brian_griffin said:
The SCS data is wrong. It must not factor tie-breakers into account. BUF does not have a 50% chance of finishing 30th with 57 points. BUF has a 66% chance of finishing 30th with 57 points. For the 2 possible combinations where BUF finishes with 57 points (W,OTL or OTL,W), ARI could in turn finish with 56, 57, or 58 points. Two of those ARI outcomes put BUF in 30th place.
Its not wrong. The 50/50 is not calculated the same way you are running your random distribution odds.
I actually just checked this out earlier to confirm, but the 50/50 method does truly assume 50% chance of winning and 50% chance of losing. However it breaks the losing into 37.6% chance the loss will be in regulation and 12.4% chance the loss will be in OT/SO for a loser point. This seems to come from the actual data of outcomes in the league - 2462 game "outcomes" (2 outcomes per GP), and 302 loser points have been awarded. 302/2426 = 12.4%. I ran the numbers and this gives 82.8% chance for the sabres finishing 30th, which matches SCS. It all matches for all the what if scenarios.
I think this is probably more accurate than assuming a 33% chance across all outcomes in a game, since the data does not show a loser point has a 33% chance of occuring across the league. In all outcomes of every game played this year, 50% of the time there will have been a win outcome, 37.6% of the time there will have been a regulation loss outcome, and 12.4% of the time there will have been a loser point outcome. This I believe is what the 50/50 method is trying to represent.
There's some information in the "what is this" link that supports this. The author states that he uses a 22% chance of going into OT, but this seems out of date since he references this was the rate of OT games in 07-08. It makes sense he would be using an up-to-date OT rate today and that probably the page just hasn't been updated.
By the way everybody really should read that "what is this" link in detail, especially before calling the site trash or its numbers trash. It knows what it is and it knows its limitations.
http://www.sportsclubstats.com/WhatIsThis.html