[crinkledaces]

I came across this system and was wondering if anyone thinks it would work. Or if someone could point me in the right direction as to what I can do to formulate my own line based on stats only and not hunches.



1. The first thing we will do is take all the American League teams and all the National League teams and calculate each team's average runs per game and secondly, calculate the median runs per team per league. With 14 AL teams, the median would simply be averaging the seventh and eighth highest-scoring teams. With 16 NL teams, it's averaging the eighth and ninth highest-scoring teams.
Because we're dealing with 14 and 16 teams,medians make a bit more sense than using average, as one or two teams could throw things out of sync. Since we already have calculated the averages for each team, we're not really adding much additional work, either.

2. Take the average number of runs a team allows with each pitcher and divide that by the offensive median of the league he pitches in. If the median number of runs scored per team is 4.1 and a pitcher allows 4.5 runs every time he takes the mound, the pitcher's rating is a 1.10. If our pitcher allows 3.6 runs per start, his number would be a .88, which is merely 3.6 divided by 4.1.

3. Multiply each team's offensive average by the opposing pitcher's rating to come up with the predicted number of runs for each team.

4. Subtract .07 cents from the road team and add .07 cents to the home team to account for home field advantage.

5. Take the team with the highest number of predicted runs and subtract the lower rated team's predicted runs. As each run is worth roughly 50 cents on the money line we will multiply that difference by .5 and to come up with our predicted line on the game.

Now, we'll jump to the next page and look at some examples and some basic rules.


Examples

Before we get into some rules, let's look at an few example, as the system is one of those that may sound harder than it actually is.

Example 1: If Team A averages 5.13 runs per game and is at Team B, which scores 4.58 runs per game. Pitcher A has a rating of .93 and Pitcher B has a rating of 1.22.

First, we will take Team A's 5.13 runs per game and multiply that by Pitcher B's 1.22 to get 6.26. Then we take Team B's rating of 4.58 and multiply that by Pitcher A's .93 to get 4.26.

When we account for home field, Team A's 6.26 becomes a 6.19 and Team A's 4.26 becomes a 4.33.

If we subract 4.34 from 6.18 we get 1.86, which is the number of runs we predict Team A will win by. Multiplied by .5 gives us .93, which is what we say Team A should be favored by, so our line would be Team A -193 over Team B and our predicted total on the game is 10.52.

Rules

Before we get into any rules, I have to say I have no idea how this will work. On the surface it makes sense, but in handicapping things don't always work out the way they should.

This is one of those times where I'm basically throwing an idea out there and you're free to use it as is, make your own adjustments and tweeks, add some new wrinkles, etc.

My thoughts:
1. I wouldn't use the current season's stats until one-third of the season is played, primarily due to wanting pitchers to have a decent amount of starts, so that one or two exceptionally good or exceptionally poor, doesn't influence the ratings too much.

2. For the first 54 games, you could use last year's stats, with a few minor adjustments for players gained or lost due to trades or free agency. If I was using last year's stats for the first 54 games of this season, I would only consider pitchers with at least 10 starts.

3. As the season progresses, you could use separate offensive stats for right-handed pitchers and left-handed pitchers depending on how much work you want to put into it.

4. In theory, the greater our predicted line is from the actual line, the better we should do, but we'll have to essentially wait and see if that pans out with the actual odds.

5. Finally, this will probably be a work in process and it may take a bit of time to get some wrinkles ironed out, but it should be a good starting point and get us moving in the right direction.

Thanks in advance for your time and responses

[inrenokid]

This is a small variation on the old DeCune style explaind at "about.com" ; these simple systems are out striped by contempary computer programs.

[flsaders85]

What do you do when Mike Trout has to miss 2 months due to injury? You need player-based models to beat the market.

[crinkledaces]

Any idea where I can find information on a player based model?

[James Marques]

What do you do when Mike Trout has to miss 2 months due to injury? You need player-based models to beat the market.

I disagree. If he's out of the lineup, I'm sure there is a way to measure his impact on the team. I'm no baseball expert, but I hear a lot about this "wins above replacement" stat. I use a similar stat for pro basketball, as I believe there's a huge advantage in NBA lines when a team's star player(s) is/are out.

[James Marques]

If we subract 4.34 from 6.18 we get 1.86, which is the number of runs we predict Team A will win by. Multiplied by .5 gives us .93, which is what we say Team A should be favored by, so our line would be Team A -193 over Team B and our predicted total on the game is 10.52.

I'm having a little trouble following this logic.

[flsaders85]

it sounds like you are agreeing with me...a player based model will absolutely measure the impact because he isn't in your lineup...

[crinkledaces]

does anyone know where to find a statistical baseball betting strategy that works

[James Marques]

it sounds like you are agreeing with me...a player based model will absolutely measure the impact because he isn't in your lineup...

I was actually. Must've misread something somewhere. But basically I don't think it boils down to individuals all the way. Model a team, and make sure you can account for one or two "big time" players being out. So yes, I agree.

[crinkledaces]

You agree on what? That this is a decent system? Or at least a good base?

[D W S]

Any idea where I can find information on a player based model?

Jim Jasper's Sports Betting book used something similar as the method you're describing, but instead of team averages used individual players and factored their runs per game - RBI and runs scored for his team averages - changing the line-ups depending on if the team faced right or left handed pitchers. Really, you have to weigh the time involved for player-based lineups or using team averages. You could always use a weighted system that places more emphasis on a team's last 10 to 15 games or so, which would factor in lengthy injuries, but would reduce your sample size a bit and be dependent on luck of the draw when it comes to strength of pitchers.

[crinkledaces]

Thank u for the info

[magyarsvensk]

I just skimmed this a bit (see #2) but here are a few thoughts

1. Baseball scores are not normally distributed -- not even close. Use a median-based distribution with different formulas for each tail.
2. With any static system, you can pretty much count on the handicappers already doing it and doing it better than you. Use a dynamic system with multiple variables.
3. It seems to me that the specific ballpark they are playing in would have a greater effect on the outcome than the teams or home field advantage.
4. You use the words "in theory" a lot, which is generally a bad sign when you are developing a system. The more basic and comprehensive a theory you use (e.g. "This happened this much over the last 1,000 games so it will probably turn out that way in the future"), the better chance you will have of succeeding.

[ScreaminPain]

bullpens?.......