Support Board
Date/Time: Sun, 24 Nov 2024 04:44:00 +0000
Post From: .depth Files Clarification
| [2024-06-25 14:09:39] |
| skalaydzhiyski - Posts: 58 |
|
Hi guys, I am currently getting external historical MBO data and encoding it in .depth files. Everything seems to work fine when it comes to Sierra and I can replay days as far back as 2017 in the DOM. So far so good apart from my paranoia about the static const uint8_t FLAG_END_OF_BATCH = 0x01 flag... Can someone please explain what is meant by Batch here and what is the significance of a batch ? I am currently doing the following to generate the .depth file. 1. Generate .scid file with Trades from my external provider 2. Create a pandas dataframe and Add 1 row of CLEAR_BOOK command 3. Add ADD_BID/ADD_ASK 200 rows (1 for each level, 100 bid and 100 ask levels) 4. From here on I have noticed Sierra's depth files have timestamp resolution of a millisecond, so I append to the dataframe in the following manner: I get the diff between book state (external) for timestamp 3902734068951000 (for example) and timestamp 3902734068952000, for all prices not found in the first but found in the second - ADD_BID/ASK, for all prices found in the first but not found in the second, CANCEL_BID/ASK, for all prices found in both - MODIFY. 4. Use this dataframe to encode the .depth file in bytes. Note for the initial rows (ADDS immediately after CLEAR_BOOK) I have the batch flag set to 0 (replicating what .depth files usually have) for all the rest I have the flag set to 1. I know this is a bit of a confusing explanation, but I just need to know more about how the flag might impact my encoding of the files and how this affects Sierra's replays, because I need to be able to trust them... All the best, Spas |