TomG's Robot Professional MLB Betting - Page 27

JoshEngleman · #**261** 06-29-2007, 04:34 PM

yeah, there is one sheet per team which has the bullpen stats pulled from fangraphs.com

I just cant figure out how to type Royals in cell A1 and have the IP/HR/BB/KK show up in cell A2,A3...

whipsaw · #**262** 06-29-2007, 04:36 PM

[ QUOTE ]
yeah, there is one sheet per team which has the bullpen stats pulled from fangraphs.com

I just cant figure out how to type Royals in cell A1 and have the IP/HR/BB/KK show up in cell A2,A3...

[/ QUOTE ]

The easiest way would be to have a separate sheet that compiles every team's bullpen stats into one table.

TomG · #**263** 06-30-2007, 03:36 AM

Updated YTD:
123.53 units wagered
+16.52 units
+13.4% ROI

Alright I'm bringing this back. I've made some tweaks and improvements. Specifically...

1) Improved measurement of homefield advantage
2) Current season offensive performance is now regressed against a pre-season projection
3) I am no longer using xERA for pitching. Instead, I am using a true defense independent measure of pitcher/bullpen performance.

I am now more confident in the model. The only major area where I am lacking is defensive modeling. I would like to use BABIP or DER to account for fielder performance but I am currently unsure how to convert an efficiency stat such as these into marginal runs saved/allowed.

Saturday MLB...

Oakland at NY Yankees
Oakland +139

Minnesota at Detroit
Minnesota +140

Milwaukee at Chicago Cubs
Milaukee -110

Tampa Bay at Cleveland
Tampa Bay +230

Toronto at Seattle
Toronto -137

JoshEngleman · #**264** 06-30-2007, 11:55 AM

I just entered my data or the As/Yankees game, and since we have similar system I'm not surprised they are in agreement here. I'm really wondering what vegas is thinking on that line, because I just don't see it at all.

TomG · #**265** 06-30-2007, 12:27 PM

I calculated out a "true line of Oakland +115. So they should be a dog, but I see the line shaded about 10% against the Yankees (not too surprising).

irisheyes · #**266** 06-30-2007, 12:42 PM

Tom, first--good luck. I am glad you're back.

Are you concerned that using DIPS without defensive stats factored into the model could skew in favor of poor defensive teams?

JoshEngleman · #**267** 06-30-2007, 01:02 PM

I calc'd Oakland to -109. Some of my numbers have been pretty drastic lately and I'm not too sure why. I'm starting to think that I may be doing something wrong. I have the As giving up 4.24 runs and scoring 4.39 runs on average. I have the yankees giving up 5.76 and scoring 5.26. Normally, I would first plug those totals into the pythagenpat equation.

(4.24 + 4.39)^.285=1.85

That gives me the run environment exponent. I then take that exponent and plug it into the Pythagorean theorm.

(4.24^1.85)/((4.24^1.85)+(4.39^1.85))=.516

I would do the same for the yankees numbers, which is .455

I then plug those two winning percentages into the log5 method, to determine how these teams would fair playing against each other. I add HFA in before that calc, btw.

Based on my numbers, log5 says the a's would win this matchup 52.09% of the time which corresponds to a -109 line. I put that line and the vegas line in the kelly equation to see the value of the bet.

Does anything seem wrong with this?

TomG · #**268** 06-30-2007, 01:11 PM

Josh, your math is correct. The differences are going to be in the inputs, not the calculations. Little things are going to throw out much different lines... for example, if you give Chad Gauden an ERA of 5.0 and Kei Igawa and ERA of 5.5 you might get a line of Oak +110. But if you give Gauden and Igawa a similar ERA of 5.5 then all of a sudden Oakland is +131 and there is no value in the bet. That 0.5 run has made all of the difference. So you can see how sensitive the model is to the input values. Hitting, pitching, fielding, etc. could all have a 0.5 run impact and thus a difference in ROI. It is critical that you use measures that capture the expected runs scored/allowed as accurately as possible.

JoshEngleman · #**269** 06-30-2007, 01:20 PM

Well, at least the math is correct. Lol.

The Yankees game is probably a bad example just because it's the Yankees and public opinion sways that line more than anything.

I guess I'm just happy to know I'm not making any big errors.

TomG · #**270** 06-30-2007, 01:27 PM

[ QUOTE ]
Tom, first--good luck. I am glad you're back.

Are you concerned that using DIPS without defensive stats factored into the model could skew in favor of poor defensive teams?

[/ QUOTE ]
Good question. The answer is yes... and no. Since I am using a defense independent pitching stat I am thus excluding fielder performance. I am missing that part of the picture. I would like to account for defense, but I feel the model is still reasonable even without it. Here's why...

Defensive Efficiency

The team with the best defense, the New York Mets has a DER of 0.735. The team with the worst defense, Tampa Bay has a DER of 0.667. The difference is 0.068. To put it into words, the difference between the best defensive team in baseball and the worst defensive team in baseball results in a average of 6.8 extra hits per 100 balls batted in play.

Basically, if the difference between an excellent defense and the worst defense is not that large, then on average the difference will be even smaller. At least compared to hitting and pitching, defense is not very important.