Predictivity Review
If you are unfamiliar with predictivity basics, please check the following card
Predictivity Pricing Page
Can be found at https://[compass hostname]:8400/dashboard/pricing/predictiveness (need to edit the hostname to your Compass host server)
For the top 5 most traded instruments in your portfolio. Review each instrument's predictivity as described below
Pricing Model Predictivity
Identifying that there is a problem
negative predictivity
negative predictivity in an often traded instrument warrants a pricing backtest, to gain an indication of what is causing this.
drop in period on period performance
Select the "Compare timeframe button". This compares week to week predictivity vs the same point in the trading week last week. Sort by largest change to identify what has dropped off the most this week
pricing model performance has lower predictivity than source signals
if the signals are performing better than the overall pricing predictivity, what is inhibiting their performance / stopping them from being published. Next step is to identify what is going on via a pricing backtest:
Running an Echo Pricing Backtest
Signals Predictivity
System referenced signals
How are the system referenced signals doing this week?
A system referenced signal is one which is referenced in the configuration of the system.
Candidate Signals
Do we have any other candidate signals that are outperforming the ones referenced by the system?
Market making signals vs Risk Increasing Signals
Market making signals are evaluated on the overall improvement they have made on the pricing. They may not be incredibly high yielding but if they work on every single tick, in aggregate their improvement may be huge.
Risk increasing signals are not necessarily always active and are more focused on when they are active, the chance and size of the return of their prediction is larger.