Adaptive-fee-tuning.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "38f5f740",
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import math\n",
    "import plotly.graph_objects as go\n",
    "from plotly.subplots import make_subplots\n",
    "pd.options.plotting.backend = \"plotly\"\n",
    "\n",
    "def init_fee_config():\n",
    "    return get_fee_config(\n",
    "        min_fee = 100, \n",
    "        medium_fee = 500, \n",
    "        max_fee = 3000, \n",
    "        beta1 = 4200, \n",
    "        beta2 = 90000, \n",
    "        gamma1 = 700, \n",
    "        gamma2 = 15000\n",
    "    )\n",
    "\n",
    "def get_fee_config(min_fee, medium_fee, max_fee, beta1, beta2, gamma1, gamma2):\n",
    "    config = {}\n",
    "    config['baseFee'] = min_fee\n",
    "    config['alpha1'] = medium_fee - config['baseFee']\n",
    "    config['alpha2'] = max_fee - config['alpha1'] - config['baseFee']\n",
    "    config['beta1'] = beta1\n",
    "    config['beta2'] = beta2\n",
    "    config['gamma1'] = gamma1\n",
    "    config['gamma2'] = gamma2\n",
    "    return config    \n",
    "\n",
    "def sigmoid(volatility, alpha, beta, gamma):\n",
    "    if (volatility > beta):\n",
    "        volatility = volatility - beta\n",
    "        if (volatility >= 6 * gamma):\n",
    "            return alpha\n",
    "        ex = math.exp(volatility / gamma) * gamma**4\n",
    "        return (alpha * ex) // (gamma**4 + ex)\n",
    "    else:\n",
    "        volatility = beta - volatility\n",
    "        if (volatility >= 6 * gamma):\n",
    "            return 0\n",
    "        ex = gamma**4 + math.exp(volatility / gamma) * gamma**4\n",
    "        return (alpha * gamma**4) // ex;\n",
    "        \n",
    "\n",
    "def getFee(volatility, config):\n",
    "    volatility = volatility // 15\n",
    "    result = config['baseFee']\n",
    "    result += sigmoid(volatility, config['alpha1'], config['beta1'], config['gamma1'])\n",
    "    result +=  sigmoid(volatility, config['alpha2'], config['beta2'], config['gamma2'])\n",
    "    return result"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "9801a2bb",
   "metadata": {},
   "outputs": [],
   "source": [
    "def plot_with_ticks(df, resample_rate='1h'):\n",
    "    df_resampled_1h = df.resample(resample_rate).mean()\n",
    "\n",
    "    fig = make_subplots(specs=[[{\"secondary_y\": True}]])\n",
    "    fig.add_trace(go.Scatter(x=df_resampled_1h.index, y=df_resampled_1h['tick'], name='Tick'))\n",
    "    fig.add_trace(go.Scatter(x=df_resampled_1h.index, y=df_resampled_1h['tickAverage'], name='Average tick, 24h'))\n",
    "    fig.add_trace(go.Scatter(x=df_resampled_1h.index, y=df_resampled_1h['fee'], name='Fee'), secondary_y=True)\n",
    "\n",
    "    fig.update_xaxes(title_text=\"Time\")\n",
    "    fig.update_yaxes(title_text=\"Tick value\", secondary_y=False)\n",
    "    fig.update_yaxes(title_text=\"Fee value\", secondary_y=True)\n",
    "    return fig\n",
    "    \n",
    "def plot_with_prices(df, decimals0, decimals1, inverse, resample_rate='1h'):\n",
    "    df_resampled_1h = df.resample(resample_rate).mean()\n",
    "    df_resampled_1h['price'] = (1.0001) ** df_resampled_1h['tick'] / (10**(decimals1 - decimals0))\n",
    "    df_resampled_1h['priceAverage'] = (1.0001) ** df_resampled_1h['tickAverage'] / (10**(decimals1 - decimals0))\n",
    "    if (inverse):\n",
    "        df_resampled_1h['price'] = 1 / df_resampled_1h['price']\n",
    "        df_resampled_1h['priceAverage'] = 1 / df_resampled_1h['priceAverage']\n",
    "    \n",
    "    fig = make_subplots(specs=[[{\"secondary_y\": True}]])\n",
    "    fig.add_trace(go.Scatter(x=df_resampled_1h.index, y=df_resampled_1h['price'], name='Price'))\n",
    "    fig.add_trace(go.Scatter(x=df_resampled_1h.index, y=df_resampled_1h['priceAverage'], name='Average price, 24h'))\n",
    "    fig.add_trace(go.Scatter(x=df_resampled_1h.index, y=df_resampled_1h['fee'], name='Fee'), secondary_y=True)\n",
    "\n",
    "    fig.update_xaxes(title_text=\"Time\")\n",
    "    fig.update_yaxes(title_text=\"Price value\", secondary_y=False, type=\"log\")\n",
    "    fig.update_yaxes(title_text=\"Fee value\", secondary_y=True)\n",
    "    return fig"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "aa0f6f9a",
   "metadata": {},
   "outputs": [],
   "source": [
    "df = pd.read_json('input/timepoints.json')\n",
    "df.set_index('timestamp', inplace=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "49f3cf24",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Make plot with ticks, average ticks and corresponding fee values (default config)\n",
    "plot_with_ticks(df)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "d775a46f",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Make plot with price, average prices and corresponding fee values (default config)\n",
    "# Values are for ETH/USDC pair on Polygon. You should change them if you use other data\n",
    "plot_with_prices(df, decimals0 = 6, decimals1 = 18, inverse = True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "961fb619",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Experimenting with fee configs\n",
    "\n",
    "feeConfig = get_fee_config(\n",
    "        min_fee = 100, \n",
    "        medium_fee = 3000, \n",
    "        max_fee = 15000, \n",
    "        beta1 = 360, \n",
    "        beta2 = 60000, \n",
    "        gamma1 = 59, \n",
    "        gamma2 = 8500\n",
    "    )\n",
    "\n",
    "feeConfig = get_fee_config(\n",
    "        min_fee = 100, \n",
    "        medium_fee = 3000, \n",
    "        max_fee = 5000, \n",
    "        beta1 = 403, \n",
    "        beta2 = 8574, \n",
    "        gamma1 = 67, \n",
    "        gamma2 = 570\n",
    "    )\n",
    "\n",
    "default_new_token_config = get_fee_config(\n",
    "        min_fee = 100, \n",
    "        medium_fee = 3000, \n",
    "        max_fee = 5000, \n",
    "        beta1 = 550, \n",
    "        beta2 = 8574, \n",
    "        gamma1 = 87, \n",
    "        gamma2 = 570\n",
    "    )\n",
    "\n",
    "df['fee_calculated'] = df.apply(lambda x: getFee(x['volatilityAverage'], default_new_token_config), axis=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "114d9ad9",
   "metadata": {},
   "outputs": [],
   "source": [
    "# plot new fee over price history\n",
    "resampling_rate = '15min'\n",
    "df_calculated = df.copy()\n",
    "df_resampled = df_calculated.resample(resampling_rate).mean()\n",
    "fig = plot_with_prices(df_calculated, decimals0 = 6, decimals1 = 18, inverse = True, resample_rate=resampling_rate)\n",
    "fig.add_trace(go.Scatter(x=df_resampled.index, y=df_resampled['fee_calculated'], name='Fee (new)'), secondary_y=True)\n",
    "fig.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "f6b3ff7e",
   "metadata": {},
   "outputs": [],
   "source": [
    "# plot difference between default and new fee\n",
    "\n",
    "df_resampled_1h = df.resample('1h').mean()\n",
    "\n",
    "fig = go.Figure()\n",
    "fig.add_trace(go.Scatter(x=df_resampled_1h.index, y=df_resampled_1h['fee'], name='Fee'))\n",
    "fig.add_trace(go.Scatter(x=df_resampled_1h.index, y=df_resampled_1h['fee_calculated'], name='Fee calculated'))\n",
    "\n",
    "fig.update_xaxes(title_text=\"Time\")\n",
    "fig.update_yaxes(title_text=\"Fee value\")\n",
    "fig.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "5325558e",
   "metadata": {},
   "outputs": [],
   "source": [
    "# plot new fee over price history with heatmap\n",
    "import plotly.express as px\n",
    "\n",
    "def plot_with_heatmap(value_field_name = 'price'):\n",
    "    resampling_rate = '1h'\n",
    "    df_calculated = df.copy()\n",
    "    df_resampled = df_calculated.resample(resampling_rate).mean()\n",
    "\n",
    "    decimals0 = 6\n",
    "    decimals1 = 18\n",
    "    inverse = True\n",
    "\n",
    "    df_resampled['price'] = (1.0001) ** df_resampled['tick'] / (10**(decimals1 - decimals0))\n",
    "    df_resampled['priceAverage'] = (1.0001) ** df_resampled['tickAverage'] / (10**(decimals1 - decimals0))\n",
    "    if (inverse):\n",
    "        df_resampled['price'] = 1 / df_resampled['price']\n",
    "        df_resampled['priceAverage'] = 1 / df_resampled['priceAverage']\n",
    "\n",
    "    fig = go.Figure()\n",
    "    fig.add_trace(go.Scatter(\n",
    "        x=df_resampled.index, \n",
    "        y=df_resampled[value_field_name], \n",
    "        name=value_field_name,\n",
    "        )\n",
    "    )\n",
    "\n",
    "    df_resampled['medium_fee_price'] = df_resampled[df_resampled['fee_calculated'] >= 1500][value_field_name]\n",
    "    df_resampled['high_fee_price'] = df_resampled[df_resampled['fee_calculated'] >= 3200][value_field_name]\n",
    "    df_resampled['low_fee_price'] = df_resampled[df_resampled['fee_calculated'] <= 1500][value_field_name]\n",
    "\n",
    "    df_resampled.fillna(0, inplace= True)\n",
    "\n",
    "    fig.add_traces(go.Scatter(x=df_resampled.index, y = df_resampled['medium_fee_price'],\n",
    "                                  line = dict(color='rgba(255, 255, 120, 1)'), line_width = 0,\n",
    "                                  fill='tozeroy', \n",
    "                                  fillcolor = 'rgba(255, 255, 120, 1)', name='Medium fee'))\n",
    "\n",
    "    fig.add_traces(go.Scatter(x=df_resampled.index, y = df_resampled['low_fee_price'],\n",
    "                                  line = dict(color='rgba(120, 255, 120, 1)'), line_width = 0,\n",
    "                                  fill='tozeroy', \n",
    "                                  fillcolor = 'rgba(120, 255, 120, 1)', name='Low fee'))\n",
    "\n",
    "    fig.add_traces(go.Scatter(x=df_resampled.index, y = df_resampled['high_fee_price'],\n",
    "                                  line = dict(color='rgba(255, 120, 120, 1)'), line_width = 0,\n",
    "                                  fill='tozeroy', \n",
    "                                  fillcolor = 'rgba(255, 120, 120, 1)', name='High fee'))\n",
    "\n",
    "    fig.update_xaxes(title_text=\"Time\")\n",
    "    fig.update_yaxes(title_text=\"Price value\", type=\"log\")\n",
    "    \n",
    "    return fig"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "1930bb75",
   "metadata": {},
   "outputs": [],
   "source": [
    "plot_with_heatmap(value_field_name = 'price')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "97a97e78",
   "metadata": {},
   "outputs": [],
   "source": [
    "plot_with_heatmap(value_field_name = 'tick')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "fbcaa87f",
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.9.13"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}