search

bug-or-feature / pysystemtrade-fsb

Systematic spread betting in Python
https://pysystemtrade-fsb.bugorfeature.net/
GNU General Public License v3.0
10 stars 2 forks source link

Epic variation checker #11

Closed bug-or-feature closed 1 year ago

bug-or-feature commented 2 years ago

Need a new report that occasionally (~monthly?) checks for the existence/usage of epic variations. There is a suspicion that IG do not always use the same roll calendar each year, per instrument.

This was illustrated recently when investigating a roll issue with SOYOIL_fsb showed that there were 4 epics, not 3, and I had to reconfigure the instrument and rebuild the data. It's possible that I made a mistake with the original config, but also that the contracts for which IG offer bets varies over time. There was also an occasion in 2021 when the GOLD_fsb bets varied unexpectedly

There is a process in the data project that does this, but it has only been run once or twice. Need to formalise it into a regular thing, and build it in such a way that the rate limits are not exceeded

markgahagan1 commented 2 years ago

IG do not maintain consistency of codes across any roll calendars. Apparently it is just up to the dealing desk to decide what code is used next and there is no predetermination. I just search the codes on a regular basis to get the furthest ones out. Clunky but works for me but not in pysystemtrade.

On 7 Jun 2022, at 12:31, Andy Geach @.***> wrote:

 Need a new report that occasionally (~monthly?) checks for the existence/usage of epic variations. There is a suspicion that IG do not always use the same roll calendar each year, per instrument.

This was illustrated recently when investigating a roll issue with SOYOIL_fsb showed that there were 4 epics, not 3, and I had to reconfigure the instrument and rebuild the data. It's possible that I made a mistake with the original config, but also that the contracts for which IG offer bets varies over time. There was also an occasion in 2021 when the GOLD_fsb bets varied unexpectedly

There is a process in the data project that does this, but it has only been run once or twice. Need to formalise it into a regular thing, and build it in such a way that the rate limits are not exceeded

— Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you are subscribed to this thread.

bug-or-feature commented 1 year ago

This is done. There's a new overnight process that scans IG for variations in the known epics. Ignores FX due to their regular HMUZ periods. It's slow due to the rate limits; can do 25 instruments per night without clashing with other processes. The least recently calculated epics are written to mongo with a timestamp. A new section in the FSB report shows any mismatches with config