Framework for Baseball Approach

[blogguy]

I've been playing around with a basic framework for baseball handicapping the past few days that's been showing some good potential (10-2 between the Airbets contest and Leo's Play of the Day thread). Figured I'd write it up so everyone can take a look at it and help me think it through.

The basic goal was to make sure that we were taking into account starting pitching, offense, and the bullpen. A lot of handicappers tend to focus on starting pitching...then kind of gloss over the other two figuring that stuff cancels out. I myself was kind of glossing over bullpens because of time constraints. Now that we've got a way of monitoring all three fairly reasonably, I wanted to put together a way of using them for each game.

The starting point was this:
*Rate each offense on a scale of 1-3 (3 for the best, 1 for the worst)
*Rate each starting pitcher on a scale of 1-3 (3 for quality starters, 1 for guys with ERA's over 5.00...2 for innings munchers who are okay but not really aces)
*Rate each bullpen on a scale of 1-3 (3 for the top 10, 1 for the bottom 10)
*Add 1 point for home field advantage (I may decide later that 0.5 is better...one point seems too high for some teams...but it's got to be worth SOMETHING)

Then...simply add up the points and see what we have for each game. I know that sounds ridiculously simple. But...it's a simplicity that forces offense and bullpens to be considered. And, as a result...it's a simplicity that can possibly find holes in Vegas lines that are based more on starting pitchers and less on the other things.

In actual practice, it's important to allow for some flexibility. If you can't decide between a 2 and a 3...just put in 2.5 instead of arguing with yourself for hours. Is Tom Glavine a 3 pitcher? No...not if you think of Roy Halladay as a 3 pitcher. Is Glavine just a 2 who's an innings muncher? Well, his stats are better than that. Stick in a 2.5 for now and change it down the road once you've got more data.

Some offenses are also "tweeners" like that. A 3 seems to high...but you know they've got explosive potential in the right spots. A 1 seems to low, but you can't really justify a 2 either. Just put in the halfway number and move on.

How do you rate the elements?
*Starting pitchers: Look at the full season ERA's in your morning newspaper...adjust based on recent form if you wish based on the last 3 starts number that's also in your paper.

*Bullpen: Use the chart from covers.com that's down in the THINK TANK thread, or go to the stats page at covers.com for a daily update on bullpen ERA's. This past weekend I simply used 3 points for the top 10...2 points for the middle 10, and 1 point for the bottom 10. I "tweaked" if a starter for a good team wasn't likely to go deep in the game. So, the Yankees got penalized a bit on the bullpen in Randy Johnson's starts because he's been getting knocked out so early. The worst part of the Yankees' pen was likely to pitch (though it was ultimately the best part that lost the game late)

*Offense: Use the median from the ROAD scoring line down in the THINK TANK thread that shows what each team has scored this year ranked from high to low. Using the ROAD numbers helps take out home park illusions...and forces you to see that the Dodgers have a GREAT offense and the Reds are kind of timid when they're not playing at home with the wind blowing out. That scoring line is the truest sense of real offensive ability I know of.

Once you've got the numbers in place, you don't have to just look at the point totals you calculate. You can look for certain combinations of things that lead to proper decision-making:

*A 3 pitcher with a 1 bullpen is a guy you only want to take for the 5-inning line if you like him because the bullpen is terrible

*A 3 pitcher with a 3 bullpen is a guy you can back either way because your pick is less likely to get burned after the starter leaves

*A 1 pitcher with a 1 bullpen is a recipe for disaster. The other team is likely to score a bunch of runs, and is a good take at a cheap price

*A pair of offenses that add up to 4 or more vs. a pair of pitching combinations that add up to 3.5 or less, and you want to consider the Over

*A pair of pitching combinations that add up to 8 or more against a pair of offenses that add up to 3.5 or less and you want to consider the Under

You can also look for side and total tandem combinations this way:
*Take any 2.5 offense AND the Over vs. a pitching combination of starter and bullpen that adds up to 3.5 or less

*Take any 3.5 pitcher AND the Under vs. any offense that registers at 1.5 or less

For those of you who are applying the "cocktail" approach with umpires, weather, and ballparks...look for conditions that favor what the numbers are showing. Obviously the pitching indicators are stronger in pitchers parks, with the wind blowing in, or with Under umpires. The offensive indicators are stronger in hitter's parks, with the wind blowing out, or with Over umpires.

Of course, the most basic way to deal with the data is to simply add up the full team numbers and look for value against the line. This is still a work in progress, so I can't say that you've definitely got to do x with and advantage of 3 or more, or y with an advantage of 2 or less.

My general goal is to find:
*A range that defines" comparable" teams so we can use the coin flip theory that's worked so well during March Madness and the NHL playoffs lately (not to mention the bowls forever) and just play the underdogs where there's a comeback.

*A range that defines "non-comparable" with corresponding moneyline price ranges that make sense. You've got to have a HUGE edge to lay 200-250 (St. Louis did have huge edges and went 3-0 with a 28-11 scoreboard edge even with the big moneylines versus Kansas City). A big edge allows for plays at 160-190. Something at least 3 or more at 150 or less suggests a play on that price favorite.

*Teams who qualify as "non-comparable" who are playing in hitter's parks (index of 102 or higher) where victory margins of 1 run are rare. You can get these teams at underdog prices at -1.5 runs where the evidence seems to make it clear that it's much better than a 50/50 shot that the superior team will win by 2 runs or more.

I know this sounds ultra-simplified. But...the fact that includes both offenses and bullpens I believe forces everyone to see things they weren't seeing before. I've heard a few people who liked the Angels over the Dodgers this weekend, apparently unaware that the Dodgers offense was significantly better going in. The hosts had the much better offense, a starting pitching advantage in every game, and home field...but people were backing the slumping Angels in each game. A slumping team with clear disadvantages on the road? (Dodgers go 3-0 with a 31-7 scoreboard edge) There have been complaints about a few of the bullpens this year. Forcing yourself to use an equation that penalizes the bad bullpens keeps you from investing in teams like that. No reason to take a flyer on Florida (0-3) or Kansas City (0-3). Toronto has awful bullpen stats, and the team got crunched in Colorado. You could go against Towers and a minor league call-up very cheaply with a Colorado team that had better starting pitchers in those games and a much better bullpen along with a unique home field that Toronto hadn't dealt with before (Rockies go 3-0 with an 18-7 scoreboard edge).

This is the general framework I'll try to be building from in the coming days in the TANK. We'll see what time permits on a daily basis. Please feel free to share your thoughts. I know there will be different views on how to shade all the different data. I know Spraguer prefers a leveraged approach to bullpens...and that jjd prefers the ERA's of the current bullpen pitchers rather than the full season numbers. For now, I'm going to use the full season rankings and we can make changes if evidence suggests we should.

[BuzzRavanaugh]

I love this approach. Think this sounds good.

I wish I could find my old copy of Lem Bankers book on Sports Betting. His baseball system was very good. It gave you a value number. I used it 20 years ago with info out of the USA Today.

I would get interested in baseball if there was a way to do something simple like this blogguy. I think you got something here.

[cash]

<div class="FTQUOTE"><begin quote>Originally posted by: BuzzRavanaugh

I love this approach. Think this sounds good.



I wish I could find my old copy of Lem Bankers book on Sports Betting. His baseball system was very good. It gave you a value number. I used it 20 years ago with info out of the USA Today.



I would get interested in baseball if there was a way to do something simple like this blogguy. I think you got something here.</end quote></div>

I agree Buzz. Indexing teams this way should bear some fruit.
Without cocktailing totals just using medians with a 2:1 bias
this weekend yielded a 10-4 result.

[blogguy]

Thanks for the votes of support guys. I'll try to put something together for Monday's games by midafternoon so we can look at the possibilites together for Monday Night. Good luck finding that Lem Banker book Buzz.

[count zero]

I have the Lem Banker book (I think -- "Lem Banker's Book Of Sportsbetting," 1986, right?), and I have to say I never found it particularly useful. It's like every other sportsbetting book, they're never going to tell you everything. Anyway, FWIW, his baseball system is:

1) give every team a HFA varying from 0 to 20c.

2) compare the starting pitchers' ERAs and give 20c for each run of difference to the better pitcher.

3) compare the teams' OBA, which is (walks + hits + safe on error) / (plate appearances), subtract the smaller from the larger and add that many cents to the better team.

4) total everything up, then modify (he does not specify how) the final figure to reflect the difference in the teams' bullpen strength, paying special attention to fatigue factors for the closer and setup man. His examples suggest 5c for each case of having a much better closer or much better setup crew.

A bet is said to be generated when the betting line is 15c different from Banker's line. That is obviously simplistic, and I would think that any English-speaking person could do better.

Banker also mentions that he specifiacally ignores a) managing, b) injuries, except to superstars or cluster injuries, and c) matchup factors, like how a player does against certain pitchers or in certain ballparks. But he then says he also uses two angles to generate additional bets: 1) bet against teams that are coming home off a long trip, and 2) bet against pitchers in their second start following an injury that caused them to miss time. Finally, he advises you to shop and not be afraid to pass games. Duh.


It's not that I didn't enjoy the book, and back in the day I saw with my own eyes how good Banker was at pro football (from his weekly picks in the newspaper), but I didn't learn anything from it, just like I rarely learn anything about handicapping from books. I guess it wouldn't take a huge database to be able to calculate all the data his system uses (although it would be a large and unpleasant task, IMO), so in principle you could test how the method has performed in the past.

[topspinner]

Blog, interesting approach....one thing I might suggest adding is how the teams have done over the last ten games. Maybe if a team is two or more games better than an opponent, you add a point. Baseball teams seem to be very streaky...both hot and cold.

[Spraguer]

Bloggy, while I think there are some problems with this approach - its simplicity is both its blessing and its curse - I wanted to contribute something to your idea.

So, straight from my spreadsheets, here are my ratings divided into groups to correspond to "The Blogguy 1-2-3 System".

AL Offense
3: Bos, NYY, Cle, Det
2: Tex, Oak, Tor, Sea, ChW, Bal
1: LAA, Min, TB, KC

NL Offense
3: NYM, Cin, Phi, Atl StL
2: Mil, SF, LAD, SD, Hou, Fla
1: Pit, Arz, Was, Col, ChC

AL Starting Pitchers
3: Vazquez, Schilling, Johnson, Santana, Halladay, Hernandez, Verlander, Liriano, Escobar, Mussina, Lackey, Harden, Haren, Janssen, Bonderman, Buehrle, Gregg, Radke, Burnett, Garcia, Millwood, Baker, Beckett, Contreras, Kazmir, Weaver, Hammel, Clement, Pavano, Colon.

2: Carmona, DiNardo, Wilson, Bedard, Loe, Wang, Blanton, Bautista, Towers, Santana, Garland, Lilly, Loaiza, Zito, Cabrera, Carrasco, Robertson, Greinke, Eaton, Affeldt, Tejeda, Downs, Wells, Lee, Padilla, Pineiro, Sabathia, Koronka, Wakefield, Rogers, Washburn, Halsey.

1: Fossum, Maroth, Chacin, Lopez, Wright, Byrd, Chen, Saarloos, Westbrook, Silv, Redman, Johnson, Chacon, Moyer, Bonser, McClung, Taubenheim, Small, Halama, Halama, Hendrickson, Lohse, Hernandez, Meche, Benson, Dickey, Waechter, Bauer, Elarton, Mays, Haeger.

NL Starting Pitchers
3: Sheets, Carpenter, Prior, Martinez, Oswalt, Peavy, Cruz, Pettitte, Lieber, Wood, Bush, Maddux, Reyes, Smoltz, Webb, Schmidt, Zambrano, Cain, Hill, Eveland, Guzman, Nieve, Young, Nolasco, Bannister, Hamels, Madson, Myers, Davis, Harang, Lidle, Thomson, Maine, Olse, Penny, Perez, Lowry.

2: Capuano, Francis, Kim, Hudson, Snell, Patterson, Hernandez, Morris, Hensley, Johnson, Willis, Arroyo, Ramirez, Traber, Davies, Ponson, Batista, Hendrickson, Lowe, Mitre, Duke, O'Connor, Kim, Vargas, Marshall, Rusch, Seo, Williams, Williams, Perez, Maholm, Mulder, Rodriguez, Zambrano, Park, Milton, Tomko.

1: Suppan, Claussen, Floyd, Ohka, Backe, Marquis, Glavine, Hernandez, Gonzalez, Jennings, Astacio, Wright, Thompson, Smith, Trachsel, Vargas, Estes, Buchholz, Santos, Ramirez, Moehler, Hennessey, Wilson, Ascensio, Cook, Ortiz, Fogg, Day, Sosa, Armas, Williams, Drese, Brazelton, Ortiz, Lima, Sele, Hernandez.

MLB Bullpens

3: Col, Det, Bos, NYM, Oak, Tor, Phi, LAA, StL, Arz
2: Cin, Min, SD, Mil, Sea, NYY, ChW, Hou, Bal, Cle
1: Pit, ChC, LAD, Atl, Was, Tex, SF, TB, KC, Fla

A few notes:
The starting pitcher's numbers are based on Defence independant pitching stats, projected and weighted. There are some counterintuitive categorizations there, based more on these defence independant indices and base performance value than on projected (or especially past!) runs allowed. I've found that there is value in the skill guys.

Also... beware of interleague play. I lost a season's worth of profits this past weekend. It was extremely chalky. Seriously. I'll built it back up over the next few weeks, though, but a frustrating setback nonetheless.

[blogguy]

Thanks Spraguer!

I haven't had a chance to go through yet and compare where you've got things to where I've got them...so there are bound to be some differences. I've given myself the ability to fudge a half a point here or there. And...I'm also giving myself some flexibility on the offenses depending on the park. I'll probably have Cincy as a 3 at home but a 1.5 or 2 on the road. Minnesota's home medians suggest a 3 for them at home, but they score like a 1 on the road.

Once I get a chance to go through them I'll try to use your numbers for guidance in spots where there are question marks. Perhaps I'll even do a blog version and a spraguer version each day on the chance that agreement adds some weight to the matter.

Regarding interleague play. I know it throws some people for a loop. But this particular approach did very well. As I mentioned elsewhere my contest plays down in the Baseball folder went 10-2 between the Airbets and Leo events (no duplications)...and that was kind of the tip of the iceberg for the weekend. That's why I wanted to get things up quickly on Monday for peer review. It was doing that well out of the gate.

As you said...the weekend was chalky...and this is an approach that tends to be chalky because you're just totalling up measures of superiority. Could be a misleading read long term. But, there's also the possibility that the oddsmakers aren't taking into full account the impact of offenses and bullpens when making their lines. This serves to understate the advantages. When you watch games like Cincy/Milwaukee last night, and Detroit/KC...it sure felt that way in the gut. We'll see what an extended look shows.

Thanks again for putting all of that in the 1-2-3 format. I've already worked up the numbers for today...was about to type them in to a new thread. I'll see if I can incorporate your stuff beginning tomorrow.

Oh...TOPSPINNER...thanks for sharing your thoughts. I used to play a lot of tennis before my meniscus started complaining...so I know tennis players are BRILLIANT!

It's something to think about. I think the grading system allows for some flexibility because teams on a good or bad run will often see their offensive and bullpen numbers affected. I was surprised how much the bullpen ERA's jumped around just form Friday to Monday this past week.

Also...some of what we see as streaks are just superior teams beating inferior teams the way they're supposed to. The Cubs became a 1 offense with the injury to Lee, and had a "slump" The Angels haven't been scoring much, and they've been "slumping" lately. Sometimes "hot" and "cold" is just simply clustering of results that are already encompassed in "good" and "bad." I think we're partly incorporating that issue anyway as a result.

But it is something to keep an eye on. Streaks may be "red flags" that an adjustment needs to be made with an offense or bullpen because of recent changes that aren't being reflected in the longer term numbers. Thanks for bringing it up!

[BuzzRavanaugh]

count--thats it. I thought that was pretty good back in the day.