When you auto-trading guys do back testing, how do you calculate the cost for spread? For one trade(one buy order and one sell order), do you need to double the spread as cost? or only one fold of the spread for 2 orders?
If you had thought about it for one second before posting you would have figured that out yourself. Once of course. Example: current market 100.0x100.2 You buy at 100.2 and instantly sell at 100.0. You paid the spread once. There is no mathematical nor logical reason why this does not hold no matter the time span between entering and exiting.
LOL. And if you do it even faster, maybe your broker will not even notice you and will forget to charge you commission for the second trade. Lightening speeds means single costs anytime. Anyway. In that example of yours you can make 4 possible trades. If you are are patient you would be making a market, if you are aggressive you would be paying the spread. How many times do you need to do what in one whole cycle is up to you? Spreads and slippage costs are like commissions. You will incur them every time you transact.
Spread is not charged twice like slippage and commisions. You kind of buy at ask and sell at bid, or mid... Although it's 2 transaction with commissions often charged each way, and slippage always possible, spread is not double charged.
In the example above you have 4 possible trades. BuyxSell 100x100.2 100x100 100.2x100 100.2x100.2 All different pnl outcomes. In this simplistic example, the biggest difference is the spread. Now, feel free to model your trades as close as you think the actual execution will happen in real time. This surely will make all the difference.
Nonsense. The spread is paid on average once per trade for liquidity takers. This this absolute basic 101 for any algo trader and quant who learns to test ideas and needs to factor in cost of execution. You don't have 4 possible trades. You have one trade. If you buy to open you sell to close and vice versa. As liquidity taker you pay on average the spread once, you can do even better if you provide liquidity, but the question was about standard business practices which err on the conservative side.
It's all about algo evaluation. In the example above your pnl could swing from -.20 to +.20. Metrics are all about the base price. If you are a pure execution trader or a long term investor, your base price would be a zero transaction cost. All you would care about then is passing the spread to the client or doing better than that. If you are a short term intraday high frequency trader, I doubt breaking even is your base metric. In that case you may want to take into consideration the full spread swing on entry and exit in order to optimise execution. I you keep taking the full spread all the time you are not going to last long here. This will make or break a market making strategy for example.
You are conflating many different issues that are not relevant to the actual question. You need to model cost of execution via TCA or by harnessing actual execution statistics of past trades but that was not the question. The question was very well framed and posed. Read it again.
LOL\that maybe is why my broker tends to notice most every time. Actually in his number example ,that could be done in a liquid ETF market. And IF ONLY we could keep price close to his number+ make a good trend profit/LOL For one buy + sell, I would figure double the spread+ when we get positive slippage, count your blessings. Add the SEC fees, however low they go.......................................
Based on the original OP first post, the answer is: (InsideAsk - InsideBid) * BigpointValue ; However, the OP has modified the request making the above incorrect.