FiveThirtyEight MLB Prediction results so far

[AJisKewl]

Take this with a small grain of salt gents...

I know a lot of people don't think much about FiveThirtyEight.com's predictions (I mean, they completely screwed up on the 2016 election), but they are moderately successful for some teams at predicting the winner in MLB matchups so far this year. Basically what they do is come up with a team rating for each team, then adjust the rating based on pitcher quality for a given match up, then adjust the rating one last time for home-field advantage, rest, and travel.

I built a scraper that gets each team's game history with an accompanying FiveThirtyEight prediction and determines if the prediction was correct or wrong. For example: If FiveThirtyEight predicts Rockies = 55% and Padres = 45%, and Rockies win, then this is a correct prediction. If FiveThirtyEight predicts Rockies = 39% and Padres = 61%, and Rockies win, then this is a wrong prediction. If a teams' percentage > 50%, they are predicted to win the game.

Here are the results from my analysis of FiveThirtyEight's daily MLB predictions.
Rangers = 73.17
Phillies = 72.97
Padres = 66.67
Mariners = 65.85
Nationals = 64.1
Astros = 63.41
Angels = 62.79
Athletics = 62.5
Yankees = 62.16
Dodgers = 60.98
White Sox = 57.89
Braves = 56.76
Royals = 56.41
Blue Jays = 56.1
Pirates = 52.5
Twins = 51.43
Red Sox = 51.28
Rays = 51.16
Dbacks = 50
Cardinals = 50
Orioles = 50
Tigers = 50
Indians = 48.72
Marlins = 48.72
Mets = 48.72
Cubs = 46.15
Rockies = 45
Brewers = 43.9
Reds = 43.59
Giants = 42.86

You can read each entry as: "If FiveThirtyEight predicts the Rangers will win (based on giving them a percentage>50%) or lose (based on giving them a percentage <= 50%), their prediction is correct 73.17% of the time."

I know it's still early in the season, and this doesn't mean much, but I wanted to share it with fellow SBR guys. I value everyone's insight.

[2daBank]

Here the thing tho the lines in themselves give teams probability to win. Per your example if they give the Rockies a 55% chance to win vs the pads then that a ml of roughly -122.. so if Rockies were -150 that day are you still giving them credit for being correct if they win?

The fact they have been on the right side of 50% to say rangers will win or lose 73% of the time doesn't really mean anything.

[StackinGreen]

That's the first thing that stood out to me. It's not about making a winning pick, per se. It's about what they are winning when they make the winning pick, and losing when they make the losing pick.

Ideally, you'd like to see if they are overestimating the dawg/underestimating the favorite, and vice versa. This is very simply shown by anyone who ever created a line for anything. For example, if the Orioles are rated by 538 to be 60% chance to win, then any line under -150 is a bet. Got it? That's the easiest way to see how well they are doing. The discrepancy of ML to their ratings will show how much better they are doing monetarily, over time.

[AJisKewl]

I understand what you guys are saying here, and I agree that we should check for their bias. I'll add the lines and money winnings to this analysis.

This was just simply seeing how good they were at picking a straight up winner.

[2daBank]

I understand what you guys are saying here, and I agree that we should check for their bias. I'll add the lines and money winnings to this analysis.

This was just simply seeing how good they were at picking a straight up winner.

what be sweet is you find they have some teams they rating above your 50% threshold (or even close) for a win when said team is winning as a underdog. Then you on to something!!

I don't know anything about the site (never heard of), maybe they do some solid work beats me. Glad you get what me and SG were saying as far too many want to go nuts on win percentages when they don't mean all that much by themselves. I kinda suspect their teams percentage chance of winning/losing a game gonna fall fairly close to the line but heck I'm curious to see what you come up with.