Update daily stable token pairs query

.sqlfluff

@@ -16,6 +16,8 @@ indented_on_contents = False
 blockchain = 'ethereum'
 starttime = '2023-01-01'
 endtime = '2024-01-01'
+start_date = '2023-01-01'
+end_date = '2024-01-01'
 start_time = '2023-01-01'
 end_time = '2024-01-01'
 period = 'last 3m'

cowprotocol/lookup_tables/daily_stable_token_pairs_5719467.sql

@@ -0,0 +1,98 @@
+-- this query returns a daily list of buy and sell tokens which have a stable daily xrate between them in the previous 7d
+-- stability criteria = relative volatility < 0.7% (stddev_price / avg_price) and price range < 2.5%  (max_price-min_price)/avg_price
+-- these criteria are based on empiric evidence, but it's been observed that for some dates, pairs like stETH-WETH will show similar stats to USDT-WBTC, 
+-- so at the end we force ETH to ETH variations to be included here
+-- even though we'd like to categorise them respectively as stable and variable. 
+with
+relevant_tokens as (
+    select 
+        block_date as ref_date
+        , '{{blockchain}}' as blockchain
+        -- the following logic makes fetching the native token price possible
+        , if(
+            sell_token_address!=0xeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeee
+            , sell_token_address
+            , (select token_address from dune.blockchains where name = '{{blockchain}}')
+        ) as sell_token_address 
+        , if(
+            buy_token_address!=0xeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeee
+            , buy_token_address
+            , (select token_address from dune.blockchains where name = '{{blockchain}}')
+        ) as buy_token_address
+        , sell_token
+        , buy_token
+        , token_pair
+        , sum(usd_value) as daily_volume
+    from cow_protocol_{{blockchain}}.trades
+    where block_date between timestamp '{{start_date}}' and timestamp '{{end_date}}'
+    group by 1,2,3,4,5,6,7
+    having sum(usd_value) > 100 -- performance improvement w/ minimal impact
+)
+, daily_relevant_prices as (
+    select 
+        timestamp as day
+        , contract_address
+        , price
+    from prices.day as p 
+    join (
+        select distinct buy_token_address  as token from relevant_tokens
+        union distinct
+        select distinct sell_token_address as token from relevant_tokens
+    ) as t
+        on p.contract_address = t.token
+    where 
+        timestamp between date_add('day', -6, timestamp '{{start_date}}') and timestamp '{{end_date}}'
+        and blockchain = '{{blockchain}}'
+)
+, history as (
+    select
+        t.*,
+        pb.day as price_date,
+        pb.price as buy_token_price,
+        ps.price as sell_token_price,
+        ps.price / pb.price as xrate
+    from relevant_tokens as t
+    join daily_relevant_prices as pb
+        on pb.contract_address = t.buy_token_address
+        and pb.day between date_add('day', -6, t.ref_date) and t.ref_date
+    join daily_relevant_prices as ps
+        on ps.contract_address = t.sell_token_address
+        and ps.day = pb.day -- avoid cartesian explosion
+)
+, final_stats as (
+    select 
+        ref_date
+        , blockchain
+        -- following logis is to undo what was done before and keep it joinable with the cp trades tables
+        , if(
+            sell_token_address!=(select token_address from dune.blockchains where name = '{{blockchain}}')
+            ,sell_token_address
+            ,0xeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeee
+        ) as sell_token_address        
+        , if(
+            buy_token_address!=(select token_address from dune.blockchains where name = '{{blockchain}}')
+            ,buy_token_address
+            ,0xeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeee
+        ) as buy_token_address
+        , sell_token
+        , buy_token
+        , token_pair
+        , daily_volume
+        , stddev(xrate) / avg(xrate) as relative_volatility
+        , (max(xrate)-min(xrate)) / avg(xrate) as xrate_range
+        , avg(xrate)    as avg_xrate_l7d
+        , stddev(xrate) as stddev_xrate_l7d
+        , max(xrate)    as max_xrate_l7d
+        , min(xrate)    as min_xrate_l7d
+        , count(if(xrate is not null,1,null)) as days_of_data
+    from history
+    group by 1,2,3,4,5,6,7,8
+)
+select*
+from final_stats
+where (
+    relative_volatility < 0.007   -- relative volatility < 0.7%
+    and xrate_range < 0.025    -- xrate range within 2.5%
+    and days_of_data > 4    -- at least 5 days of xrate data
+) 
+or token_pair like '%ETH%ETH'