Author Topic: NL East Race Watch (Read 203212 times)

tomterp · « **Reply #2500 on:** September 01, 2014, 07:50:25 pm »

Quote from: PebbleBall on September 01, 2014, 06:36:36 pm

Why are the division odds lower now than at the time of the post I quoted?

Not sure, I looked for your post but it must be 30 pages back.

Anyway, the odds are better than when I posted. And note that even a simulation with 1m iterations will yield different results each time, though usually no more than a small percentage point or two.

Slateman · « **Reply #2501 on:** September 01, 2014, 11:23:03 pm »

Mathguy · « **Reply #2502 on:** September 02, 2014, 04:51:46 am »

Oh please don't put so much faith in this level of detail - there are so many unknowns (like dropping 3 to Philly) that a tenth of a percent prediction is worthless. Better to look at something like odds, such as 9 to 1

Quote from: tomterp on September 01, 2014, 06:04:58 pm

I like BP - not sure if you can see this or not, please advise.

As I recall, they play out the remainder of the season 1 million times each night in a Monte Carlo simulation, using PECOTA projections to drive the results, then just compute the # of times each team makes the playoffs in each category - Div, WC, WS etc.

http://www.baseballprospectus.com/odds/

tomterp · « **Reply #2503 on:** September 02, 2014, 08:32:04 am »

Quote from: Mathguy on September 02, 2014, 04:51:46 am

Oh please don't put so much faith in this level of detail - there are so many unknowns (like dropping 3 to Philly) that a tenth of a percent prediction is worthless.

I'm going to guess that plenty of those million simulations had the Nats dropping 3 straight to Philly. Sure, the rounding to the nearest tenth of a % implies an accuracy that just isn't realistic. In fact, I bet if they ran the simulation 1 million times, then ran it again another 1m times, the odds would change by more than 0.1%