'dex_polygon_sandwiches` exclude failing model

[jeff-dude]

fyi @hildobby
this model started to fail today on duplicates, returning three total tx_hash values:
0x65d573be0ae948cd50d213382d1bff6470098167995fc352da90d6d65c279b9a
0xa251d71c08c1d8eabdc76944342ccec4e740705dd87ea1414fa8945c3e08921f
0x66c7e9a2c879df410e9f1125a19d80fa90d59e0ac2b52634f1188a2a03351505

screenshot of results to be safe:

query i used to find dupes (although this runs on spark, so you likely can't replicate unless you change to dunesql):

with
  src as (
    WITH
      trades AS (
        SELECT
          dt.project,
          dt.version,
          date_trunc("day", dt.block_time) AS block_date,
          dt.block_time,
          t.block_number,
          dt.token_sold_address,
          dt.token_bought_address,
          dt.token_sold_symbol,
          dt.token_bought_symbol,
          dt.taker,
          dt.tx_hash,
          dt.tx_from,
          dt.project_contract_address,
          dt.evt_index,
          t.index,
          MIN(t.gas_price) AS gas_price,
          MIN((t.gas_price / POWER(10, 18)) * t.gas_used) AS tx_fee,
          SUM(COALESCE(dt.token_sold_amount_raw, 0)) AS token_sold_amount_raw,
          SUM(COALESCE(dt.token_bought_amount_raw, 0)) AS token_bought_amount_raw,
          SUM(COALESCE(dt.token_sold_amount, 0)) AS token_sold_amount,
          SUM(COALESCE(dt.token_bought_amount, 0)) AS token_bought_amount,
          SUM(COALESCE(dt.amount_usd, 0)) AS amount_usd
        FROM
          dex.trades dt
          INNER JOIN polygon.transactions t ON t.block_time = dt.block_time
          AND t.hash = dt.tx_hash
        WHERE
          dt.blockchain = 'polygon'
        GROUP BY
          1,
          2,
          3,
          4,
          5,
          6,
          7,
          8,
          9,
          10,
          11,
          12,
          13,
          14,
          15
      ),
      sandwiches AS (
        SELECT
          'polygon' AS blockchain,
          s1.project,
          s1.version,
          s1.block_date,
          s1.block_time,
          s1.block_number,
          SUM(s2.token_sold_amount_raw) / SUM(s1.token_bought_amount_raw) AS ratio_traded_token,
          SUM(s2.token_bought_amount_raw) / SUM(s1.token_sold_amount_raw) AS profit_percentage_of_initial,
          SUM(s2.token_bought_amount_raw) - SUM(s1.token_sold_amount_raw) AS profit_amount_raw,
          SUM(s2.token_bought_amount) - SUM(s1.token_sold_amount) AS profit_amount,
          s1.token_sold_address AS currency_address,
          s1.token_sold_symbol AS currency_symbol,
          SUM(s1.token_bought_amount_raw) - SUM(s2.token_sold_amount_raw) AS profit_traded_currency_amount_raw,
          SUM(s1.token_bought_amount) - SUM(s2.token_sold_amount) AS profit_traded_currency_amount,
          s1.token_bought_address AS traded_currency_address,
          s1.token_bought_symbol AS traded_currency_symbol,
          s1.taker AS frontrun_taker,
          s1.tx_from AS frontrun_tx_from,
          MAX_BY(s2.taker, s2.index) AS backrun_taker,
          MAX_BY(s2.tx_from, s2.index) AS backrun_tx_from,
          s1.index AS frontrun_index,
          MAX_BY(s2.index, s2.index) AS backrun_index,
          s1.project_contract_address AS sandwiched_pool,
          s1.tx_hash AS frontrun_tx_hash,
          MAX_BY(s2.tx_hash, s2.index) AS backrun_tx_hash,
          MAX_BY(s2.index, s2.index) - s1.index -1 AS amount_trades_between,
          s1.gas_price,
          s1.tx_fee AS frontrun_tx_fee,
          MAX_BY(s2.tx_fee, s2.index) AS backrun_tx_fee
        FROM
          trades s1
          INNER JOIN trades s2 ON s1.block_time = s2.block_time
          AND s1.project = s2.project
          AND s1.version = s2.version
          AND s1.tx_hash != s2.tx_hash
          AND s1.index < s2.index
          AND s1.project_contract_address = s2.project_contract_address
          AND (
            s1.tx_from = s2.tx_from
            OR s1.taker = s2.taker
          )
          AND s1.token_sold_address = s2.token_bought_address
          AND s1.token_bought_address = s2.token_sold_address
          AND s2.token_sold_amount BETWEEN s1.token_bought_amount * 0.9 AND s1.token_bought_amount  * 1.1
          --AND s2.token_bought_amount > s1.token_sold_amount -- Removed to also include trades where the sandwiched trade was unprofitable
        GROUP BY
          s1.project,
          s1.version,
          s1.block_date,
          s1.block_time,
          s1.block_number,
          s1.token_sold_address,
          s1.token_sold_symbol,
          s1.token_bought_address,
          s1.token_bought_symbol,
          s1.taker,
          s1.tx_from,
          s1.index,
          s1.project_contract_address,
          s1.tx_hash,
          s1.gas_price,
          s1.tx_fee
      )
    SELECT
      s1.*
    FROM
      sandwiches s1
      LEFT JOIN sandwiches s2 ON s1.block_time = s2.block_time
      AND s1.frontrun_tx_hash = s2.backrun_tx_hash
    WHERE
      s2.backrun_index IS NULL
  )
select
  sandwiched_pool,
  frontrun_tx_hash,
  frontrun_taker,
  frontrun_index,
  currency_address,
  count(1)
from
  src
group by
  sandwiched_pool,
  frontrun_tx_hash,
  frontrun_taker,
  frontrun_index,
  currency_address
having
  count(1) > 1

took one tx, proof of no dupes in dex trades:

select *
from dex.trades
where blockchain = 'polygon'
and tx_hash = 0xa251d71c08c1d8eabdc76944342ccec4e740705dd87ea1414fa8945c3e08921f
and block_date = date '2023-04-02'