choose the symbol and chart timeframe first.
- Set period, shift, averaging method and applied price.
- Confirm the separate-window line after the bar closes.
Indicators → Technicals → Standard Deviation
- Choose the exact symbol feed and interval.
- Open Settings → Inputs; record Length and Source.
- Keep the built-in study distinct from community scripts.
No universal best value
Treat 14, 20 and 50 as a small comparison grid, not as three recommendations. Shorter windows react faster and turn over observations sooner; longer windows respond more slowly and blend more regimes.
- Period / length
- 14 faster · 20 middle comparison · 50 slower
- Applied price / source
- Close is a clear baseline; Open, High, Low, HL2, HLC3 and HLCC4 answer different questions.
- MA method
- SMA is the plain baseline used in this page's demos. EMA or other methods change the centre and output.
- Shift
- Keep shift 0 for ordinary research unless displacement itself is part of the written hypothesis.
- Convention
- Population divides by N; sample estimates often divide by N - 1. Verify the implementation before comparing tools.
Add missing information, not another vote.
A useful stack has one explicit job per tool. Correlated transforms of the same closes are not separate confirmation just because they have different names.
Moving average slope
Answers which side is allowed. SD still answers only whether dispersion is expanding or contracting.
Watch the overlap: SD may already be calculated around a moving average of the same closes.Price structure
Pre-marked swings or a fixed range supply boundaries, invalidation and room-to-target.
This adds chart location instead of another transformation of volatility.ATR — with one job
ATR can flag an oversized candle or scale a buffer, but it is another volatility measure.
Do not award one vote to rising SD and a second vote to rising ATR unless the rules give them distinct jobs.Bollinger Bands / z-score
Both directly reuse standard deviation: bands scale envelopes; z-score divides distance by dispersion.
Band widening plus SD expansion is one mathematical family, not two confirmations.From chart idea to review.
Recreate one method, freeze the card, replay unseen candles, size the fictional risk and record every qualified setup.
Standard deviation FAQ
The shortest answers to the questions most likely to distort a chart test.
01Does rising standard deviation mean price will rise?+
No. Standard deviation measures how widely the selected values are spread around their average. A sharp rise can accompany an up move, a down move or two-sided whipsaw.
02What is the best standard deviation setting for forex?+
There is no universal best period, average method or price input. A useful research process compares a small fixed set, such as 14, 20 and 50 completed closes, then evaluates the same rule across unseen charts and realistic costs.
03Is standard deviation the same as ATR?+
No. Standard deviation measures dispersion around an average of the selected input. ATR averages true range and explicitly includes gaps through its true-range calculation. Both describe volatility, so assigning each a separate confirmation vote can double-count the same market property.
04Does very high standard deviation predict a reversal?+
No. High standard deviation says recent observations are widely dispersed. It does not identify a ceiling, floor or direction, and the reading can stay high while a trend continues.
05Why do MT4, MT5 and TradingView values differ?+
Check the symbol feed, candle boundaries, period, applied price or source, moving-average method, shift and calculation convention. A small difference is not automatically an error; document the exact platform and inputs used for a test.
06How is standard deviation related to Bollinger Bands?+
Bollinger Bands place envelopes a chosen number of standard deviations around a moving average. Band widening and a rising standard-deviation line are therefore mathematically related, not two independent confirmations.