def stats_update():
upnl_l = x if (x := calc_long_pnl(long_pprice, tick[0], long_psize, xk['inverse'],
xk['contract_multiplier'])) == x else 0.0
upnl_s = y if (y := calc_shrt_pnl(shrt_pprice, tick[0], shrt_psize, xk['inverse'],
xk['contract_multiplier'])) == y else 0.0
stats.append({'timestamp': tick[2],
'balance': balance, # Redundant with fills, but makes plotting easier
'equity': balance + upnl_l + upnl_s})
async def fetch_position(self) -> dict:
position = {}
if self.market_type == 'linear_perpetual':
fetched, bal = await asyncio.gather(
self.private_get(self.endpoints['position'], {'symbol': self.symbol}),
self.private_get(self.endpoints['balance'], {'coin': self.quot})
)
long_pos = [e for e in fetched['result'] if e['side'] == 'Buy'][0]
shrt_pos = [e for e in fetched['result'] if e['side'] == 'Sell'][0]
position['wallet_balance'] = float(bal['result'][self.quot]['wallet_balance'])
else:
fetched, bal = await asyncio.gather(
self.private_get(self.endpoints['position'], {'symbol': self.symbol}),
self.private_get(self.endpoints['balance'], {'coin': self.coin})
)
position['wallet_balance'] = float(bal['result'][self.coin]['wallet_balance'])
if self.market_type == 'inverse_perpetual':
if fetched['result']['side'] == 'Buy':
long_pos = fetched['result']
shrt_pos = {'size': 0.0, 'entry_price': 0.0, 'leverage': 0.0, 'liq_price': 0.0}
else:
long_pos = {'size': 0.0, 'entry_price': 0.0, 'leverage': 0.0, 'liq_price': 0.0}
shrt_pos = fetched['result']
elif self.market_type == 'inverse_futures':
long_pos = [e['data'] for e in fetched['result'] if e['data']['position_idx'] == 1][0]
shrt_pos = [e['data'] for e in fetched['result'] if e['data']['position_idx'] == 2][0]
position['long'] = {'size': float(long_pos['size']),
'price': float(long_pos['entry_price']),
'leverage': float(long_pos['leverage']),
'liquidation_price': float(long_pos['liq_price'])}
position['shrt'] = {'size': -float(shrt_pos['size']),
'price': float(shrt_pos['entry_price']),
'leverage': float(shrt_pos['leverage']),
'liquidation_price': float(shrt_pos['liq_price'])}
position['long']['upnl'] = calc_long_pnl(position['long']['price'], self.price,
position['long']['size'], self.xk['inverse'],
self.xk['contract_multiplier']) \
if position['long']['price'] != 0.0 else 0.0
position['shrt']['upnl'] = calc_shrt_pnl(position['shrt']['price'], self.price,
position['shrt']['size'], self.xk['inverse'],
self.xk['contract_multiplier']) \
if position['shrt']['price'] != 0.0 else 0.0
upnl = position['long']['upnl'] + position['shrt']['upnl']
position['equity'] = position['wallet_balance'] + upnl
return position
def backtest(config: dict,
ticks: np.ndarray,
do_print=False) -> (list, list, bool):
if len(ticks) <= config['ema_span']:
return [], [], False
long_psize, long_pprice = 0.0, 0.0
shrt_psize, shrt_pprice = 0.0, 0.0
liq_price, liq_diff = 0.0, 1.0
balance = config['starting_balance']
if all(x in config for x in
['long_pprice', 'long_psize', 'shrt_pprice', 'shrt_psize']):
long_pprice, long_psize, shrt_pprice, shrt_psize = (
config["long_pprice"],
config["long_psize"],
config["shrt_pprice"],
config["shrt_psize"],
)
else:
long_pprice, long_psize, shrt_pprice, shrt_psize = 0.0, 0.0, 0.0, 0.0
pnl_plus_fees_cumsum, loss_cumsum, profit_cumsum, fee_paid_cumsum = 0.0, 0.0, 0.0, 0.0
xk = {k: float(config[k]) for k in get_keys()}
if config['exchange'] == 'binance':
calc_liq_price = calc_liq_price_binance
elif config['exchange'] == 'bybit':
calc_liq_price = calc_liq_price_bybit
prev_long_close_ts, prev_long_entry_ts, prev_long_close_price = 0, 0, 0.0
prev_shrt_close_ts, prev_shrt_entry_ts, prev_shrt_close_price = 0, 0, 0.0
latency_simulation_ms = config['latency_simulation_ms'] \
if 'latency_simulation_ms' in config else 1000
next_stats_update = 0
stats = []
def stats_update():
upnl_l = x if (x := calc_long_pnl(
long_pprice, tick[0], long_psize, xk['inverse'],
xk['contract_multiplier'])) == x else 0.0
upnl_s = y if (y := calc_shrt_pnl(
shrt_pprice, tick[0], shrt_psize, xk['inverse'],
xk['contract_multiplier'])) == y else 0.0
stats.append({
'timestamp': tick[2],
'balance':
balance, # Redundant with fills, but makes plotting easier
'equity': balance + upnl_l + upnl_s
})
all_fills = []
fills = []
bids, asks = [], []
ob = [min(ticks[0][0], ticks[1][0]), max(ticks[0][0], ticks[1][0])]
ema_span = int(round(config['ema_span']))
# emas = calc_emas(ticks[:, 0], ema_span)
# price_stds = calc_stds(ticks[:, 0], ema_span)
# volatilities = price_stds / emas
ema_std_iterator = iter_indicator_chunks(ticks[:, 0], ema_span)
ema_chunk, std_chunk, z = next(ema_std_iterator)
volatility_chunk = std_chunk / ema_chunk
zc = 0
closest_liq = 1.0
prev_update_plus_delay = ticks[ema_span][2] + latency_simulation_ms
update_triggered = False
prev_update_plus_5sec = 0
tick = ticks[0]
stats_update()
# tick tuple: (price, buyer_maker, timestamp)
for k, tick in enumerate(ticks[ema_span:], start=ema_span):
chunk_i = k - zc
if chunk_i >= len(ema_chunk):
ema_chunk, std_chunk, z = next(ema_std_iterator)
volatility_chunk = std_chunk / ema_chunk
zc = z * len(ema_chunk)
chunk_i = k - zc
# Update the stats every 1/2 hour
if tick[2] > next_stats_update:
closest_liq = min(closest_liq, calc_diff(liq_price, tick[0]))
stats_update()
next_stats_update = tick[2] + 1000 * 60 * 30
fills = []
if tick[1]:
if liq_diff < 0.05 and long_psize > -shrt_psize and tick[
0] <= liq_price:
fills.append({
'qty':
-long_psize,
'price':
tick[0],
'pside':
'long',
'type':
'long_liquidation',
'side':
'sel',
'pnl':
calc_long_pnl(long_pprice, tick[0], long_psize,
xk['inverse'], xk['contract_multiplier']),
'fee_paid':
-calc_cost(long_psize, tick[0], xk['inverse'],
xk['contract_multiplier']) *
config['taker_fee'],
'long_psize':
0.0,
'long_pprice':
0.0,
'shrt_psize':
0.0,
'shrt_pprice':
0.0,
'liq_price':
0.0,
'liq_diff':
1.0
})
long_psize, long_pprice, shrt_psize, shrt_pprice = 0.0, 0.0, 0.0, 0.0
else:
if bids:
if tick[0] <= bids[0][1]:
update_triggered = True
while bids:
if tick[0] < bids[0][1]:
bid = bids.pop(0)
fill = {
'qty':
bid[0],
'price':
bid[1],
'side':
'buy',
'type':
bid[4],
'fee_paid':
-calc_cost(bid[0], bid[1], xk['inverse'],
xk['contract_multiplier']) *
config['maker_fee']
}
if 'close' in bid[4]:
fill['pnl'] = calc_shrt_pnl(
shrt_pprice, bid[1], bid[0], xk['inverse'],
xk['contract_multiplier'])
shrt_psize = min(
0.0,
round_(shrt_psize + bid[0],
config['qty_step']))
fill.update({
'pside': 'shrt',
'long_psize': long_psize,
'long_pprice': long_pprice,
'shrt_psize': shrt_psize,
'shrt_pprice': shrt_pprice
})
prev_shrt_close_ts = tick[2]
else:
fill['pnl'] = 0.0
long_psize, long_pprice = calc_new_psize_pprice(
long_psize, long_pprice, bid[0], bid[1],
xk['qty_step'])
if long_psize < 0.0:
long_psize, long_pprice = 0.0, 0.0
fill.update({
'pside': 'long',
'long_psize': bid[2],
'long_pprice': bid[3],
'shrt_psize': shrt_psize,
'shrt_pprice': shrt_pprice
})
prev_long_entry_ts = tick[2]
fills.append(fill)
else:
break
ob[0] = tick[0]
else:
if liq_diff < 0.05 and -shrt_psize > long_psize and tick[
0] >= liq_price:
fills.append({
'qty':
-shrt_psize,
'price':
tick[0],
'pside':
'shrt',
'type':
'shrt_liquidation',
'side':
'buy',
'pnl':
calc_shrt_pnl(shrt_pprice, tick[0], shrt_psize,
xk['inverse'], xk['contract_multiplier']),
'fee_paid':
-calc_cost(shrt_psize, tick[0], xk['inverse'],
xk['contract_multiplier']) *
config['taker_fee'],
'long_psize':
0.0,
'long_pprice':
0.0,
'shrt_psize':
0.0,
'shrt_pprice':
0.0,
'liq_price':
0.0,
'liq_diff':
1.0
})
long_psize, long_pprice, shrt_psize, shrt_pprice = 0.0, 0.0, 0.0, 0.0
else:
if asks:
if tick[0] >= asks[0][1]:
update_triggered = True
while asks:
if tick[0] > asks[0][1]:
ask = asks.pop(0)
fill = {
'qty':
ask[0],
'price':
ask[1],
'side':
'sel',
'type':
ask[4],
'fee_paid':
-calc_cost(ask[0], ask[1], xk['inverse'],
xk['contract_multiplier']) *
config['maker_fee']
}
if 'close' in ask[4]:
fill['pnl'] = calc_long_pnl(
long_pprice, ask[1], ask[0], xk['inverse'],
xk['contract_multiplier'])
long_psize = max(
0.0,
round_(long_psize + ask[0],
config['qty_step']))
fill.update({
'pside': 'long',
'long_psize': long_psize,
'long_pprice': long_pprice,
'shrt_psize': shrt_psize,
'shrt_pprice': shrt_pprice
})
prev_long_close_ts = tick[2]
else:
fill['pnl'] = 0.0
shrt_psize, shrt_pprice = calc_new_psize_pprice(
shrt_psize, shrt_pprice, ask[0], ask[1],
xk['qty_step'])
if shrt_psize > 0.0:
shrt_psize, shrt_pprice = 0.0, 0.0
fill.update({
'pside': 'shrt',
'long_psize': long_psize,
'long_pprice': long_pprice,
'shrt_psize': shrt_psize,
'shrt_pprice': shrt_pprice
})
prev_shrt_entry_ts = tick[2]
liq_diff = calc_diff(liq_price, tick[0])
fill.update({
'liq_price': liq_price,
'liq_diff': liq_diff
})
fills.append(fill)
else:
break
ob[1] = tick[0]
if tick[2] > prev_update_plus_delay and (
update_triggered or tick[2] > prev_update_plus_5sec):
prev_update_plus_delay = tick[2] + latency_simulation_ms
prev_update_plus_5sec = tick[2] + 5000
update_triggered = False
bids, asks = [], []
liq_diff = calc_diff(liq_price, tick[0])
closest_liq = min(closest_liq, liq_diff)
for tpl in iter_entries(balance, long_psize, long_pprice,
shrt_psize, shrt_pprice, liq_price, ob[0],
ob[1], ema_chunk[k - zc], tick[0],
volatility_chunk[k - zc], **xk):
if len(bids) > 2 and len(asks) > 2:
break
if tpl[0] > 0.0:
bids.append(tpl)
elif tpl[0] < 0.0:
asks.append(tpl)
else:
break
if tick[0] <= shrt_pprice and shrt_pprice > 0.0:
for tpl in iter_shrt_closes(balance, shrt_psize, shrt_pprice,
ob[0], **xk):
bids.append(list(tpl) + [shrt_pprice, 'shrt_close'])
if tick[0] >= long_pprice and long_pprice > 0.0:
for tpl in iter_long_closes(balance, long_psize, long_pprice,
ob[1], **xk):
asks.append(list(tpl) + [long_pprice, 'long_close'])
bids = sorted(bids, key=lambda x: x[1], reverse=True)
asks = sorted(asks, key=lambda x: x[1])
if len(fills) > 0:
for fill in fills:
balance += fill['pnl'] + fill['fee_paid']
upnl_l = calc_long_pnl(long_pprice, tick[0], long_psize,
xk['inverse'],
xk['contract_multiplier'])
upnl_s = calc_shrt_pnl(shrt_pprice, tick[0], shrt_psize,
xk['inverse'],
xk['contract_multiplier'])
liq_price = calc_liq_price(balance, long_psize, long_pprice,
shrt_psize, shrt_pprice,
xk['inverse'],
xk['contract_multiplier'],
config['max_leverage'])
liq_diff = calc_diff(liq_price, tick[0])
fill.update({'liq_price': liq_price, 'liq_diff': liq_diff})
fill['equity'] = balance + upnl_l + upnl_s
fill['available_margin'] = calc_available_margin(
balance, long_psize, long_pprice, shrt_psize, shrt_pprice,
tick[0], xk['inverse'], xk['contract_multiplier'],
xk['leverage'])
for side_ in ['long', 'shrt']:
if fill[f'{side_}_pprice'] == 0.0:
fill[f'{side_}_pprice'] = np.nan
fill['balance'] = balance
fill['timestamp'] = tick[2]
fill['trade_id'] = k
fill['gain'] = fill['equity'] / config['starting_balance']
fill['n_days'] = (tick[2] - ticks[ema_span][2]) / (1000 * 60 *
60 * 24)
fill['closest_liq'] = closest_liq
try:
fill['average_daily_gain'] = fill['gain'] ** (1 / fill['n_days']) \
if (fill['n_days'] > 0.5 and fill['gain'] > 0.0) else 0.0
except:
fill['average_daily_gain'] = 0.0
all_fills.append(fill)
if balance <= 0.0 or 'liquidation' in fill['type']:
return all_fills, stats, False
if do_print:
line = f"\r{k / len(ticks):.3f} "
line += f"adg {all_fills[-1]['average_daily_gain']:.4f} "
line += f"closest_liq {closest_liq:.4f} "
print(line, end=' ')
tick = ticks[-1]
stats_update()
return all_fills, stats, True