Bar magnifier

Table of Contents

• When to enable it
• Enabling
• The intrabars a Pine strategy walks
• Distribution modes
• Diagnostics
• How it interacts with request.security()
• Reproducibility

The bar magnifier sub-samples each parent bar into synthetic intra-bar slices, allowing intra-bar order fills to resolve at finer granularity than the source feed. It's the runtime's mechanism for matching TradingView's "Bar Magnifier" feature.

When to enable it

Enable the magnifier when:

• Your strategy uses intra-bar stop / limit / take-profit orders whose fills depend on the OHLC path inside the parent bar.
• You're parity-checking against a TV backtest that has Bar Magnifier on.

Skip it when:

• Orders are bar-close-only (process_orders_on_close = true).
• You're running purely on a 1m or sub-minute feed where no extra resolution exists to recover.

Enabling

run_backtest_full(s, bars, n,
                  /*input_tf */ "5",
                  /*script_tf*/ "60",
                  /*bar_magnifier  */ 1,
                  /*magnifier_samples*/ 4,
                  PF_MAGNIFIER_ENDPOINTS,
                  &report);

Arg	Effect
bar_magnifier	Boolean toggle.
magnifier_samples	Sub-bar samples per parent bar. Typical: 4. Higher = finer fills, more CPU.
magnifier_dist	Sampling density profile — see below.

Add volume-weighted sample density as a sticky toggle on the handle:

strategy_set_magnifier_volume_weighted(s, 1);

The intrabars a Pine strategy walks

When the input feed is finer than the script timeframe, a compiled Pine strategy's magnifier does not walk the input bars: it walks the intrabars TradingView's bar magnifier walks, which the Pine host builds from the input (tradingview_magnifier_bars magnifier_intrabars.hpp:46). TradingView's help centre ("What is bar magnifier backtesting mode") fixes the intrabar timeframe per chart timeframe, each row covering the charts from it up to the next row (tradingview_intrabar_timeframe magnifier_intrabars.hpp:21):

Chart	Intrabar	Buildable from 1-minute bars
1 minute	10 seconds	no — needs a 10-second feed
5 minutes	30 seconds	no — needs a 30-second feed
10 minutes	1 minute	yes (the feed itself)
15 minutes	2 minutes	yes
30 minutes	5 minutes	yes
1 hour	10 minutes	yes
4 hours	30 minutes	yes
1 day	1 hour	yes
3 days	4 hours	yes
1 week and up	1 day	yes

Each intrabar is the symbol's regular bar of that timeframe — anchored at the session day's open (00:00 UTC on a 24x7 symbol, 09:15 on NSE, 09:30 on NYSE) and cut at the session's close — and belongs to the chart bar that holds its last minute. On a 15-minute chart the 2-minute intrabar from 10:14 to 10:16 belongs to the 10:15 bar, so a chart bar's path is its own open, the intrabars it owns (each an open, its nearer extreme, its other extreme, its close), then its own close. The host marks the chart bar's open and close as one-price bars that traded nothing where an intrabar straddles its first minute or leaves its last one to the next bar, and the kernel walks such a bar as one point. The rows from 1 minute to 1 day were measured against TradingView's own request.security_lower_tf bars and magnified exports (tests/test_adapter_magnifier_intrabar_tapes.cpp, the private fixture set magnifier_intrabars).

The host walks the input bars themselves when it cannot build TradingView's: an intrabar finer than the input or not a multiple of it (a 1- or 5-minute chart on 1-minute bars), or a session it cannot resolve. A daily chart of an exchange-listed stock is built from the minutes, but TradingView prices that chart bar's open and close from its daily feed (the official open and close), which a 1-minute feed does not carry.

Distribution modes

See also

pf_magnifier_distribution_t for the enum.

Mode	Density profile	Use case
PF_MAGNIFIER_UNIFORM	Even spacing across the bar.	Symmetric noise; no prior on where fills cluster.
PF_MAGNIFIER_COSINE	Dense at the bar's two ends (cosine spacing), sparse in the middle.	Smooth volatility profile.
PF_MAGNIFIER_TRIANGLE	Linear taper from a peak.	Single-peaked intraday activity.
PF_MAGNIFIER_ENDPOINTS (default)	Exact O,H,L,C points + uniform fill between.	TV-parity default. Best for stop/limit fill realism.
PF_MAGNIFIER_FRONT_LOADED	Density biased toward bar open.	Open-driven assets (futures session opens).
PF_MAGNIFIER_BACK_LOADED	Density biased toward bar close.	Close-driven assets (equities rebalance).

Diagnostics

Every report carries two magnifier counters:

Field	Meaning
magnifier_sub_bars_total	Synthetic intra-bar slices generated.
magnifier_sample_ticks_total	Sample ticks visited.

The kernel's intrabar driver counts both for every run that walks an intrabar path, so a bare C++ or C native host reports them as a compiled strategy does; a bar that was assigned no lower-feed bars counts nothing. The Pine host mirrors the same counts at its bar callbacks, and no state hash folds them.

Quick sanity check: with magnifier_samples = 4 and PF_MAGNIFIER_ENDPOINTS, expect four sample ticks per sub-bar and one per one-price bar that traded nothing. A Pine strategy walks TradingView's intrabars, not the input bars (above): a 15-minute chart on 1-minute bars visits about 4 * 7.5 + 1 ticks per chart bar, not 4 * 15.

How it interacts with request.security()

A higher-timeframe feed brought in via request.security() follows the parent timeframe of the security site, not the script TF. The magnifier still runs on the script-TF parent bar; the security feed delivers complete-bar values when the parent bar closes, partial-bar values otherwise. See pf_report_t::security_partial_total to count partial-bar evaluations.

Reproducibility

The magnifier is deterministic for a given (bars, magnifier_samples, magnifier_dist, volume_weighted) tuple. Two runs with identical inputs produce bit-identical trade lists. There is no internal RNG seeded from time or address.