PyneCore 6.5.1 still differs from TradingView strategy results for intraday request.security() HTF futures sessions

[cald3ra]

Describe the bug

After upgrading to pynesys-pynecore==6.5.1, PyneCore still does not fully match saved TradingView Strategy Tester results for a Pine strategy that uses historical request.security(..., lookahead=barmerge.lookahead_off) on intraday higher-timeframe bars for a non-24h futures symbol.

This is a follow-up to #66. The 6.5.1 release note says historical barmerge.lookahead_off now confirms the closed HTF bar on the period's own last chart bar instead of one bar late. That appears to address the specific one-bar-late timing bug, but strategy-level parity is still not reached on the tested V2.12.2 configurations.

The proprietary strategy and full market CSV are not attached publicly. The isolated Pine expression remains:

midLineHTF = request.security(
     syminfo.tickerid,
     midHTF,
     (ta.highest(high, midLen) + ta.lowest(low, midLen)) / 2.0,
     lookahead=barmerge.lookahead_off)

The attached minimal Pine/PyneCore scripts only plot this expression. They do not include proprietary entry/exit logic.

To Reproduce

1. Install PyneCore 6.5.1:

   python -m pip install -U pynesys-pynecore==6.5.1 typer

2. Use a 30-minute futures-like symbol with non-24h sessions and partial intraday HTF candles.
3. Run the attached minimal Pine script in TradingView:
   • minimal_htf_midpoint_repro.pine
4. Run the attached minimal PyneCore script:
   • minimal_htf_midpoint_repro.py
5. For strategy-level evidence, see:
   • comparison_pynecore_651_vs_tv.csv
   • pynecore_651_summary_L30_X30_HTF720.csv
   • pynecore_651_summary_L1_X0_HTF60.csv
   • tv_metrics_L30_X30_HTF720.csv
   • tv_metrics_L1_X0_HTF60.csv

The two validated strategy configurations are:

Config	Chart TF	midLen	midX	useMidHTF	midHTF
L30_X30_HTF720	30m	30	30	true	720
L1_X0_HTF60	30m	1	0	true	60

The exact comparison uses saved TradingView XLSX reports ending 2026-06-05. The local PyneCore CSV also ends at 2026-06-05, so later live TradingView data is not part of this comparison.

Observed behavior

PyneCore 6.5.1 still differs from TradingView at strategy-summary level.

Config	Window	TradingView trades	PyneCore 6.5.1 trades	Delta trades	TradingView PnL	PyneCore 6.5.1 PnL	Delta PnL	TradingView PF	PyneCore 6.5.1 PF
L30_X30_HTF720	2023	18	21	+3	3336.20	7525.00	+4188.80	1.348	1.907
L30_X30_HTF720	2024	17	21	+4	5957.80	8500.00	+2542.20	1.828	2.172
L30_X30_HTF720	2025	16	21	+5	5179.40	1800.00	-3379.40	2.715	1.278
L30_X30_HTF720	Jun2022-Jun2026	74	80	+6	23426.60	22875.00	-551.60	1.713	1.705
L30_X30_HTF720	2020-Jun2026	108	68	-40	12742.20	-13850.00	-26592.20	1.222	0.683
L1_X0_HTF60	2023	27	26	-1	12441.80	13050.00	+608.20	2.185	2.286
L1_X0_HTF60	2024	30	28	-2	8577.00	10175.00	+1598.00	1.763	2.025
L1_X0_HTF60	2025	27	27	0	7591.80	8375.00	+783.20	1.952	2.120
L1_X0_HTF60	Jun2022-Jun2026	111	108	-3	52377.40	55950.00	+3572.60	2.244	2.410
L1_X0_HTF60	2020-Jun2026	168	163	-5	69871.20	75650.00	+5778.80	1.874	1.999

The HTF720 continuous 2020-Jun2026 case is the largest remaining divergence: TradingView has 108 trades and +12742.20 PnL, while PyneCore 6.5.1 has 68 trades and -13850.00 PnL.

Expected behavior

For historical data, PyneCore should expose HTF values from request.security(..., lookahead_off) on the same lower-timeframe bars as TradingView and should produce matching strategy trades when the same native 30-minute and HTF data are supplied.

The remaining parity should hold around:

• partial intraday HTF candles
• midday or overnight session gaps
• shortened bars
• non-24h futures sessions
• native HTF data supplied through --security

Screenshots / evidence

Attached ZIP contains:

• comparison_pynecore_651_vs_tv.csv
• pynecore_651_summary_L30_X30_HTF720.csv
• pynecore_651_summary_L1_X0_HTF60.csv
• tv_metrics_L30_X30_HTF720.csv
• tv_metrics_L1_X0_HTF60.csv
• minimal_htf_midpoint_repro.pine
• minimal_htf_midpoint_repro.py
• TV_LIVE_CHECK_2026-06-13.txt
• tv_period_dialog.png
• after_ctrl_enter.png

Environment

• PyneCore package: pynesys-pynecore==6.5.1
• Required CLI dependency installed separately: typer
• OS: Windows 11
• Python: 3.14.4
• Execution: PyneCore CLI
• Chart timeframe: 30 minutes
• Requested HTFs tested: 60 minutes and 720 minutes
• lookahead: barmerge.lookahead_off
• Symbol type: futures-like, non-24h session
• Data end date: 2026-06-05

Additional context

I understand that 6.5.1 fixed the previously reported one-bar-late historical HTF confirmation. The remaining mismatch may be caused by another piece of the TradingView parity chain rather than the exact same issue, for example:

• native HTF feed mapping through --security
• strategy simulator order handling after HTF timing changes
• shortened/partial session bar closing semantics
• barmerge behavior on sparse non-24h futures sessions

Could you advise which area to isolate next, or whether you want a smaller anonymized CSV extract around the first divergent trade/bar for one of the two configs?

[cald3ra]

pynecore_651_followup_issue.zip