True Strength Index Strategy: Calculation, Crossovers, and Limits
True Strength Index is a momentum oscillator built from double-smoothed price changes. It is distinct from TradingView's similarly abbreviated Trend Strength Index, which measures price/time correlation. Confirm the full name and calculation before copying settings or interpreting a TSI signal.
Calculation
Compute the change between successive closes. Apply two stages of EMA smoothing to that change, and separately apply the same smoothing stages to its absolute value. Divide the smoothed change by the smoothed absolute change and multiply by 100. A signal line can then smooth the resulting TSI.
Common settings use 25 and 13 smoothing periods, with a separate signal-line period. They are conventions, not a required 1:2 ratio or proof of an optimal strategy. Use the actual script inputs and handle insufficient history or a zero denominator. TradingView documents the calculation in its True Strength Index guide.
Double smoothing reduces some short-term variation and adds lag. It does not guarantee a fixed one- or two-bar delay, superiority to RSI, or earlier warning than MACD.
Three different signal ideas
| Signal | What it describes | Main limitation |
|---|---|---|
| Zero-line crossing | Smoothed signed momentum changes sign | Can repeat in sideways conditions |
| Signal-line crossing | TSI crosses its own smoothed average | Frequent crossings may have little trading value |
| Divergence | Selected price swings and oscillator swings differ | Swing selection and confirmation timing can be subjective |
An oscillator above zero is a state, while crossing above zero is an event. Decide whether repeated signals may add to a position. For divergence, specify how pivots are detected and when they become known; a marker drawn on an earlier pivot is not necessarily information available at that earlier bar.
Thresholds are hypotheses
Values such as ±25 can be used as reference zones, but their meaning depends on the instrument and settings. A strong reading can persist in a trend. It is not an automatic reversal instruction.
There is no established rule that high-beta stocks require ±70 or low-beta stocks require ±50. Beta relative to a benchmark is also not identical to the volatility of the oscillator. Test proposed thresholds rather than assuming a category label determines them.
Build a complete experiment
Choose one signal definition, a position direction, an exit, protective orders, size, and costs. State whether the rule waits for the chart bar to close. If adding a moving-average filter, compare the filtered and unfiltered versions under identical conditions.
Inspect complete trade records, including missed moves and losing signals. Reserve data that was not used to choose the thresholds, and compare real-time observations with historical output. RSI, MACD, and TSI all depend on price, so agreement is not independent confirmation by itself.
A backtest analysis tool can summarize the sample, but resampling historical trades does not establish an edge or predict all future losses. No reproducible evidence supports the former TSLA/NVDA/AMD/META trade anecdotes, 60% success claim, or claimed 1996–2023 mean-reversion study. They have been removed rather than presented as personal experience.
