What slippage model will be most realist to live trading? I am conducting my own backtesting and simulation for a trading algorithm. I intend to make "small" order sizes. The approximate number of roundtrip orders would be 50 - 100 per day. The ordersize will be 100 shares, and the average volume of the stock is 12 million. I want to quantify the price slippage based on market impact and time slippage. I have 4 ideas how to do this. I only intend to have 1 model active at a time. 1. Slippage by number of ticks after order is sent 2. Slippage by basis points per share. (15 bps min) 3. Slippage by using bid/ask spread 4. Slippage as a percentage of the tickbar. These 4 models do not account for the ordersize or the daily volume of the stock. Since my order size is "small", I would think that there would be a simple model for the relationship of ordersize vs. volume traded. Does anyone use a slippage model with this type of consideration? What slippage model do you use? Thanks!
How exactly are you defining slippage here? Are you already using bid/ask data to handle transaction costs or are you just using last traded? If the latter, if the spreads are relatively stable - typically I just check my equity curve by removing avg_spread, avg_spread*spread_sd1, and avg_spread*spread_sd2 which is fairly crude but is fine in a few cases. Now if you're talking about latency, in my event driven backtest framework I simply introduce latency into the system, allowing for execution after a defined period (say 100 ms). If you don't want to get that fancy and don't work with data quite that high frequency, suppose you're using minute bars, just delay execution to your next bar and that's just a super conservative estimate. Hope that helps!
1. Slippage by number of ticks after order is sent. Impliment by specifying a quantity of ticks for your order entry to "slip" before getting filled. 2. Slippage by basis points per share. (15 bps min) Impliment by specifying a number of basis points (bps) per share for your order entry to "slip" before getting filled. This would be applied to the entry and exit. 3. Slippage by using bid/ask spread. Assume order filled at bid/ask with no additional slippage. 4. Slippage as a percentage of the tickbar. Your slippage would be a set percentage of the range of the tick bar (from high to low). I am using last traded tick data which I am treating as market orders for my transactions. I don't fully understand how you are treating slippage based on your description. Can you elaborate? My slippage model has to be fairly realistic. If the model is too conservative, the strategy will not be profitable. I need to replicate slippage as precise as I can for my strategy to give the strategy credibility. I believe my tick data timestamps are only in seconds. So I can't do any slippage associated with ms unfortunately. If I were slipping only 100 ms, I believe that my strategy should be viable. Thanks for your comment! And Sorry about the late reply!
Wait it sounds like you're using barred data? Are you barring this from raw tick data? Or do you have snapshot data? Re: Bid/Ask - Starting point would be to reconstruct the order book, then simulate your own interaction with it.
I believe this is raw data. Its not a snapshot. The time stamps are only in seconds, but I have multiple orders for the same timestamps. I found a free data source for the tick data I am using. I have compared to other data sources, and it coorelates with the historic market tick data of other data providers. I am kind of a newbie so I apologize for my ignorance. I appreciate you sharing how you treat slippage. I still don't know how I would implement what you are describing. Can you direct me to some literature about this subject. I have scoured the internet about slippage to get the ideas stated above. Thanks for your reply!