bid_price == -9.223372036854776e+16?

[LJingnan]

hbt = HftBacktest(
    [
        'btcusdt_20240522.npz',
    ],
    tick_size=0.01,
    lot_size=0.001,
    maker_fee=-0.00005,
    taker_fee=0.0007,
    order_latency=FeedLatency(),
    queue_model=SquareProbQueueModel(),
    asset_type=Linear,
    trade_list_size=10_000,

)

@njit
def print_bbo(hbt):
    t = 0
    bid_prices = np.zeros(11721077)
    ask_prices = np.zeros(11721077)

    while hbt.elapse(100000):
        bid_prices[t] = hbt.best_bid
        ask_prices[t] = hbt.best_ask
        t = t + 1
        print(t)

        # !!!!
        hbt.clear_last_trades()

        if not hbt.run and t >= bid_prices :
            break
    return bid_prices, ask_prices
bid_prices, ask_prices = print_bbo(hbt)
max(bid_prices)  # 70642.7
min(bid_prices)  # ???? -9.223372036854776e+16

I download data using python, But I fould some bid_price where strange. Is it a collection error? Here is the data sample

[nkaz001]

At the start of the backtesting or right after a depth clear event, the market depth is empty, resulting in no bid or ask. In such cases, it returns INVALID_MIN or INVALID_MAX based on the side, defined as -sys.maxsize or sys.maxsize in marketdepth.py. You can check this by using from hftbacktest.marketdepth import INVALID_MIN, INVALID_MAX.

[LJingnan]

At the start of the backtesting or right after a depth clear event, the market depth is empty, resulting in no bid or ask. In such cases, it returns INVALID_MIN or INVALID_MAX based on the side, defined as -sys.maxsize or sys.maxsize in marketdepth.py. You can check this by using from hftbacktest.marketdepth import INVALID_MIN, INVALID_MAX.

Would this affect backtesting,because I find out that my collecting program restart often.

[nkaz001]

Collecting program data doesn't have to do with backtesting. Connection loss often happens. It's recommended to handle invalid values to ensure that erroneous actions are not taken in backtesting.

[LJingnan]

Collecting program data doesn't have to do with backtesting. Connection loss often happens. It's recommended to handle invalid values to ensure that erroneous actions are not taken in backtesting.

I see, could you please give me some advise? When I encounter INVALID_MIN or INVALID_MAX then I clear all position and wait until a DEPTH_SNAPSHOT_EVENT? Is this logic ok?

Now I use the method: I split data on depth_clear event, and bacltest them one by one and collect them together at the end. Is this a good method?

[nkaz001]

I'd rather wait until both a valid bid and ask are received. Ultimately, the same situation will occur in live trading, so there's no use in applying methods that only work in backtesting.

[LJingnan]

I'd rather wait until both a valid bid and ask are received. Ultimately, the same situation will occur in live trading, so there's no use in applying methods that only work in backtesting.

I see, Thank You!