def bid_exceeds_offer_before_drop(data, max_drop = .00001, min_profit=0.00005, max_time = 10000, network_delay=30):
ts = data.get_col('t')
bids = data.get_col('bid/100ms')
offers = data.get_col('offer/100ms')
signal = np.zeros(len(ts))
max_per_millisecond = analysis.density_per_millisecond(max_time-network_delay)
find_index = analysis.find_future_index
for idx in xrange(len(ts)):
t = ts[idx]
start_bid = bids[idx]
# average prices over the time window in which a buy message may reach
# the server
find_index(ts, idx, t+network_delay, max_per_millisecond=max_per_millisecond)
delayed_offers = offers[idx:last_delay_index]
avg_delayed_offer = np.mean(delayed_offers)
last_future_index = find_index(ts, idx, max_time, max_per_millisecond=max_per_millisecond)
future_bids = bids[last_delay_index:last_future_index]
sale_price_prct = (future_bids - avg_delayed_offer) / avg_delayed_offer
profit_index = find_first (sale_price_prct >= min_profit)
if profit_index:
neg_bid_prct = (future_bids - start_bid) / start_bid
quit_index = find_first( neg_bid_prct <= -max_drop)
if quit_index is None or quit_index > profit_index:
signal[idx] = 1
return signal
def min_bid_exceeds_offer_single_timeframe(data,
wait_time=1500,
hold_time=250,
target_prct=1.0001):
ts = data['t']
bids = data['bid/100ms']
offers = data['offer/100ms']
signal = np.zeros(len(ts))
max_per_millisecond = analysis.density_per_millisecond(hold_time)
last_time = wait_time + hold_time
find_index = analysis.find_future_index
for idx in xrange(len(ts)):
t = ts[idx]
start_offer = offers[idx]
start_hold_idx = find_index(ts,
idx,
t + wait_time,
max_per_millisecond=max_per_millisecond)
# prefilter
if bids[start_hold_idx - 1] > start_offer:
target_price = target_prct * start_offer
end_hold_idx = find_index(ts,
idx,
t + last_time,
max_per_millisecond=max_per_millisecond)
future_bids = bids[start_hold_idx:end_hold_idx]
if len(future_bids) > 0 and np.min(future_bids) > target_price:
signal[idx] = 1
return signal
def bid_exceeds_offer_before_drop(data,
max_drop=.00001,
min_profit=0.00005,
max_time=10000,
network_delay=30):
ts = data.get_col('t')
bids = data.get_col('bid/100ms')
offers = data.get_col('offer/100ms')
signal = np.zeros(len(ts))
max_per_millisecond = analysis.density_per_millisecond(max_time -
network_delay)
find_index = analysis.find_future_index
for idx in xrange(len(ts)):
t = ts[idx]
start_bid = bids[idx]
# average prices over the time window in which a buy message may reach
# the server
find_index(ts,
idx,
t + network_delay,
max_per_millisecond=max_per_millisecond)
delayed_offers = offers[idx:last_delay_index]
avg_delayed_offer = np.mean(delayed_offers)
last_future_index = find_index(ts,
idx,
max_time,
max_per_millisecond=max_per_millisecond)
future_bids = bids[last_delay_index:last_future_index]
sale_price_prct = (future_bids - avg_delayed_offer) / avg_delayed_offer
profit_index = find_first(sale_price_prct >= min_profit)
if profit_index:
neg_bid_prct = (future_bids - start_bid) / start_bid
quit_index = find_first(neg_bid_prct <= -max_drop)
if quit_index is None or quit_index > profit_index:
signal[idx] = 1
return signal
def min_bid_exceeds_offer_single_timeframe(data, wait_time=1500, hold_time=250, target_prct=1.0001):
ts = data['t']
bids = data['bid/100ms']
offers = data['offer/100ms']
signal = np.zeros(len(ts))
max_per_millisecond = analysis.density_per_millisecond(hold_time)
last_time = wait_time+hold_time
find_index = analysis.find_future_index
for idx in xrange(len(ts)):
t = ts[idx]
start_offer = offers[idx]
start_hold_idx = find_index(ts, idx, t+wait_time, max_per_millisecond=max_per_millisecond)
# prefilter
if bids[start_hold_idx-1] > start_offer:
target_price = target_prct * start_offer
end_hold_idx = find_index(ts, idx, t+last_time, max_per_millisecond=max_per_millisecond)
future_bids = bids[start_hold_idx:end_hold_idx]
if len(future_bids) > 0 and np.min(future_bids) > target_price:
signal[idx] = 1
return signal
