Adds observation (orderbook,trades) and reward

Author: Adeleet

RL/observation.py

@@ -0,0 +1,39 @@
+import numpy as np
+import pandas as pd
+
+from stats_columns import stats_columns
+
+
+class Observation:
+    def __init__(self, client):
+        self.client = client
+
+    def new(self):
+        # stats = __ob_stats__(client)
+        orderbook = self.__ob_orderbook__()
+        trades = self.__ob_trades__()
+        return np.concatenate([trades, orderbook])
+
+    def __ob_orderbook__(self):
+        raw_orderbook = self.client.orderbook(
+            params={"symbol": "XBTUSD", "depth": "20"})
+        df_orderbook = pd.DataFrame(raw_orderbook).drop(
+            ["symbol", "id"], axis=1)
+        df_orderbook = df_orderbook.sort_values(by="side", ascending=False)
+        return df_orderbook[["price", "size"]].values.flatten()
+
+    def __ob_stats__(self):
+        return pd.DataFrame(self.client.instruments(
+            params={"symbol": "XBTUSD",
+                    "columns": ",".join(stats_columns())}))
+
+    def __ob_trades__(self):
+        df = pd.DataFrame(self.client.trades_bucketed(
+            params={"binSize": "1m",
+                    "symbol": "XBTUSD",
+                    "count": "2",
+                    "reverse": "true",
+                    "partial": "true"}))
+        df = df.select_dtypes(exclude=["object"])
+        self.last_vwap = df["vwap"][0]
+        return df.apply(lambda x: x * np.array([0.75, 0.25])).sum().values

RL/reward.py

@@ -0,0 +1,33 @@
+import pandas as pd
+
+
+class Reward:
+    def __init__(self, client):
+        self.client = client
+
+    def __last_o_book__(self, o_book, side=None):
+        if side:
+            o_book = o_book[o_book["side"] == side]
+        return o_book[o_book["timestamp"] == o_book["timestamp"].max()]
+
+    def get(self, prev_act, vwap):
+        orders = self.client.get_orders(params={"symbol": "XBTUSD",
+                                                "reverse": "true",
+                                                "count": "10"})
+
+        df_orders = pd.DataFrame(orders)[["cumQty",
+                                          "price",
+                                          "side",
+                                          "timestamp"]]
+        df_orders["timestamp"] = pd.to_datetime(df_orders["timestamp"])
+
+        self.last_order = self.__last_o_book__(df_orders)
+        self.last_buy = self.__last_o_book__(df_orders, "Buy").values[0]
+        self.last_sell = self.__last_o_book__(df_orders, "Sell").values[0]
+
+        if prev_act != self.last_order["side"].values[0] or prev_act == "Hold":
+            return 0
+        elif prev_act == "Buy":
+            return vwap - self.last_buy
+        elif prev_act == "Sell":
+            return self.last_sell - self.last_buy

RL/stats_columns.py

@@ -0,0 +1,68 @@
+def stats_columns():
+    return ['optionStrikePcnt',
+            'optionStrikeRound',
+            'optionStrikePrice',
+            'optionMultiplier',
+            'maxOrderQty',
+            'maxPrice',
+            'lotSize',
+            'tickSize',
+            'multiplier',
+            'underlyingToPositionMultiplier',
+            'underlyingToSettleMultiplier',
+            'quoteToSettleMultiplier',
+            'isQuanto',
+            'isInverse',
+            'initMargin',
+            'maintMargin',
+            'riskLimit',
+            'riskStep',
+            'limit',
+            'capped',
+            'taxed',
+            'deleverage',
+            'makerFee',
+            'takerFee',
+            'settlementFee',
+            'insuranceFee',
+            'fundingRate',
+            'indicativeFundingRate',
+            'prevClosePrice',
+            'limitDownPrice',
+            'limitUpPrice',
+            'bankruptLimitDownPrice',
+            'bankruptLimitUpPrice',
+            'prevTotalVolume',
+            'totalVolume',
+            'volume',
+            'volume24h',
+            'prevTotalTurnover',
+            'totalTurnover',
+            'turnover',
+            'turnover24h',
+            'homeNotional24h',
+            'foreignNotional24h',
+            'prevPrice24h',
+            'vwap',
+            'highPrice',
+            'lowPrice',
+            'lastPrice',
+            'lastPriceProtected',
+            'lastChangePcnt',
+            'bidPrice',
+            'midPrice',
+            'askPrice',
+            'impactBidPrice',
+            'impactMidPrice',
+            'impactAskPrice',
+            'hasLiquidity',
+            'openInterest',
+            'openValue',
+            'fairBasisRate',
+            'fairBasis',
+            'fairPrice',
+            'markPrice',
+            'indicativeTaxRate',
+            'indicativeSettlePrice',
+            'optionUnderlyingPrice',
+            'settledPrice']