Fix backtest (#719)

* modify FileStorage to support multiple freqs

* modify backtest's sample documentation

* change the logging level of read data exception from error to debug

* fix the backtest exception when volume is 0 or np.nan

* fix test_storage.py

* add backtest_daily

* modify backtest_daily's docstring

* add __repr__/__str__ to Position

* fix the bug of nested_decision_execution example

Co-authored-by: Young <afe.young@gmail.com>
Co-authored-by: you-n-g <you-n-g@users.noreply.github.com>

qlib/backtest/__init__.py

@@ -186,8 +186,10 @@ def get_strategy_executor(
     trade_exchange = get_exchange(**exchange_kwargs)
 
     common_infra = CommonInfrastructure(trade_account=trade_account, trade_exchange=trade_exchange)
-    trade_strategy = init_instance_by_config(strategy, accept_types=BaseStrategy, common_infra=common_infra)
-    trade_executor = init_instance_by_config(executor, accept_types=BaseExecutor, common_infra=common_infra)
+    trade_strategy = init_instance_by_config(strategy, accept_types=BaseStrategy)
+    trade_strategy.reset_common_infra(common_infra)
+    trade_executor = init_instance_by_config(executor, accept_types=BaseExecutor)
+    trade_executor.reset_common_infra(common_infra)
 
     return trade_strategy, trade_executor
 

qlib/backtest/account.py

@@ -29,7 +29,10 @@ rtn & earning in the Account
 
 
 class AccumulatedInfo:
-    """accumulated trading info, including accumulated return/cost/turnover"""
+    """
+    accumulated trading info, including accumulated return/cost/turnover
+    AccumulatedInfo should be shared accross different levels
+    """
 
     def __init__(self):
         self.reset()
@@ -62,6 +65,11 @@ class AccumulatedInfo:
 
 
 class Account:
+    """
+    The correctness of the metrics of Account in nested execution depends on the shallow copy of `trade_account` in qlib/backtest/executor.py:NestedExecutor
+    Different level of executor has different Account object when calculating metrics. But the position object is shared cross all the Account object.
+    """
+
     def __init__(
         self,
         init_cash: float = 1e9,
@@ -95,6 +103,8 @@ class Account:
         self.init_vars(init_cash, position_dict, freq, benchmark_config)
 
     def init_vars(self, init_cash, position_dict, freq: str, benchmark_config: dict):
+        # 1) the following variables are shared by multiple layers
+        # - you will see a shallow copy instead of deepcopy in the NestedExecutor;
         self.init_cash = init_cash
         self.current_position: BasePosition = init_instance_by_config(
             {
@@ -106,6 +116,9 @@ class Account:
                 "module_path": "qlib.backtest.position",
             }
         )
+        self.accum_info = AccumulatedInfo()
+
+        # 2) following variables are not shared between layers
         self.portfolio_metrics = None
         self.hist_positions = {}
         self.reset(freq=freq, benchmark_config=benchmark_config)
@@ -119,7 +132,8 @@ class Account:
     def reset_report(self, freq, benchmark_config):
         # portfolio related metrics
         if self.is_port_metr_enabled():
-            self.accum_info = AccumulatedInfo()
+            # NOTE:
+            # `accum_info` and `current_position` are shared here
             self.portfolio_metrics = PortfolioMetrics(freq, benchmark_config)
             self.hist_positions = {}
 

qlib/backtest/exchange.py

@@ -231,7 +231,7 @@ class Exchange:
                 self.extra_quote["limit_buy"] = False
                 self.logger.warning("No limit_buy set for extra_quote. All stock will be able to be bought.")
             assert set(self.extra_quote.columns) == set(self.quote_df.columns) - {"$change"}
-            self.quote_df = pd.concat([self.quote_df, extra_quote], sort=False, axis=0)
+            self.quote_df = pd.concat([self.quote_df, self.extra_quote], sort=False, axis=0)
 
     LT_TP_EXP = "(exp)"  # Tuple[str, str]
     LT_FLT = "float"  # float
@@ -736,7 +736,11 @@ class Exchange:
 
         # TODO: the adjusted cost ratio can be overestimated as deal_amount will be clipped in the next steps
         trade_val = order.deal_amount * trade_price
-        adj_cost_ratio = self.impact_cost * (trade_val / total_trade_val) ** 2
+        if not total_trade_val or np.isnan(total_trade_val):
+            # TODO: assert trade_val == 0, f"trade_val != 0, total_trade_val: {total_trade_val}; order info: {order}"
+            adj_cost_ratio = self.impact_cost
+        else:
+            adj_cost_ratio = self.impact_cost * (trade_val / total_trade_val) ** 2
 
         if order.direction == Order.SELL:
             cost_ratio = self.close_cost + adj_cost_ratio

qlib/backtest/executor.py

@@ -130,7 +130,7 @@ class BaseExecutor:
 
         if common_infra.has("trade_account"):
             # NOTE: there is a trick in the code.
-            # copy is used instead of deepcopy. So positions are shared
+            # shallow copy is used instead of deepcopy. So positions are shared
             self.trade_account: Account = copy.copy(common_infra.get("trade_account"))
             self.trade_account.reset(freq=self.time_per_step, port_metr_enabled=self.generate_portfolio_metrics)
 

qlib/backtest/position.py

@@ -223,6 +223,12 @@ class BasePosition:
         """
         raise NotImplementedError(f"Please implement the `settle_commit` method")
 
+    def __str__(self):
+        return self.__dict__.__str__()
+
+    def __repr__(self):
+        return self.__dict__.__repr__()
+
 
 class Position(BasePosition):
     """Position

qlib/backtest/utils.py

@@ -70,7 +70,7 @@ class TradeCalendarManager:
         - If self.trade_step >= self.self.trade_len, it means the trading is finished
         - If self.trade_step < self.self.trade_len, it means the number of trading step finished is self.trade_step
         """
-        return self.trade_step >= self.trade_len - 1
+        return self.trade_step >= self.trade_len
 
     def step(self):
         if self.finished():
@@ -222,7 +222,7 @@ class CommonInfrastructure(BaseInfrastructure):
 
 
 class LevelInfrastructure(BaseInfrastructure):
-    """level instrastructure is created by executor, and then shared to strategies on the same level"""
+    """level infrastructure is created by executor, and then shared to strategies on the same level"""
 
     def get_support_infra(self):
         """