254 lines
9.7 KiB
Python
254 lines
9.7 KiB
Python
"""Review-aware factor scoring.
|
|
|
|
This module is the bridge between daily review results and the next screening
|
|
run. Strategy code should score named factors here instead of hard-coding every
|
|
``score += N`` forever.
|
|
"""
|
|
|
|
from __future__ import annotations
|
|
|
|
from dataclasses import dataclass, field
|
|
from typing import Any
|
|
|
|
from app.config.config_loader import get_signal_weights
|
|
from app.core.signal_taxonomy import signal_label_for_code
|
|
|
|
|
|
DEFAULT_FACTOR_WEIGHTS = {
|
|
"vp_fly_1h_current": 5.0,
|
|
"volume_consecutive_1h": 4.0,
|
|
"cex_top_gainer_24h": 2.0,
|
|
"volume_divergence_1h": 1.0,
|
|
"static_accum_4h": 5.0,
|
|
"higher_lows_4h": 2.0,
|
|
"compression_surge_4h": 2.0,
|
|
"box_breakout_pullback_1h": 6.0,
|
|
"box_breakout_pullback_4h": 8.0,
|
|
"ignition_1h_current": 4.0,
|
|
"ignition_4h_current": 3.0,
|
|
"ignition_d1_current": 6.0,
|
|
"dynamic_k_1h_bull": 3.0,
|
|
"dynamic_k_d1_bull": 5.0,
|
|
"breakout_pullback_d1": 8.0,
|
|
"breakout_pullback_w1": 8.0,
|
|
"breakout_15m_current": 3.0,
|
|
"pullback_15m_confirm": 2.0,
|
|
"strong_resonance_bypass": 3.0,
|
|
"entry_quality_gate": 2.0,
|
|
"top_trader_long": 1.0,
|
|
"sector_rotation": 2.0,
|
|
"sentiment_resonance": 2.0,
|
|
"dex_volume_spike": 2.0,
|
|
"liquidity_add": 1.5,
|
|
"exchange_outflow": 2.0,
|
|
"whale_accumulation": 2.5,
|
|
"smart_money_buying": 2.5,
|
|
"liquidity_remove_risk": 2.5,
|
|
"exchange_inflow_risk": 2.5,
|
|
"holder_concentration_risk": 2.0,
|
|
"funding_extreme": 2.0,
|
|
"trend_exhaustion": 3.0,
|
|
"false_breakout": 5.0,
|
|
"high_position_reject": 5.0,
|
|
"risk_reward_bad": 2.0,
|
|
}
|
|
|
|
FACTOR_GROUPS = {
|
|
"vp_fly_1h_current": "momentum",
|
|
"volume_consecutive_1h": "participation",
|
|
"cex_top_gainer_24h": "momentum",
|
|
"volume_divergence_1h": "risk",
|
|
"static_accum_4h": "structure",
|
|
"higher_lows_4h": "structure",
|
|
"compression_surge_4h": "structure",
|
|
"box_breakout_pullback_1h": "structure",
|
|
"box_breakout_pullback_4h": "structure",
|
|
"ignition_1h_current": "momentum",
|
|
"ignition_4h_current": "momentum",
|
|
"ignition_d1_current": "momentum",
|
|
"dynamic_k_1h_bull": "momentum",
|
|
"dynamic_k_d1_bull": "momentum",
|
|
"breakout_pullback_d1": "structure",
|
|
"breakout_pullback_w1": "structure",
|
|
"breakout_15m_current": "entry_quality",
|
|
"pullback_15m_confirm": "entry_quality",
|
|
"strong_resonance_bypass": "structure",
|
|
"entry_quality_gate": "entry_quality",
|
|
"top_trader_long": "positioning",
|
|
"sector_rotation": "narrative",
|
|
"sentiment_resonance": "narrative",
|
|
"dex_volume_spike": "onchain_flow",
|
|
"liquidity_add": "onchain_flow",
|
|
"exchange_outflow": "onchain_flow",
|
|
"whale_accumulation": "onchain_flow",
|
|
"smart_money_buying": "onchain_flow",
|
|
"liquidity_remove_risk": "risk",
|
|
"exchange_inflow_risk": "risk",
|
|
"holder_concentration_risk": "risk",
|
|
"funding_extreme": "risk",
|
|
"trend_exhaustion": "risk",
|
|
"false_breakout": "risk",
|
|
"high_position_reject": "risk",
|
|
"risk_reward_bad": "risk",
|
|
}
|
|
|
|
GROUP_CAPS = {
|
|
"momentum": 16.0,
|
|
"participation": 6.0,
|
|
"structure": 16.0,
|
|
"positioning": 4.0,
|
|
"narrative": 5.0,
|
|
"onchain_flow": 6.0,
|
|
"entry_quality": 7.0,
|
|
"risk": 12.0,
|
|
}
|
|
|
|
WEIGHT_ALIASES = {
|
|
"vp_fly_1h_current": ("量价齐飞", "1H当前量价齐飞"),
|
|
"volume_consecutive_1h": ("连续3x放量", "连续3x放量(≥3根)", "1H连续放量"),
|
|
"volume_divergence_1h": ("量价背离", "1H量价背离"),
|
|
"static_accum_4h": ("静K蓄力", "4H静K蓄力"),
|
|
"box_breakout_pullback_1h": ("1H箱体突破回踩", "1H底部箱体突破回踩"),
|
|
"box_breakout_pullback_4h": ("4H箱体突破回踩", "4H底部箱体突破回踩"),
|
|
"ignition_1h_current": ("静K动K转折", "静K→动K转折", "1H当前起爆点"),
|
|
"ignition_4h_current": ("静K动K转折", "静K→动K转折", "4H当前起爆点"),
|
|
"ignition_d1_current": ("静K动K转折", "静K→动K转折", "日线当前起爆点"),
|
|
"breakout_15m_current": ("15min当前突破",),
|
|
"pullback_15m_confirm": ("15min回踩确认",),
|
|
"top_trader_long": ("大户偏多",),
|
|
"sector_rotation": ("板块联动",),
|
|
"sentiment_resonance": ("舆情共振",),
|
|
"dex_volume_spike": ("DEX 放量", "链上成交放量"),
|
|
"liquidity_add": ("流动性增加",),
|
|
"exchange_outflow": ("交易所流出",),
|
|
"whale_accumulation": ("鲸鱼增持",),
|
|
"smart_money_buying": ("聪明钱买入",),
|
|
"liquidity_remove_risk": ("流动性撤出风险",),
|
|
"exchange_inflow_risk": ("交易所流入风险",),
|
|
"holder_concentration_risk": ("持仓集中风险",),
|
|
"funding_extreme": ("资金费率极端",),
|
|
"trend_exhaustion": ("趋势衰减",),
|
|
"false_breakout": ("假突破",),
|
|
"high_position_reject": ("高位拒绝",),
|
|
"risk_reward_bad": ("盈亏比不合格",),
|
|
}
|
|
|
|
|
|
def _safe_float(value: Any, default: float = 0.0) -> float:
|
|
try:
|
|
if value is None or value == "":
|
|
return default
|
|
return float(value)
|
|
except Exception:
|
|
return default
|
|
|
|
|
|
def _clamp(value: float, low: float, high: float) -> float:
|
|
return max(low, min(high, value))
|
|
|
|
|
|
@dataclass
|
|
class FactorScorer:
|
|
"""Apply review-adjusted factor weights and keep an audit trail."""
|
|
|
|
weights: dict[str, Any] = field(default_factory=dict)
|
|
breakdown: list[dict[str, Any]] = field(default_factory=list)
|
|
group_totals: dict[str, float] = field(default_factory=dict)
|
|
|
|
@classmethod
|
|
def from_runtime(cls) -> "FactorScorer":
|
|
try:
|
|
weights = get_signal_weights() or {}
|
|
except Exception:
|
|
weights = {}
|
|
return cls(weights=weights)
|
|
|
|
def _runtime_weight(self, code: str) -> float | None:
|
|
keys = [code, signal_label_for_code(code), *WEIGHT_ALIASES.get(code, ())]
|
|
for key in keys:
|
|
if key in self.weights:
|
|
value = self.weights[key]
|
|
if isinstance(value, dict):
|
|
value = value.get("weight")
|
|
return max(0.0, _safe_float(value))
|
|
return None
|
|
|
|
def _apply_group_cap(self, code: str, adjusted: float) -> tuple[float, str, float, float]:
|
|
group = FACTOR_GROUPS.get(code, "other")
|
|
cap = _safe_float(GROUP_CAPS.get(group), 99.0)
|
|
if cap <= 0:
|
|
return adjusted, group, cap, 0.0
|
|
used = abs(_safe_float(self.group_totals.get(group)))
|
|
remaining = max(0.0, cap - used)
|
|
capped = max(-remaining, min(remaining, adjusted))
|
|
self.group_totals[group] = round(_safe_float(self.group_totals.get(group)) + capped, 6)
|
|
return capped, group, cap, remaining
|
|
|
|
def delta(self, code: str, base: float, *, evidence: str = "", value: Any = None) -> float:
|
|
"""Return the review-aware score delta for one factor.
|
|
|
|
``base`` remains the strategy designer's default score. Once reviews
|
|
have enough samples, ``signal_performance.weight`` becomes a multiplier
|
|
against the factor's canonical baseline. A reviewed weight of 0 means
|
|
the factor is effectively淘汰 for scoring.
|
|
"""
|
|
base = _safe_float(base)
|
|
runtime_weight = self._runtime_weight(code)
|
|
baseline = max(0.1, _safe_float(DEFAULT_FACTOR_WEIGHTS.get(code), abs(base) or 1.0))
|
|
if runtime_weight is None:
|
|
adjusted = base
|
|
multiplier = 1.0
|
|
source = "base"
|
|
else:
|
|
multiplier = _clamp(runtime_weight / baseline, 0.0, 1.8)
|
|
adjusted = base * multiplier
|
|
source = "review_weight"
|
|
adjusted, group, group_cap, group_remaining = self._apply_group_cap(code, adjusted)
|
|
adjusted = round(adjusted, 3)
|
|
self.breakdown.append(
|
|
{
|
|
"factor_code": code,
|
|
"factor_name": signal_label_for_code(code),
|
|
"factor_group": group,
|
|
"base_delta": base,
|
|
"score_delta": adjusted,
|
|
"runtime_weight": runtime_weight,
|
|
"baseline_weight": baseline,
|
|
"multiplier": round(multiplier, 4),
|
|
"group_cap": group_cap,
|
|
"group_remaining_before": round(group_remaining, 3),
|
|
"source": source,
|
|
"evidence": evidence,
|
|
"value": value,
|
|
}
|
|
)
|
|
return adjusted
|
|
|
|
def add_existing(self, code: str, observed_score: float, *, evidence: str = "", value: Any = None, cap: float | None = None) -> float:
|
|
base = _safe_float(observed_score)
|
|
if cap is not None:
|
|
base = min(base, _safe_float(cap, base))
|
|
return self.delta(code, base, evidence=evidence, value=value)
|
|
|
|
def summary(self) -> dict[str, Any]:
|
|
total = round(sum(_safe_float(item.get("score_delta")) for item in self.breakdown), 3)
|
|
groups = {}
|
|
for item in self.breakdown:
|
|
group = item.get("factor_group") or "other"
|
|
bucket = groups.setdefault(group, {"score_delta": 0.0, "items": 0})
|
|
bucket["score_delta"] = round(bucket["score_delta"] + _safe_float(item.get("score_delta")), 3)
|
|
bucket["items"] += 1
|
|
opportunity_groups = {"momentum", "participation", "structure", "positioning", "narrative", "onchain_flow"}
|
|
opportunity_score = round(sum(_safe_float(v.get("score_delta")) for k, v in groups.items() if k in opportunity_groups), 3)
|
|
entry_score = round(_safe_float(groups.get("entry_quality", {}).get("score_delta")), 3)
|
|
risk_score = round(abs(min(0.0, _safe_float(groups.get("risk", {}).get("score_delta")))), 3)
|
|
return {
|
|
"total_delta": total,
|
|
"opportunity_score": opportunity_score,
|
|
"entry_score": entry_score,
|
|
"risk_score": risk_score,
|
|
"groups": groups,
|
|
"group_caps": GROUP_CAPS,
|
|
"items": self.breakdown,
|
|
}
|