Last Updated: 2026-10-26
An indicator can summarize a price series correctly and still produce a losing trading rule. Before choosing one, decide what you need to measure: trend, dispersion, recent momentum, traded volume or derivatives positioning. Three differently coloured versions of the same closing prices are not three independent reasons to trade.
The calculations below use invented bars, not a trading record. They make the inputs visible so you can check why two charts disagree. They do not rank indicators by profitability.
Match the tool to the question
Scroll sideways to see every column. Keyboard: focus the table region and use the arrow keys.
| Question | Useful reference | Boundary |
|---|---|---|
| How does recent price compare with its past? | SMA, EMA, HMA | Smoothing and lag depend on the formula and length |
| Is recent upward movement larger than downward movement? | RSI | A high reading is not a reversal deadline |
| Are two smoothed series separating? | MACD | It reuses price data; it is not an independent confirmation of an EMA |
| How large have recent ranges been? | ATR | Range has no bullish or bearish direction |
| How dispersed are recent closes? | Bollinger Bands | A band touch is not automatically an entry or a probability bound |
| Where did recorded volume occur? | VWAP, volume profile | Venue, session, price input and aggregation matter |
| How does volume accumulate under a close-direction rule? | OBV | It does not identify institutions or net buying by a group |
| What is happening in perpetual contracts? | Funding and open interest | Contract units, venue and interval must match; neither reveals every holder's intent |
Supertrend combines an ATR-based band with a switching rule. Ichimoku combines several price-derived lines, some plotted at shifted positions. Fibonacci retracements mark ratios between selected anchors. A footprint classifies trades by its chosen execution-side method. Each can organize a chart; none proves that a level must hold. For a proprietary "AI signal", require the input timing, version and out-of-sample evidence rather than treating the label as an explanation.
Reproduce a small calculation
Use these four synthetic closed bars, in chronological order. Prices and volumes use arbitrary consistent units; this is not a particular coin or exchange.
Scroll sideways to see every column. Keyboard: focus the table region and use the arrow keys.
| Bar | Close | High | Low | Volume |
|---|---|---|---|---|
| 1 | 100 | 101 | 99 | 10 |
| 2 | 102 | 103 | 99 | 20 |
| 3 | 101 | 102 | 100 | 30 |
| 4 | 104 | 105 | 100 | 40 |
SMA over the last three closes: (102 + 101 + 104) / 3 = 102.3333…. This discards bar 1. A three-bar EMA with weight 2 / (3 + 1) = 0.5, explicitly seeded at the first close of 100, instead progresses through 100, 101, 101 and 102.5. Another seed or longer warm-up history can change the EMA; this is not a claim that every platform must show 102.5.
A volume-weighted close proxy over all four bars: (100×10 + 102×20 + 101×30 + 104×40) / 100 = 102.3. This uses each bar's close as its representative price. It is not the exact transaction-level VWAP, because the trades inside each bar are absent. A platform using typical price, another session boundary or another venue can legitimately differ.
OBV, seeded at zero on bar 1: add 20 for the higher close, subtract 30 for the lower close, then add 40: 0, 20, -10, 30. A flat close would add zero under this convention. This classifies whole-bar volume using close direction. It does not show that institutions bought 30 units: every executed trade has a buyer and a seller.
Momentum and range use different information
The three close changes are +2, -1 and +3. For a deliberately short, initial three-change RSI calculation, average gains are 5/3, average losses 1/3, their ratio is 5 and RSI is 83.3333…. This is not a standard 14-period reading and does not say a reversal is due. TradingView's RSI documentation specifies the gain/loss smoothing; subsequent Wilder-style updates differ from repeatedly taking a new simple mean.
For bars 2–4, true range is the largest of high minus low, absolute high minus previous close, and absolute low minus previous close. The values are 4, 2 and 5. Their simple mean is 11/3, or about 3.6667 price units. This demonstrates the range calculation, not a fully warmed-up Wilder ATR. TradingView documents the formula and selectable smoothing. A multiple of ATR can define a planned distance; it does not cap losses when prices gap or execution fails.
An RSI and an ATR can therefore add different descriptions of the same bars. An RSI and a price-moving-average crossover share more of their input. Counting agreeing indicators does not establish independent evidence or a win probability.
Diagnose mismatches before changing settings
Save the symbol and venue, spot or perpetual contract, timezone, candle duration, closed-bar rule, price source, volume units, session reset, smoothing and warm-up. A current unfinished candle changes as trades arrive. Historical plots that use future-confirmed pivots or shifted lines must be aligned to the information actually available at decision time.
For example, an alert that uses a 15-minute closing price cannot be assumed executable at that same close before it is known. Record the alert timestamp and the next available quote, then include spread, fees and slippage. A screenshot of the eventual candle is not an execution record.
Turn a reading into a testable question
Write the complete rule before inspecting its outcome: instrument set, data window, entry timing, exit, size limit, missing-data behaviour and costs. Keep a later period unused while choosing settings. Compare the rule with a simple baseline on identical timestamps and capital assumptions, reporting net results, drawdown, number of trades and unresolved data gaps.
If many parameter combinations were tried, keep that search history. Selecting the best-looking combination and reporting only its winning period exaggerates the evidence. Stop at a descriptive chart when the data cannot support a strategy test. There is no requirement to turn every indicator reading into an order.
Method: synthetic arithmetic and linked primary formula documentation reviewed on 26 October 2026. No personal trading results or predictive performance claimed. AITR may earn commissions from commercial links elsewhere on the site; these documentation links are not affiliate recommendations. Trading can lose money.