Using new gnosis dataset and updating DAA graph

app.py

@@ -26,7 +26,6 @@ from tabs.trader_plots import (
     get_interpretation_text,
     plot_total_bet_amount,
     plot_active_traders,
-    plot_rolling_average,
     plot_rolling_average_roi,
 )
 from tabs.daily_graphs import (
@@ -34,6 +33,7 @@ from tabs.daily_graphs import (
     plot_daily_metrics,
     trader_daily_metric_choices,
     default_daily_metric,
+    plot_rolling_average,
 )
 from scripts.utils import get_traders_family
 from tabs.market_plots import (
@@ -136,7 +136,15 @@ def load_all_data():
         repo_type="dataset",
     )
     df7 = pd.read_parquet(all_mech_calls_df)
-    return df1, df2, df3, df4, df5, df6, df7
+
+    # gnosis_txs_dataset.parquet
+    gnosis_txs_df = hf_hub_download(
+        repo_id="valory/Olas-predict-dataset",
+        filename="gnosis_txs_dataset.parquet",
+        repo_type="dataset",
+    )
+    df8 = pd.read_parquet(gnosis_txs_df)
+    return df1, df2, df3, df4, df5, df6, df7, df8
 
 
 def prepare_data():
@@ -149,8 +157,12 @@ def prepare_data():
         retention_df,
         active_traders,
         all_mech_calls,
+        gnosis_txs_df,
     ) = load_all_data()
-
+    all_trades["creation_timestamp"] = all_trades["creation_timestamp"].dt.tz_convert(
+        "UTC"
+    )
+    all_trades = all_trades.sort_values(by="creation_timestamp", ascending=True)
     all_trades["creation_date"] = all_trades["creation_timestamp"].dt.date
 
     # nr-trades variable
@@ -191,6 +203,41 @@ def prepare_data():
         .dt.to_period("W")
         .dt.start_time.dt.strftime("%b-%d-%Y")
     )
+
+    # prepare daa dataset concatenating all_trades and gnosis_txs_df
+    selected_trades = all_trades[all_trades["staking"] != "non_Olas"].copy(deep=True)
+    selected_trades = selected_trades[
+        [
+            "market_creator",
+            "trader_address",
+            "creation_timestamp",
+        ]
+    ]
+    selected_trades.drop_duplicates(inplace=True)
+    selected_trades["tx_date"] = pd.to_datetime(
+        selected_trades["creation_timestamp"].dt.date, format="%b-%d-%Y"
+    )
+
+    gnosis_txs_df["tx_date"] = pd.to_datetime(gnosis_txs_df["tx_datetime"]).dt.date
+    gnosis_txs_df = gnosis_txs_df[
+        [
+            "trader_address",
+            "tx_date",
+            "market_creator",
+        ]
+    ]
+    gnosis_txs_df["tx_date"] = pd.to_datetime(
+        gnosis_txs_df["tx_date"], format="%b-%d-%Y"
+    )
+    selected_trades = selected_trades[
+        [
+            "trader_address",
+            "tx_date",
+            "market_creator",
+        ]
+    ]
+    daa_df = pd.concat([selected_trades, gnosis_txs_df], ignore_index=True)
+    daa_df.drop_duplicates(inplace=True)
     return (
         traders_data,
         closed_markets,
@@ -199,6 +246,7 @@ def prepare_data():
         retention_df,
         active_traders,
         all_mech_calls,
+        daa_df,
     )
 
 
@@ -210,6 +258,7 @@ def prepare_data():
     raw_retention_df,
     active_traders,
     all_mech_calls,
+    daa_df,
 ) = prepare_data()
 
 retention_df = prepare_retention_dataset(
@@ -642,14 +691,14 @@ with demo:
                 gr.Markdown(" # Daily active agents in Pearl markets")
             with gr.Row():
                 rolling_avg_plot = plot_rolling_average(
-                    active_traders,
+                    daa_df,
                     market_creator="pearl",
                 )
             with gr.Row():
                 gr.Markdown(" # Daily active agents in QS markets")
             with gr.Row():
                 rolling_avg_plot = plot_rolling_average(
-                    active_traders,
+                    daa_df,
                     market_creator="quickstart",
                 )
 

tabs/daily_graphs.py

@@ -229,3 +229,55 @@ def plot_daily_metrics_v2(
 
     # Update y-axes to have the same range
     fig.update_yaxes(matches="y")
+
+
+def get_sevenday_rolling_average(daa_df: pd.DataFrame) -> pd.DataFrame:
+    """Function to get the 7-day rolling average of the number of unique
+    trader_address"""
+    # Create a local copy of the dataframe
+    local_df = daa_df.copy()
+
+    # Sort the dataframe by date
+    local_df = local_df.sort_values(by="tx_date").set_index("tx_date")
+
+    # Group by market_creator and calculate rolling average of unique trader_address
+    rolling_avg = (
+        local_df.resample("D")["trader_address"]
+        .nunique()
+        .rolling(window=7)
+        .mean()
+        .reset_index()
+    )
+    rolling_avg.rename(columns={"trader_address": "rolling_avg_traders"}, inplace=True)
+    return rolling_avg
+
+
+def plot_rolling_average(
+    daa_df: pd.DataFrame,
+    market_creator: str = None,
+) -> gr.Plot:
+    """Function to plot the rolling average of daily active traders by markets"""
+    if market_creator is not None:
+        filtered_traders_df = daa_df.loc[daa_df["market_creator"] == market_creator]
+        rolling_avg_df = get_sevenday_rolling_average(filtered_traders_df)
+    else:
+        rolling_avg_df = get_sevenday_rolling_average(daa_df)
+    print(rolling_avg_df.head())
+
+    # Ensure 'creation_date' is a column, not an index
+    if "creation_date" not in rolling_avg_df.columns:
+        rolling_avg_df = rolling_avg_df.reset_index()
+
+    fig = px.bar(
+        rolling_avg_df,
+        x="tx_date",
+        y="rolling_avg_traders",
+    )
+    fig.update_layout(
+        xaxis_title="Date",
+        yaxis_title="7-day rolling average of DAA",
+    )
+
+    return gr.Plot(
+        value=fig,
+    )

tabs/trader_plots.py

@@ -446,69 +446,6 @@ def plot_rolling_average_roi(
     )
 
 
-def get_sevenday_rolling_average(active_traders_df: pd.DataFrame) -> pd.DataFrame:
-    """Function to get the 7-day rolling average of the number of unique
-    trader_address removing the unknown traders"""
-    # Create a local copy of the dataframe
-    local_df = active_traders_df.copy()
-
-    # Remove unknown traders
-    local_df = local_df.loc[local_df["trader_type"] != "unknown"]
-
-    # Convert string dates to datetime
-    local_df["creation_date"] = pd.to_datetime(
-        local_df["creation_date"], format="%b-%d-%Y"
-    )
-    # Sort the dataframe by date
-    local_df = local_df.sort_values(by="creation_date").set_index("creation_date")
-
-    # Group by market_creator and calculate rolling average of unique trader_address
-    rolling_avg_by_market = (
-        local_df.resample("D")["trader_address"]
-        .nunique()
-        .rolling(window=7)
-        .mean()
-        .reset_index()
-    )
-    rolling_avg_by_market.rename(
-        columns={"trader_address": "rolling_avg_traders"}, inplace=True
-    )
-    return rolling_avg_by_market
-
-
-def plot_rolling_average(
-    active_traders_df: pd.DataFrame,
-    market_creator: str = None,
-) -> gr.Plot:
-    """Function to plot the rolling average of active traders for the different categories and markets"""
-    if market_creator is not None:
-        filtered_traders_df = active_traders_df.loc[
-            active_traders_df["market_creator"] == market_creator
-        ]
-        rolling_avg_df = get_sevenday_rolling_average(filtered_traders_df)
-    else:
-        rolling_avg_df = get_sevenday_rolling_average(active_traders_df)
-    print(rolling_avg_df.head())
-
-    # Ensure 'creation_date' is a column, not an index
-    if "creation_date" not in rolling_avg_df.columns:
-        rolling_avg_df = rolling_avg_df.reset_index()
-
-    fig = px.bar(
-        rolling_avg_df,
-        x="creation_date",
-        y="rolling_avg_traders",
-    )
-    fig.update_layout(
-        xaxis_title="Date",
-        yaxis_title="7-day rolling average of active traders",
-    )
-
-    return gr.Plot(
-        value=fig,
-    )
-
-
 def plot_active_traders(
     active_traders_data: pd.DataFrame,
     market_creator: str = None,