TomG's Robot Professional MLB Betting

[TomG]

Introduction

I've built an Excel worksheet to handicap based strictly on the analytic methods outlined in Michael Murray's book Betting Baseball 2007. Here is the methodology for those who aren't familiar...

Offensive power ratings are generated using the batting lineup's OB% and SLG% adjusted for the league average.

Pitching power rating uses the starting pitcher's xERA and his average distance. Whenever possible, I use the starting pitcher's average xERA from the last 3 seasons. There is room for subjectivity here as sometimes sample size forces me to manually adjust a pitcher's xERA or average innings pitched upward or downward towards a more "normal" number. After accounting for the starting pitcher's average distance, I use the team's bullpen average xERA. The team's total pitching power rating is the weighted average of the starting pitcher's xERA and the bullpen's xERA.

Using Michael Murray's methods, I calculate each teams expected runs scored by multiplying the team's offensive power rating by the opposing team's pitching power rating and dividing by the league's average runs scored. Bill James's Pythagorean Formula (using exponents taken from Pinnacle's total runs line) is used to convert each team's expected runs scored into each team's expected win %. I then add a flat 4% to the home team to account for home field advantage.

The result is an expected "true line" that I compare against the posted line for value.

My goals...

1) Bet overnight MLB moneylines using this system
2) Beat the closing line on the majority of games
3) Achieve a positive ROI over a large sample size

I will start out flat betting but I may switch to a Kelly system if/when I gain confidence in my ability to quantify the approximate value of each play. I will be betting my picks with my personal bankroll, but I cannot guarantee the accuracy and profitability of this system. It is as much an experiment as anything. Coat tail at your own risk!

Finally I would appreciate any suggestions/comments on this methodology. My interest is in creating and tweaking mathematical models. I generally have little/no personal opinion on any of the games. As such, I won't be able to comment on any subjective aspects of the matchups.

Good luck everyone!

[TomG]

Picks for the next day's games will be posted in this thread between 8 pm and 10 pm PST.

[Girchuck]

What is a large sample size?
Is it 1000 games, 10000?
How many seasons?

[TomG]

[ QUOTE ]
What is a large sample size?
Is it 1000 games, 10000?
How many seasons?

[/ QUOTE ]
For practical purposes, I'll probably bet for a few hundred games and then re-evaluate at that time if I want to continue onward.

[rjp]

If you're worried about over estimating your edges then consider using a quarter kelly. Even though you aren't confident in your edges, you should still be betting more for a larger favorite than you would for a larger underdog.

[ImStillBen]

You will get crushed.

[TomG]

Wednesday, May 16th

Chicago Cubs at New York Mets
Chicago Cubs +113
Risk 1 unit to win 1.13 units

San Francisco at Houston
San Francisco +102
Risk 1 unit to win 1.02 units

Arizona at Colorado
Colorado +104
Risk 1 unit to win 1.04 units

St. Louis at LA Dodgers
LA Dodgers -145
Risk 1.45 units to win 1 unit

Baltimore at Toronto
Toronto -157
Risk 1.57 units to win 1 unit

LA Angels at Seattle
Seattle +127
Risk 1 unit to win 1.27 units

Good luck.

[Girchuck]

Would it be too conservative to add 5-6% to your "true line" to get an "unknown error adjusted line" and use that?

[TomG]

[ QUOTE ]
Would it be too conservative to add 5-6% to your "true line" to get an "unknown error adjusted line" and use that?

[/ QUOTE ]
Hmm it doesn't make sense to me to add additional noise to the model to account for existing noise. If this model truly is flawed/broken I'd rather just ditch it.

Here they are the "true lines" for the picks I posted...

Chicago Cubs +113 (True Line -122)
San Francisco +102 (True Line -140)
Colorado +104 (True Line -112)
LA Dodgers -145 (True Line -222)
Toronto -157 (True Line -274)
Seattle +127 (True Line -102)

I'm concerned when I saw Crockpot (a poster I respect a lot) fading my Dodgers pick and Giants picks.

[Girchuck]

I've failed to communicate.
The additional 5-6% is an engineering "safety margin", not additional noise.

[ QUOTE ]
[ QUOTE ]
Would it be too conservative to add 5-6% to your "true line" to get an "unknown error adjusted line" and use that?

[/ QUOTE ]
Hmm it doesn't make sense to me to add additional noise to the model to account for existing noise. If this model truly is flawed/broken I'd rather just ditch it.

Here they are the "true lines" for the picks I posted...

Chicago Cubs +113 (True Line -122)
San Francisco +102 (True Line -140)
Colorado +104 (True Line -112)
LA Dodgers -145 (True Line -222)
Toronto -157 (True Line -274)
Seattle +127 (True Line -102)

I'm concerned when I saw Crockpot (a poster I respect a lot) fading my Dodgers pick and Giants picks.

[/ QUOTE ]