eczachly / video-game-training-sql Goto Github PK

Loved the SQL window session, definately learned some cool and useful stuff from that. Thanking you for sharing the updated SQL so I can have a further deep dive play to ensure eunderstanding.

I know you shared the PSQL data dump for the halo match dataset. Would it be possible to sahre as CSV or simialr, as I suppose this would be the easiest way to dump into my db of choice (Azure SQL).

Thanks!

Hi Zach, thanks much for making this tutorial. It was very useful to me!

However, I wanted to point out some strange behavior in the streak identification logic for your tutorial. I have attached a screenshot of the relevant columns obtained from the lengths table. The completion_date column does not seem to give any hour/minute/second granularity for matches completed on the same day. Therefore, the window function (in the grouped table) that is ORDERED BY completion_date would return the same result for each row in every distinct player_gamertag/playlist group that happens to take place on the same day. You can see in the attached screenshot how the downstream logic in subsequent CTEs indicates a streak of 17 for this player even though he/she had actual breaks in their streak on the given day.

Hey Zach just watched your most recent video and noticed an issue with how you identify streaks. The window function in the streak_identified CTE calculates only the total count of times the player's streak changed in the playlist, rather than assigning a unique identifier to each streak. For example the player "My Regret" has a streak_identifier that seems to stay the same across several different win/loss streaks.

I think you can add a window frame rows between unbounded preceding and current row to your window here and get the desired behavior, where different streaks have their own unique streak_identifier (unique to the player/playlist).

https://github.com/EcZachly/video-game-training-sql/blob/master/window-function-training/rank_filter_streaks.sql#L27