Skip to content

Combine Portfolio Pine — running multiple strategies in one Pine

alpha-forge pine generate --combine-strategies packages several buy-hold-overlay strategies into a single Pine v6 Indicator that runs on TradingView. Each sub-strategy fires alert() calls that are routed to the alpha-strike webhook and executed on moomoo / OANDA.

Why an Indicator?

Pine v6 strategy() is single-instrument only, so it cannot express a multi-ticker portfolio. The combine portfolio Pine takes a different shape:

  • Sub-strategy prices are pulled with request.security() per ticker
  • A target_qty is recomputed on every bar close per sub-strategy
  • Quantity changes (or a rebalance trigger) emit alert() calls
  • The receiving side (alpha-strike) parses the webhook v2 payload and places orders on moomoo / OANDA

As a side effect the TradingView Strategy Tester does not work on combine Pine (Strategy Tester only attaches to strategy()). Use the symbolic verify flow instead to guarantee that the emitted Pine matches the alpha-forge backtest combine logic.

Quick start

1. Generate the combine Pine

alpha-forge pine generate \
  --combine-strategies tqqq_phase2,gld_bh,tlt_bh \
  --allocation equal \
  --rebalance-freq monthly \
  --rebalance-threshold 0.05 \
  --portfolio-id beat_qqq_hedged_v1 \
  --with-webhook \
  --webhook-broker moomoo \
  --webhook-run-mode paper

Options:

Option Meaning
--combine-strategies sid1,sid2,... Comma-separated strategy IDs (≥2) to combine
--allocation equal\|custom Equal split or custom weights via --weights
--weights tqqq=0.5,gld=0.3,tlt=0.2 Custom weights (sum 1.0 ± 0.01)
--rebalance-freq weekly\|monthly\|quarterly\|yearly\|none Periodic rebalance frequency
--rebalance-threshold 0.05 Trigger rebalance when weight drift exceeds ±5%
--allow-non-buy-hold Allow mean-reversion / trend-following inside combine (Phase 2 experimental)
--with-training-data Fetch training data in parallel for HMM strategies and embed Forward Algorithm parameters (issue #974)
--portfolio-id <id> Indicator name + webhook payload portfolio_id

2. Paste into the TradingView Pine Editor

Open output/pinescript/<portfolio_id>.pine, paste into the Pine Editor, and click "Add to chart".

3. Configure the alert

Field Value
Condition The <portfolio_id> indicator → Any alert() function call
Frequency Once Per Bar Close (matches alert.freq_once_per_bar_close in Pine)
Webhook URL https://strike.alforgelabs.com/webhook (replace per environment)
Message Leave empty — Pine's make_payload() returns the full JSON itself

After saving, the first 3-ticker entry fires on the next daily bar close.

Webhook payload shape

make_payload() emits the alpha-strike webhook v2 payload on a single line:

{
  "passphrase": "<i_passphrase input>",
  "broker": "moomoo",
  "asset_class": "US",
  "action": "buy",
  "ticker": "US.TQQQ",
  "quantity": 33,
  "target_qty": 47,
  "strategy_id": "beat_qqq_hedged_v1",
  "sub_strategy_id": "tqqq_phase2",
  "portfolio_id": "beat_qqq_hedged_v1",
  "signal_id": "20260529-135959",
  "run_mode": "paper",
  "timeframe": "1D"
}

run_mode is paper or live, switchable from the TradingView input panel. Keep paper during validation to route to moomoo SIMULATE.

target_qty is the sub-strategy's absolute target holding (alpha-forge #1037). alpha-strike v0.7.0+ re-resolves the order side/quantity from the difference between target_qty and the broker's actual position, so even if zero fills, partial fills, or rounding drift Pine's assumed position, the real holding converges back to the target on the next signal (closed-loop). quantity remains as the delta fallback for versions without target_qty support.

HMM Forward Algorithm reproduction (issue #974)

When the combine includes an HMM regime strategy, pass --with-training-data to fetch each HMM's training data in parallel, embed the transition matrix / means / variances, and run the Forward Algorithm inside Pine bar by bar.

  • Parallel fetch via concurrent.futures.ThreadPoolExecutor
  • Identifiers stay disjoint across HMMs through a prefix scheme (s0_hmm_transmat, s1_hmm_transmat, s0_f_hmm_step, s1_f_hmm_step, ...)
  • Without training data the volatility regime proxy (ta.stdev) is used as a fallback (backward compatible)
alpha-forge pine generate \
  --combine-strategies qqq_hmm_v1,tqqq_v1,gld_v1 \
  --allocation equal \
  --with-training-data \
  --portfolio-id combine_hmm_v1

symbolic verify (issue #975)

Paid plans only

alpha-forge pine verify is available on paid plans (Lifetime / Annual / Monthly) only. See Trial limits for Trial plan restrictions.

alpha-forge pine verify --combine-strategies validates that the generated Pine and the alpha-forge backtest combine share the same intent — without talking to TradingView at all.

alpha-forge pine verify \
  --combine-strategies tqqq_phase2,gld_bh,tlt_bh \
  --combine-allocation equal \
  --combine-rebalance-freq monthly \
  --combine-rebalance-threshold 0.05 \
  --combine-portfolio-id beat_qqq_hedged_v1 \
  --check-mode metrics

Sample output:

# Combine Portfolio Verify — beat_qqq_hedged_v1

**Status**: ✅ PASSED (0 violation(s) / 12 checks)

## Metrics

| Metric            | Value     |
|-------------------|-----------|
| total_return_pct  | 915.5400  |
| cagr_pct          | 15.3000   |
| sharpe_ratio      | 1.0401    |
| max_drawdown_pct  | 23.3200   |
| volatility_pct    | 14.7700   |

## Integrity Checks (excerpt)

| Key                                     | Expected | Observed | OK |
|-----------------------------------------|----------|----------|----|
| pine_syntax                             | 0 errors | 0 errors | ✅ |
| weight:tqqq_phase2                      | 0.333333 | 0.3333   | ✅ |
| hedge_exposure:tqqq_phase2              | 0.4      | 0.4      | ✅ |
| rebalance_freq                          | monthly  | monthly  | ✅ |
| rebalance_threshold                     | 0.05     | 0.05     | ✅ |

Integrity checks cover:

  • pine_syntax — Pine v6 syntax verify (PineV6Validator)
  • weight:<sid> — Pine s{i}_weight versus backtest combine weight_map[sid]
  • base_exposure:<sid> — exposure during normal regime
  • hedge_exposure:<sid> — hedge state target_exposure_pct
  • rebalance_freq — ta.change(month) etc. matched against CLI spec
  • rebalance_threshold — i_rebalance_threshold input value

Pass --check-mode compile_only for syntax verify only (no backtest required).

hybrid-strategy mode (issue #985)

The default indicator mode emits a single Pine v6 Indicator and the Strategy Tester will not attach. When you want partial metrics for at least the main symbol from the TradingView Strategy Tester, use --combine-mode hybrid-strategy.

alpha-forge pine generate \
  --combine-strategies tqqq_phase2,gld_bh,tlt_bh \
  --combine-mode hybrid-strategy \
  --main-strategy tqqq_phase2 \
  --portfolio-id beat_qqq_hedged_v1 \
  --with-webhook \
  --webhook-broker moomoo

Behaviour:

  • The strategy passed to --main-strategy is emitted as a chart-native strategy() (default_qty_value = weight × 100, position managed via strategy.entry / strategy.close). Apply it to the main symbol's chart and the Strategy Tester computes Sharpe / CAGR / MDD and so on.
  • The remaining sub-strategies still trade through request.security() + alert() to alpha-strike (identical to indicator mode).
  • The main strategy must be buy-hold-overlay (emitting mean-reversion / trend-following as strategy() is not supported).
Option Meaning
--combine-mode indicator\|hybrid-strategy Output mode. indicator is the default (backward compatible)
--main-strategy <sid> Main strategy ID emitted as strategy() in hybrid mode (a buy-hold-overlay included in the combine)

Metrics are an approximation

The main strategy's strategy.entry / strategy.close does not match AlphaForge backtest combine (vectorbt) exactly, and sub-strategies are not included in the Strategy Tester metrics. Keep symbolic verify or alpha-forge backtest combine as the source of truth for portfolio metrics; use hybrid-strategy as an aid to eyeball / sanity-check the main symbol's behaviour on TradingView.

Auto-verify the Strategy Tester result (issue #986)

After applying the hybrid-strategy Pine to the main symbol's chart and running the Strategy Tester, use alpha-forge pine verify --verify-mode tradingview-strategy-tester to automatically compare the main symbol component's Sharpe / Max Drawdown against alpha-forge backtest combine's same-symbol component. The Strategy Tester aggregates are fetched via the MCP server (vinicius / tradesdontlie).

alpha-forge pine verify \
  --combine-strategies tqqq_phase2,gld_bh,tlt_bh \
  --combine-mode hybrid-strategy \
  --main-strategy tqqq_phase2 \
  --check-mode metrics \
  --verify-mode tradingview-strategy-tester \
  --symbol TQQQ --interval D \
  --mcp-server-flavor vinicius

How it works:

  • alpha-forge backtests the main strategy standalone, then blends its equity to the weight allocation ((1-w) + w·ratio, equivalent to the Pine default_qty_value = weight × 100) to build the reference metrics. This keeps Sharpe / Max Drawdown consistent with the Strategy Tester (percent_of_equity sizing).
  • TradingView side fetches the Strategy Tester aggregates on the main symbol's chart via MCP.

Only the main component is comparable

Portfolio-level Sharpe cannot be derived from the Strategy Tester (single symbol only). The TradingView Strategy Tester also does not expose Calmar directly, so PASS/FAIL is gated on Sharpe + Max Drawdown and the reference-side Calmar is shown for information only. Use symbolic verify for portfolio-level validation.

Known limitations

  • No Strategy Tester in indicator mode: the default combine Pine is an Indicator, so TradingView cannot show Sharpe / CAGR / MDD natively. Use symbolic verify or alpha-forge backtest combine for portfolio-level metrics, or hybrid-strategy mode for partial metrics on the main symbol.
  • Once Per Bar Close enforced: Pine sets alert.freq_once_per_bar_close to suppress intrabar firing — please pick the same frequency in the alert dialog.
  • First entry timing: each sub-strategy alert fires only after its bar closes, so initial entries land on the first daily close after the alert is saved.