Skip to content

fix(torghut): block capital-unsafe replay feedback #6823

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
gregkonush wants to merge 1 commit into
base: main
Choose a base branch
from
+245 −21
Open

fix(torghut): block capital-unsafe replay feedback #6823

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
80 changes: 67 additions & 13 deletions services/torghut/scripts/run_whitepaper_autoresearch_profit_target.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -117,15 +117,26 @@
"positive_day_ratio_below_oracle",
"best_day_share_above_oracle",
"active_day_ratio_failed",
"positive_day_ratio_failed",
"min_daily_net_pnl_failed",
"daily_net_observed_day_count_failed",
"best_day_share_failed",
"max_single_day_contribution_share_failed",
"max_single_symbol_contribution_share_failed",
"max_cluster_contribution_share_failed",
"worst_day_loss_above_oracle",
"max_drawdown_above_oracle",
}
)
_REALIZED_CAPITAL_HARD_BLOCK_ORACLE_BLOCKERS = frozenset(
{
"max_gross_exposure_above_oracle",
"min_cash_below_oracle",
"negative_cash_observed",
"executable_replay_account_buying_power_failed",
"executable_replay_max_notional_per_trade_failed",
"executable_replay_notional_within_buying_power_failed",
}
)
_DEFAULT_FEEDBACK_ORACLE_POLICY = ProfitTargetOraclePolicy()


def _default_strategy_config_path() -> Path:
Expand Down Expand Up @@ -2562,6 +2573,21 @@ def _feedback_daily_net_has_loss(scorecard: Mapping[str, Any]) -> bool:
)


def _feedback_has_excessive_drawdown(scorecard: Mapping[str, Any]) -> bool:
oracle_blockers = _oracle_blockers(scorecard)
if oracle_blockers & {"worst_day_loss_above_oracle", "max_drawdown_above_oracle"}:
return True
if (
_decimal(scorecard.get("worst_day_loss"))
> _DEFAULT_FEEDBACK_ORACLE_POLICY.max_worst_day_loss
):
return True
return (
_decimal(scorecard.get("max_drawdown"))
> _DEFAULT_FEEDBACK_ORACLE_POLICY.max_drawdown
)


def _feedback_family_prior_has_hard_block(scorecard: Mapping[str, Any]) -> bool:
oracle_blockers = _oracle_blockers(scorecard)
if oracle_blockers & _FAMILY_PRIOR_HARD_BLOCK_ORACLE_BLOCKERS:
Expand All @@ -2572,19 +2598,23 @@ def _feedback_family_prior_has_hard_block(scorecard: Mapping[str, Any]) -> bool:
return True
if _decimal(scorecard.get("best_day_share")) > Decimal("0.50"):
return True
return _feedback_daily_net_has_loss(scorecard)
return _feedback_has_excessive_drawdown(scorecard)


def _feedback_risk_profile_has_penalty(scorecard: Mapping[str, Any]) -> bool:
oracle_blockers = _oracle_blockers(scorecard)
if oracle_blockers & _RISK_PROFILE_FEEDBACK_ORACLE_BLOCKERS:
return True
if _feedback_has_realized_capital_hard_block(scorecard):
return True
if _decimal(scorecard.get("active_day_ratio"), default="1") < Decimal("1"):
return True
if _decimal(scorecard.get("positive_day_ratio"), default="1") < Decimal("0.60"):
return True
if _decimal(scorecard.get("best_day_share")) > Decimal("0.35"):
return True
if _feedback_has_excessive_drawdown(scorecard):
return True
if _decimal(scorecard.get("max_single_day_contribution_share")) > Decimal("0.35"):
return True
if _decimal(scorecard.get("max_single_symbol_contribution_share")) > Decimal(
Expand All @@ -2596,14 +2626,21 @@ def _feedback_risk_profile_has_penalty(scorecard: Mapping[str, Any]) -> bool:
return False


def _feedback_is_blocked(scorecard: Mapping[str, Any]) -> bool:
if _feedback_scorecard_has_hard_veto(scorecard):
def _feedback_has_realized_capital_hard_block(scorecard: Mapping[str, Any]) -> bool:
oracle_blockers = _oracle_blockers(scorecard)
if oracle_blockers & _REALIZED_CAPITAL_HARD_BLOCK_ORACLE_BLOCKERS:
return True
if _decimal(scorecard.get("max_gross_exposure_pct_equity")) > Decimal("1.0"):
return True
if _decimal(scorecard.get("min_cash")) < Decimal("0"):
return True
if _decimal(scorecard.get("negative_cash_observation_count")) > Decimal("0"):
return _decimal(scorecard.get("negative_cash_observation_count")) > Decimal("0")


def _feedback_is_blocked(scorecard: Mapping[str, Any]) -> bool:
if _feedback_scorecard_has_hard_veto(scorecard):
return True
if _feedback_has_realized_capital_hard_block(scorecard):
return True
return (
_decimal(scorecard.get("net_pnl_per_day")) <= Decimal("0")
Expand Down Expand Up @@ -2939,19 +2976,23 @@ def row_selection_reason(candidate_spec_id: str) -> str:
bundle.objective_scorecard
)
if source == "feedback_real_replay" and is_blocked:
if bundle is not None and _feedback_has_nonpositive_expected_value(
bundle.objective_scorecard
if bundle is not None and (
_feedback_has_nonpositive_expected_value(bundle.objective_scorecard)
or _feedback_has_realized_capital_hard_block(bundle.objective_scorecard)
):
return "pre_replay_mlx_feedback_blocked"
return "pre_replay_mlx_feedback_penalized"
if source == "feedback_execution_signature_replay" and is_blocked:
if bundle is not None and _feedback_has_nonpositive_expected_value(
bundle.objective_scorecard
if bundle is not None and (
_feedback_has_nonpositive_expected_value(bundle.objective_scorecard)
or _feedback_has_realized_capital_hard_block(bundle.objective_scorecard)
):
return "pre_replay_mlx_signature_feedback_blocked"
return "pre_replay_mlx_signature_feedback_penalized"
if source == "feedback_shape_prior" and bundle is not None:
if _feedback_family_prior_has_hard_block(bundle.objective_scorecard):
if _feedback_family_prior_has_hard_block(
bundle.objective_scorecard
) or _feedback_has_realized_capital_hard_block(bundle.objective_scorecard):
return "pre_replay_mlx_shape_feedback_blocked"
if is_blocked:
return "pre_replay_mlx_family_feedback_penalized"
Expand All @@ -2960,6 +3001,8 @@ def row_selection_reason(candidate_spec_id: str) -> str:
and bundle is not None
and _feedback_risk_profile_has_penalty(bundle.objective_scorecard)
):
if _feedback_has_realized_capital_hard_block(bundle.objective_scorecard):
return "pre_replay_mlx_risk_profile_feedback_blocked"
return "pre_replay_mlx_risk_profile_feedback_penalized"
if (
source == "feedback_family_replay"
Expand All @@ -2977,7 +3020,10 @@ def proposal_score_for_item(candidate_spec_id: str, raw_score: float) -> float:
if (
source in {"feedback_real_replay", "feedback_execution_signature_replay"}
and bundle is not None
and _feedback_has_nonpositive_expected_value(bundle.objective_scorecard)
and (
_feedback_has_nonpositive_expected_value(bundle.objective_scorecard)
or _feedback_has_realized_capital_hard_block(bundle.objective_scorecard)
)
):
return min(-1_000_000.0, target_by_spec.get(candidate_spec_id, raw_score))
if (
Expand All @@ -2989,14 +3035,21 @@ def proposal_score_for_item(candidate_spec_id: str, raw_score: float) -> float:
if (
source == "feedback_shape_prior"
and bundle is not None
and _feedback_family_prior_has_hard_block(bundle.objective_scorecard)
and (
_feedback_family_prior_has_hard_block(bundle.objective_scorecard)
or _feedback_has_realized_capital_hard_block(bundle.objective_scorecard)
)
):
return min(-1_000_000.0, target_by_spec.get(candidate_spec_id, raw_score))
if (
source == "feedback_risk_profile_prior"
and bundle is not None
and _feedback_risk_profile_has_penalty(bundle.objective_scorecard)
):
if _feedback_has_realized_capital_hard_block(bundle.objective_scorecard):
return min(
-1_000_000.0, target_by_spec.get(candidate_spec_id, raw_score)
)
return min(-500_000.0, target_by_spec.get(candidate_spec_id, raw_score))
if (
source == "feedback_family_replay"
Expand Down Expand Up @@ -3134,6 +3187,7 @@ def _select_candidate_specs_for_replay(
"pre_replay_mlx_feedback_blocked",
"pre_replay_mlx_signature_feedback_blocked",
"pre_replay_mlx_shape_feedback_blocked",
"pre_replay_mlx_risk_profile_feedback_blocked",
"pre_replay_mlx_family_feedback_blocked",
}

Expand Down
Loading
Loading