Win the Division this week.

[deward]

6 minutes ago, dystopia said:

Best thing Szymborski could do for himself is to go touch grass.

Or, you know, continue to learn and make tweaks to the model.

[dystopia]

3 minutes ago, deward said:

Or, you know, continue to learn and make tweaks to the model.

I mean, I think these models would be pointless no matter what. 

[Frobby]

1 minute ago, deward said:

Or, you know, continue to learn and make tweaks to the model.

The Fsngraphs playoff odds model is not his model anyway.  First, it’s based on preseason projections of players from both ZiPS (which is his model) and Stesmer (which isn’t).   Second, ZiPS is continuously updated to take into account new information from the current system, but I don’t believe the Fangraphs model uses those updates, just the original projections.  

I’d still say that none of these models have any degree of accuracy that is really helpful.  
 

[AnythingO's]

I believe alot of the discrepancy between the models' prediction and our record is due to our bullpen, both during 2012-2016 and since last year. We win (won) a higher than expected number of games late and close. There is no way a model run in March can predict what happened last summer after Lopez, an Allstar closer, was traded. Nobody could predict Bautista becoming THAT guy after all his Mil struggles. Similarly, who could predict Cano, with a Mil stat line like Felix's, would find himself in ST, certainly not a model that forecasts based on past performance. Strong bullpens skew the results.