An entry sized against pending entries that did not fill uses its own quantity

[alexgrover]

A strategy.entry queued after an opposite market entry on the same bar is sized as a reversal of the position that entry would open. When that entry doesn't fill (for example, rejected for margin), the reversal size is wrong.

Example: a short of 2 rejected for margin, then a long of 1. PineTS sized the long as 1 + 2 = 3, which was rejected too, so nothing traded. TradingView fills the long with its own quantity of 1 (probes o07 / o09, BINANCE:BTCUSDT 1h).

Change

The entry records that its reversal size came from pending orders rather than the actual position (_reversesPending). If the position is flat when it fills, it uses its own quantity. A reversal sized against an actual position keeps its frozen quantity, as before (TradingView behavior established in #316).

Not included: two opposite entries that both fill (case 29)

This PR started as a fix for controlled case 29 (a market short then a market long, from flat, on one bar), where TradingView fills only the short. A rule that dropped the second entry fixed case 29 and four probes, but broke three premium backtester runs: there, TradingView does reverse on the same bar.

More probes show the outcome depends on capital and margin:

• with 100,000 of capital at 100% margin, the second entry doesn't fill;
• with 1,000,000 of capital or margin 0, the long fills first (even when placed second) and the short then closes part of it.

That needs a proper model of TradingView's execution order and margin check for same-bar orders, so it's left for a separate investigation.

Tests

tests/namespaces/strategy/opposite-entries.test.ts (2 tests): the long-after-rejected-short case and its mirror. Both fail on dev.

Results against TradingView (compared with dev)

• Probes o07 and o09: 0 trades → identical.
• Controlled cases and premium backtesters: no change.
• Public strategies: ict-daily-liquidity-sweep (already diverging at trade 0) no longer opens the spurious same-bar long that dev opened on bar 2377. It now has TradingView's short there; the trade count changes from 53 to 55 on the already-diverged ledger.
• Unit suite: all committed tests pass. Oracle suite: 0 regressions.

---

Note

Medium Risk
Changes market entry fill sizing in the strategy engine; incorrect logic would skew backtests, but the change is narrow and covered by new tests.

Overview
Fixes same-bar opposite strategy.entry orders when the first order never opens a position (e.g. margin rejection) but the second was queued as a reversal of the first’s projected size.

strategy.entry now tags those orders with _reversesPending when reversal sizing came only from earlier market entries on the bar, not from an actual open position. fillEntryOrder resets quantity to _base_qty at fill time if the account is still flat—so a long of 1 after a rejected short of 2 fills 1, not 3. Reversals against a real position keep the existing frozen reversal qty (#316).

Adds opposite-entries.test.ts (margin-too-tight mirror cases). Does not change behavior when both opposite entries fill on one bar (left for follow-up).

^{Reviewed by Cursor Bugbot for commit 40f3f8b. Bugbot is set up for automated code reviews on this repo. Configure here.}

[cursor]

Bugbot couldn't run - usage limit reached

Bugbot is counted against Cursor usage for this user or team, and this run hit a usage or spend limit.

A user or team admin can review and increase usage limits in the Cursor dashboard.

(requestId: ff01fcac-ad0d-4d01-a264-9f3eaf55bd22)