Better Tracking Of Order Trading Metrics

[nlordell]

Now that we have multiple accounts (PMMs) placing partially fillable liquidity orders, we have seen a sharp decrease in our %orders matched metric:

However, it turns out that most of these unmatched orders are just PMMs. If we query the total number of user orders (not matching the configured PMM addresses) as well as how many of those were cancelled and how many of those didn't trade, we see that our numbers are normal:

Order count query since releasing partially fillable orders:

WITH user_orders AS (
        SELECT
                o.*,
                (SELECT COALESCE(COUNT(t.*), 0) FROM trades t WHERE t.order_uid = o.uid) AS trade_count
        FROM orders o
        WHERE o.creation_timestamp >= '2022-03-22'::timestamp
          AND o.owner NOT IN (
                decode('C6093Fd9cc143F9f058938868b2df2daF9A91d28', 'hex'),
                decode('e63A13Eedd01B624958AcFe32145298788a7a7BA', 'hex'),
                decode('16760467FE05bA9B3E97E5f09E6cEaf9B88e9f13', 'hex'),
                decode('91145aa221620b22c4193AF90005ec585072F478', 'hex')
        )
)
SELECT
        COUNT(*) AS total,
        SUM(CASE WHEN cancellation_timestamp IS null THEN 0 ELSE 1 END) AS cancelled,
        SUM(CASE WHEN trade_count=0 THEN 1 ELSE 0 END) AS expired
FROM user_orders;

 total | cancelled | expired 
-------+-----------+---------
  3707 |       142 |     253

This means that we are matching around 90% of orders (which is what we had before the release).

With this in mind, this metric that we collect now is extremely low-signal. This issue captures the work to make this metric actually reflect matched order %.

Original issue gnosis/gp-v2-services#1759 by @nlordell