How to reset costs?

[Plex134]

I've been trying to figure out how to reset costs for an instrument. The issue is, I seem to have an inordinately high cost to trade CORN in pysystemtrade. Mine is 0.05 SR whereas Rob's is 0.003 SR in his reports. I'm not sure if it is because of bad fills due to overtrading in a paper account, or me playing around in 'Auto update spread cost configuration based on sampling and trades' in interactive_controls before I understood what I was doing..., or both. Given the excerpts from below, it appears the bid_ask_trades and total_trades are way too high. Is there a way to manually reset these? Would deleting CORN.parquet in data.parquet.spreads do anything, or is this just the sampled spreads (this actually looks ok to me?)? Just curious how people would proceed in trying to get these costs back in line with reality.

slippage report

                         bid_ask_trades  total_trades  bid_ask_sampled  weight_trades  weight_samples  weight_config    estimate  Configured  Difference

CORN*                            3.3750        3.4375           0.1250           0.84            0.07           0.09      3.1668      2.9000         9.2

costs report

               percentage_cost  avg_annual_vol_perc  SR_cost
CORN*                  0.00617              0.12015  0.05138

CORN.parquet

                             spread
2024-11-04 00:15:04.153021  0.25000
2024-11-04 00:15:04.153021  0.25000
2024-11-04 00:15:04.153021  0.25000
2024-11-05 00:15:04.003668  0.25000
2024-11-05 00:15:04.003668  0.25000

...

2024-12-26 00:15:02.351056  0.25000
2024-12-26 00:15:02.351056  0.25000
2024-12-27 00:15:02.528127  0.22375
2024-12-27 00:15:02.528127  0.22500
2024-12-27 00:15:02.528127  0.22500

Thanks for any help,

[vishalg]

In a production trading system these are stored in the mongo database in the spread_costs collection. Check the Instrument configuration and spread costs section in the docs. You can edit it directly in mongo if you know what you are doing.

[Plex134]

I will give this a try, thank you!

[Plex134]

Update for anyone else:
As above, used process_modified_instruments in repocsv_spread_costs.py and was able to revert.
Thanks again @vishalg