[Jay Random]

Originally Posted by Pellanor

That's a fascinatingly bad graph. By definition a team that wins a playoff game is not going to draft in the top 10 that year, so of course there's going to be a negative correlation in it. Teams that win a lot of playoff games will not draft in the top ten. The fact that its a cloud of dots, instead of a downward line actually implies that there is some correlation between drafting top 10 and winning more games, enough so to somewhat counter the really obvious negative correlation. So there's a hint that there might be some useful information is this mess.

Um, no, it doesn't. The graph takes data over a 31-year period, which is enough time for multiple generations of players to have been drafted by bad teams, develop into good teams, and win in the playoffs – if that were the normal pattern. But it isn't.

To actually get a graph that shows high drafts leads to more playoff wins you need to massage the data more. Maybe look at number of top ten picks in non-playoff years compared to wins in playoff years?

All the top-ten picks were assigned to non-playoff teams, except for the 9th and 10th picks in 1992 (when only 24 teams were in the draft). All the playoff wins were in playoff years. Done.

You also need to remove teams that got their wins at the start of your sample, and then went into a rebuild after, since those drafts can't have caused past success.

That's a valid criticism, and one that occurred to me, too. But it isn't my graph, and I don't have the raw data handy. Scraping the data off the web by hand would take an inordinate amount of time, and I've wasted too much on this already.

Ideally, you would take the draft picks from a period starting and ending (say) about 10 years earlier than the playoff wins, and correlate those: assuming that the average age of players on a winning playoff team is about 28, which seems to be in the ballpark. Of course you would use the actual number, which, again, I haven't got the data handy to calculate. (It would be a long job to get the average age of every winner of every playoff game over 30 years!)

Don't forget to remove the two expansion teams since they're special,

Those are the extreme outliers at the bottom left. The negative correlation actually gets stronger once you remove them.

and the Oilers since they're no good to learn from.

As long as it's possible for an NHL team to be like the Oilers, they are good to learn from – if only to learn what not to do. They belong in the sample set.

Once you've massaged the data that much, you can probably get the graph to show whatever you want, so it's not going to show anything actually valuable.

You've removed three dots from the graph (Oilers, Vegas, Seattle), and precisely zero of the picks or playoff games from the other 29. This doesn't change anything, except, as I said, to make the weak negative correlation slightly stronger.

I'm not saying this graph is the be-all and end-all; it's hardly even a beginning. But it is at least an example of the kind of analysis that would need to be done to establish a correlation between high picks and winning. The people yawping loudest about the need for top picks haven't done anything of the kind.