[ABanks]

Hey guys,

I know it's only December but I am starting to work on baseball and I am trying to figure out how much fielding should be worth. Say I had a model that incorporated pitching and hitting but not fielding. In attempting to price a game on a neutral field the model basically says the two teams are evenly matched and the price should be:

Team A -105
Team B -105

Well what if I was then able to find an accurate ranking of teams' fielding abilities and discovered that Team A is the best fielding team in baseball and Team B is the worst. Team A is not the best fielding team in the last 50 years and is no better or worse than the number 1 ranked fielding teams in previous years and Team B is no better or worse than the average last place fielding team over the years.

With this new fielding information what would you make the new moneyline? I doubt there is an exact answer but what would your best estimation be? I am struggling to come up with an answer. I figure it would have to be worth at least 1 cent and probably is worth more. Maybe 20 cents? So the new price would be:

Team A -115
Team B +105

Is that too much of an adjustment? Too little? Any and all thoughts would be greatly appreciated, including suggestions on how to attack this problem. Thanks.

[flsaders85]

Quantify fielding....unearned runs is a good place to start

[Kolotoure]

There are defensive metrics which are measured in runs...

[EXhoosier10]

Quantify fielding....unearned runs is a good place to start

I think this is probably the last place to start.

[lovesbaseball1]

I don't know the answer but just wanted to point out you should be adding your adjustment to the probability that a team wins not the odds.. a 10 cent adjustment is not the same at -205 +195 as is it is at -105 -105.

[Arthur-from-LV]

Ditto what lovesbaseball1 mentions. I personally run my baseball simulator for the probability then convert that into a money line.

[Sawyer]

I think this is probably the last place to start.

Agreed.

[gryfyn1]

What aspects of pitching are you incorporating into your model? If you are basing you pitching factors off of ERA or batting models using Batting average anywhere in it you are already incorporation defense into your equation.

[sbrhedge]

Manufacturing a minimal edge with fielding is a lot of work with minimal payoff.

Start work on the pitchers.

[jetsjets1028]

just a little bit...

[Arthur-from-LV]

What aspects of pitching are you incorporating into your model? If you are basing you pitching factors off of ERA or batting models using Batting average anywhere in it you are already incorporation defense into your equation.

Yup I agree, I totally forgot about that. Although, I would add that if a pitcher goes to a different team or league, you would have to adjust his ERA to account for his new defense. Then the sample size becomes an issue, and that's where regressions need to be applied. Anyways, that's just one example of adjustment you might have to make.

Ì also agree with sbrhedge, that the payoff is fairly small for the amount of work you would need to do (hence why I use a simulator), but if you do it accurately, EVERY small edge you can use will help in the long run. This is important, because the lines are getting harder and harder to beat these day.

I would also highly recommend getting this one book: Betting on Baseball by Michael Murray 2007 edition. He has more recent ones, but I haven't read them yet.

[sbrhedge]

If you use anti-fragile systems, small edges can help .. but the larger edges are the ones that stand up year after year.

Moving a pitcher between leagues doesn't make too much of a difference depending on the type of system you use.

I am finding the lines are actually getting easier to beat over the last 5 years.

Regression analysis is often misapplied in statistics, and especially in sports betting metrics. Over-optimization is a common pitfall. Vigilance in its use is a must.

[Miz]

Manufacturing a minimal edge with fielding is a lot of work with minimal payoff.

Start work on the pitchers.

This

[TheMoneyShot]

If you are seriously thinking about wagering baseball games... I highly recommend capping the games that open -105/-105 and typically end the same way. You have a better chance of hitting these games. Than laying high juice. I typically only wagers Sunday Night Baseball games and that's it! Right after the NBA season ends... Until the NFL preseason starts. Gives you about 2 months of rest. That's typically my vacation for the year. Your mind needs a break.

[matthew919]

I agree that defense isn't worth the effort it takes to model. I'd rather wrap as much defensive info into the pitching statistic as possible, and then handicap it from there - i.e. if a top defensive team is playing a terrible defensive team, consider withholding your bet. The signal to noise is just not going to pay off for you I don't think.

[BeatingBaseball]

The evaluation of team defense and its impact on the outcome expectation of any given MLB matchup is definitely more art than science. Beyond what is reflected in the pitching data, team defense simply doesn't lend itself to mathematical modeling. Hard numbers such as errors, unearned runs and outfield assists fail to adequately reflect important variables like the range of infielders, the speed of outfielders, the respect opposing runners have for the outfield arms, etc.. Other considerations such as infield defense contributing much greater value with a groundball pitcher on the mound as opposed to a flyball pitcher further complicate the assessment.

Although team defense, a critical component in winning baseball games at the big league level, can never be ignored - I think it's one area where an old school, gut sense baseball handicapper will always have a little edge over the programmer.

[Arthur-from-LV]

I have hard time believing that using a subjective approach like "gut feeling" is an optimal strategy to handicap MLB games, and probably will get you in hot waters with your bankroll eventually. Are you gonna log your feelings for each defensive matchup and use a scientific approach to all other facets of handicapping a game? If that's what you're doing, then how would you evaluate your methodology in the long run, or fine tune it? It would be a mess, IMO.

My simulator uses the basic fielding %, far from great, but more useful then "gut feeling" for many reasons. My friend uses UZR( ultimate zone rating) in his simulator to control for defense (after all he came up with that stat). Downside to UZR is it's super complicated, but it's the cream of the crop of crop of defensive stats at this time, until something better comes along. UZR is nothing more then a refinement of basic Zone Rating.

Any given stat you choose to use is nothing more then an estimate of true talent.

[BeatingBaseball]

I have hard time believing that using a subjective approach like "gut feeling" is an optimal strategy to handicap MLB games, and probably will get you in hot waters with your bankroll eventually....

I realize the term “subjective” is a dirty word in modeling - but modeling is only a tool and is not synonymous with handicapping. As the title of this forum is Handicapper Think Tank and not Modeler Think Tank, I thought it in bounds to point out that there are simply some elements in the dynamic of professional sports competition that I believe can not be quantified mathematically, even in so statistically rich a game as baseball.

As much respect as I have for good modeling - a baseball mind is also, in itself, a rather sophisticated bit of software that should likewise be respected. The “gut” sense of one with an intuitive knowledge of the game actually does have a rational basis and should not be readily dismissed or discounted.

As I’ve been successfully handicapping baseball for over forty years, if my less than “optimal” approach is ultimately going to get me in “hot waters” with my bankroll - it better do so soon.

When it comes to handicapping team defense in MLB, by the way, I would put forward that, with rare exception, basic fielding percentage by itself has a value approaching zero. It’s about as useless as the “g” in “lasgna.”

[EXhoosier10]

My simulator uses the basic fielding %, far from great, but more useful then "gut feeling" for many reasons. My friend uses UZR( ultimate zone rating) in his simulator to control for defense (after all he came up with that stat). Downside to UZR is it's super complicated, but it's the cream of the crop of crop of defensive stats at this time, until something better comes along. UZR is nothing more then a refinement of basic Zone Rating.

Any given stat you choose to use is nothing more then an estimate of true talent.

Does your simulator also regress for the humans that subjectively choose which sharp hit bobbled ball is an error and which is an infield hit?

[BeatingBaseball]

Beyond the subjective aspect of the offical scorer's rulings - a greater challenge to statistical quantification of defense would be the hits an extraordinary defender takes away that would normally go through and be scored a hit with a less talented defender. There's no error in such situations either way, so both guys still look perfect statistically in terms of fielding percentage. In fact, the shortstop with great range can take away a ton of hits, even turn many of them into DP's, yet because he gets to so many balls he'll inevitably pick up an incremental error once in a while which can cause him to actually appear statistically weaker than a lesser shortstop. Since fielding percentage is primarily a consistency measure of one's basic ability to catch the ball and throw it accurately, it has diminishing value at higher levels of the game - at the big league level those basic skills are generally a given.

Another challenge to defensive quantification via modeling would be control of the opposition's running game. Although it can be a huge component of team defense, it's generally more significant in an NL matchup than vs an AL team. With the DH and AL style play one single base or one single run is not quite as important to stop or score. If you have a catcher with great foot work, an accurate cannon of an arm and a pop to pop of 1.7 seconds - say a Yady Molina - he's a lot more valuable vs a running club than he would be vs a station to station AL club waiting for the long ball and which doesn't steal bases anyway. A great catcher like that could also be matched with a pitcher who doesn't hold runners or who is so slow to the plate that the edge you have with your catcher's talent is totally neutralized and becomes worthless - thus any statistical edge you have factored in for him becomes no edge at all.

My point is that the "subjective" is, in these cases, the better (maybe the only) method of effective evaluation. The staff pitching stats will of course reflect team defense to some degree, but in my opinion not enough.

[Arthur-from-LV]

Exhoosier10- " Does your simulator also regress for the humans that subjectively choose which sharp hit bobbled ball is an error and which is an infield hit?"

I have no idea is to what you are trying to ask me.


BettingBaseball-" Beyond the subjective aspect of the offical scorer's rulings - a greater challenge to statistical quantification of defense would be the hits an extraordinary defender takes away that would normally go through and be scored a hit with a less talented defender. There's no error in such situations either way, so both guys still look perfect statistically in terms of fielding percentage. In fact, the shortstop with great range can take away a ton of hits, even turn many of them into DP's, yet because he gets to so many balls he'll inevitably pick up an incremental error once in a while which can cause him to actually appear statistically weaker than a lesser shortstop. Since fielding percentage is primarily a consistency measure of one's basic ability to catch the ball and throw it accurately, it has diminishing value at higher levels of the game - at the big league level those basic skills are generally a given."


I completely understand what you are getting at. I all ready stated that using fielding % to model defense is far from great. It has many essential flaws. If I had to give it a grade, maybe I would give it a D or a C-. But it's what I currently use because of it's simplicity AND because it was programmed way long before ZR and other defensive stats were readily available. I will eventually convert to ZR, but it's not as some say "worthless".


BettingBaseball- "Another challenge to defensive quantification via modeling would be control of the opposition's running game. Although it can be a huge component of team defense, it's generally more significant in an NL matchup than vs an AL team. With the DH and AL style play one single base or one single run is not quite as important to stop or score. If you have a catcher with great foot work, an accurate cannon of an arm and a pop to pop of 1.7 seconds - say a Yady Molina - he's a lot more valuable vs a running club than he would be vs a station to station AL club waiting for the long ball and which doesn't steal bases anyway. A great catcher like that could also be matched with a pitcher who doesn't hold runners or who is so slow to the plate that the edge you have with your catcher's talent is totally neutralized and becomes worthless - thus any statistical edge you have factored in for him becomes no edge at all.


My sim. does take into account for some of the running game. For example, speed of the runner(s), handedness of the pitcher on the mound, players defensive positioning based on the 24 base out situation. Unfortunately, my sim doesn't account for arm strenght and accuracy for catchers and other positions.


BettingBaseball-"My point is that the "subjective" is, in these cases, the better (maybe the only) method of effective evaluation. The staff pitching stats will of course reflect team defense to some degree, but in my opinion not enough."


I strongly disagree with your assessment BB. But, if using a subjective system to account for defense works for you....GREAT! More power to ya.. I'm not going to make claims on here wether I am a succesful handicapper, or not, because I don't take anyone else's claim(s) on here seriuosly whatsoever. I'm merely trying to share what I have and how I use it. Nonetheless, I do sometimes enjoy reading other people input (questions, comments, etc.) such as yourself. Btw, I will be posting my daily picks when this new MLB seasons starts. Maybe you will still be around and we could compare our picks.

[BeatingBaseball]

Fair enough, Arthur. Hopefully we'll both still be around and can compare analyses on some matchups.

I'll look in on your thread.

[Kolotoure]

The fact that fielding % is an awful metric doesn't mean you should ignore all statistical evaluations of defense

[BeatingBaseball]

The fact that fielding % is an awful metric doesn't mean you should ignore all statistical evaluations of defense

Kolotoure - You may be familiar with the logical fallacy of equivocation. If not, and you'd like to see an example of it, your statement is a rather good one. If the statement was intended to refute anything that has been said here, it fails to do so. The the terms "fielding percentage" and "all statistical evaluations of defense" are not one in the same. The former being only a single data point within the latter, you can't make the leap from one to the other. No one said that "all statistical evaluations of defense" should be ignored. Of course they should not. It was only posited that the value of fielding percentage at the major league level is, with rare exceptions, de minimus.

The basic argument here is whether the predictive value of statistical modeling in this specific area of the game (team defense) can at times be enhanced through some incremental application of the intuitive baseball sense of one who is familiar with both the game itself and the players involved in a given matchup. Some here believe the pure mathematical model will outperform an old school handicapper and is best left to stand alone, even when it comes to team defense. I believe, on the other hand, that a thoughtful analysis of the defensive matchup by a seasoned handicapper with a deep understanding of the game followed by any delicate massage of the win expectation (thus the price) on the overall matchup that such an analysis may indicate can add to any positive EV the modeling may provide.

Our different takes on this is quite similar to the "Money Ball" clash between the Beaniacs and the old school, tobacco juice fraternity of scouts. My question is - why can't we take advantage of the strengths of each approach?

[Arthur-from-LV]

BB how do you go about evaluating defense for a MLB player or a group of players where you have a small sample size of their overall skills? Do you simply use traditional scouting reports? Same question for evaluating offensive skills where you have a limited sample size?

[BeatingBaseball]

Arthur –

Since I believe you use a purely scientific, mathematical approach I doubt that what I have to offer will be of any practical value to you - but when I haven’t seen a guy play myself I do try to get whatever scouting information is available, tend to take all his early big league data with a grain of salt (esp any outlier level performance) and over time I'll begin to both take his stats more seriously plus develop my own personal value assessment. As I say, it’s probably too Gestalt an approach for your taste and I don’t know how one would begin to model it. One basic baseball principle relating to small sample size, of course, is that unfamiliarity in the pitcher-hitter dynamic always favors the pitcher. That you can take to the bank.

There are so many things about the game that I see as critically important but seem difficult if not impossible to effectively model. There was a time when if you really knew your pitchers, which is quite amenable to statistical analysis, that you could almost come up with +EV plays on starters alone. I believe that in recent years, however, with so much analysis of pitching and the lines so heavily adjusted for the starting pitchers that the pendulum has actually swung to the point where I often find more value in the assessment of real time team dynamics. Don’t get me wrong – starting pitching is still the biggest single component in a team’s chance to win on any given day, but the current team dynamics are huge as well. Sometimes a team is just running so well that it almost (again, almost) doesn’t matter who takes the ball. At other times, the same team could be in such a funk that they could have Koufax or Gibson out there in their prime and still find a way to lose. I imagine your simulator uses some method to overweight more recent performance, but if it isn’t doing so effectively I’d say you’re missing a lot of value. There’s probably no better example of the importance of team dynamics than the World Series outcomes the last 3 years. I had the 3 dogs almost on team dynamics alone. I also overweight managers and team leadership, esp in big games, in my assessments and a lot of what would be considered “intangibles” – travel issues, etc. Again, I don’t know how you’d go about modeling some of this.

I could give you many examples of some tendencies I'm sure you wouldn't believe to be valid, but those within the game know to be true. I'll give you one - "Latin hitters hit Latin pitchers" - how is a program going to model a little edge like that?

[Arthur-from-LV]

Sounds to me like you subscribe to the idea that each game outcome is dependent on the current situation of how each team is performing. It's as if you're trying to find anomalies, like the Latin thing you mentioned, and you choose to believe that those are somehow " found edges." My sim. doesn't need to make to those type of as you call adjusments, simply because there is no +ev to be found there. Can't be a 100% sure (probably like 98% sure), because I haven't tested your theory, but I wouldn't put any value (like money) into it. Again, it seems to me that what you seem to see as "edges" are just patterns of randomness. That's what we call noise in statistics. It's like saying that- well its raining, and it seems to me that everytime it rains so-and-so team hits a ton of home runs when that happens. Of course that's nonsense, but not far from what you're implying, I think. I wouldn't be suprised if you use current or past streaks, and "clutch performance" as some of the components to your system. If you do, well.....that has been proven to be ineffective a long time ago.

[gryfyn1]

Check jamesmpiette DOTCOM - he has an article out the accuracy of defensive predictive models; you can read get the link to his Thesis that was in JQAS on the sunject

[BeatingBaseball]

Sounds to me like...you choose to believe...it seems to me that what you seem to see as "edges" are just patterns of randomness. That's what we call noise in statistics. It's like saying that- well its raining, and it seems to me that everytime it rains so-and-so team hits a ton of home runs when that happens. Of course that's nonsense, but not far from what you're implying, I think. I wouldn't be suprised if you use current or past streaks, and "clutch performance" as some of the components to your system. If you do, well.....that has been proven to be ineffective a long time ago.

You can be pretty insulting to someone who tried to treat you civilly and respecfully, arthur. Exactly what is it with you arrogant, pseudointellectual computer geeks who believe yourself to be mathematical geniuses that drives you to be so condescending?

Don't put words in my mouth, Nimrod. And don't wild guess or imagine that I use past streaks or past clutch performance or anything I haven't said I use. Everything I do use has a rational baseball basis and is anything but nonsense. There is no rational correlation between rain and home runs. To imagine I would think so is insulting. I know what noise is. I know what randomness is. And, no matter what you choose to believe, I've been putting value (serious money) at risk doing what I do with success - not ineffectively - for a very long time. I have nothing to prove to you.

You may know a little about statistics, but statistics is not synonymous with handicapping or the game of baseball - two things about which you clearlry know squat.

[Arthur-from-LV]

I'm sorry did I miss something? Lol.... Im not being civil because I believe that your system is nonsensical, but it's civil of you to call me names like nimrod? Eshhh... I don't remember calling you any names, although I will shortly. you choose to attack me because I called you out on your loonie strategy (loonie is an understatement) Im sorry if you feel offended, but you are just another troll, imo. Good luck to you and hopefully you grow some balls, or whatever you posses.

[Inspirited]

lel

[jgilmartin]

BB, do you make your own lines for games? If so, how do you arrive at them?

[BeatingBaseball]

...you are just another troll, imo. Good luck to you and hopefully you grow some balls, or whatever you posses.

Rest assured, nimrod, I have balls. They’re titanium.

I assume the “from LV” in your name is an effort to represent that you live in Las Vegas. If you really do – stop by the invited guest curtain room at the SuperBook at the former Las Vegas Hilton, now the LVH Hotel, sometime during baseball season. Tell them you would like to meet an actual, real life, respected professional baseball handicapper. You may have to bet a few dimes to gain access, but that’s likely not an issue for a sports betting genius like yourself. If for some reason you can’t swing the action, tell them you’re doing a project for your high school statistics class – or, if that doesn’t work, I’ll get you a guest pass. That way we’ll get to meet and discuss our baseball plays. There is, by the way, even a place to plug in your computer.

Now, let me take a wild guess of my own - I would say, based on what you have revealed of yourself here, that it’s a +EV play to lay -1000 on the money line that you voted for Obama. And I did that without the aid of any computer modeling whatsoever – but simply on the fact that you are clearly a classless dipshit.

Sorry, jgilmartin. But I am done with this thread.

[EXhoosier10]

lolz at how this thread progressed

[Arthur-from-LV]

The O'Mighty genius has abandoned us. R.I.P. Beating Baseball. What are we gonna do now?