[Hounds]

Hi,

Last year I performed well in two large poker tournaments, one online and one at the Borgata, and although the winnings are not life changing after buying a car and giving Uncle Sam his cut, I do have some disposable income. My goal in sports gambling is to be able to create another source of income. Ideally, I would like to not have to work at 9 to 5 jobs again.

I've spent the last several months educating myself about sabermetrics in the hopes of being able to use this knowledge of advanced baseball statistics to successfully wager on baseball. I initially had hopes of using advanced statistical techniques (such as random forests) to datamine and then model, but after extensively researching my ideas, I quite quickly realized I simply don't possess the anywhere close to the necessary expertise. I do plan on taking a stats 101 course at my community college in the Fall. (This is a fairly big commitment as I'm 35 years old!)

At this point, I'm tracking each MLB game using a custom spreadsheet that looks primarily at starting pitching using sabermetrics. I'm happy with what I have thus far as I'm able to look at a set of numbers and quickly see what's going on with each starting pitcher on a game-by-game basis in comparison with an approximation of a dime line and the over/under total. I next need to quantify hitting, fielding, relief pitching, and perhaps eventually momentum.

Relief pitching seems to be one of the great intangibles and, hence, perhaps exploitable. I plan on creating a calendar in excel that will hopefully be fully automated as I manually import boxscores from fangraphs each day that will show reliever availablity. My main problem for now, however, deals with how much/often middle relievers, set-up men, and closers can be used without rest. Is there any sort of pitch count that can be applied? The various teams often seem to have two tiers of relievers: one used in high leverage situations and one used in low leverage games. But I still can tell yet when each is available. Again, I would like to know if there is any sort of conventional wisdom on how many pitches/games relievers are available for in a given time period. Any direct answers or a point in a direction where I could find the info would be most helpful!

This is a great forum btw.... Lots of really useful info and some obviously really smart people--

Thanks!

[dwaechte]

There's little conventional wisdom on the issue, and if there is it won't be useful. This is the type of thing where if someone does know something of value they're going to want to keep it to themselves.

Most managers don't like to throw a reliever if he's been used the previous two days, so that might be a decent starting point. Patterns that include more than the previous 3 days of use would probably entail doing original research. However, I'd hypothesize that managers would rarely, if ever, look beyond the past 3 days when making this type of decision anyways.

Major league managers tend to be complete dolts who refuse to use logic when making decisions. I'd focus less on it as a matter of actual usage patterns, and more from the standpoint of effectiveness patterns. A manager might use a reliever who threw 5 days ago, 4 days ago, and the day before, but does that represent a usage pattern where a typical reliever starts to become less effective? Does the pitcher you're focusing on fit with this or does he have individual characteristics(such as being a soft tosser) that would make your data set less reliable when projecting it onto him?

The other approach would entail not looking at specific patterns in the previous week(s), but aggregrate innings or pitches divided by # of days into the season or something like that. Based on what I know about current baseball research, I doubt this would be useful... I think most sabermetricians at this point don't think overusage over the course of the season has too damaging of an effect on a pitcher's health or performance.

Really, you can see this will become a fairly extensive project, but could certainly be of value if done properly.

[Hounds]

thanks for the reply dwa...

I think most sabermetricians at this point don't think overusage over the course of the season has too damaging of an effect on a pitcher's health or performance.

i actually stumbled across an article about this just a couple of days ago on the hardball times. the article was slightly different, however, in that hypothesized and tested the idea of decreasing effeceincy when reliever pitched multiple innings in the same game.

I'd focus less on it as a matter of actual usage patterns, and more from the standpoint of effectiveness patterns.

i dunno.... i'm still kinda stuck on this idea--stuck to the point where i'm going to go through the game logs on fangraphs on a team by team basis in order to try and discern particular team/manager tendencies in relation to relievers. toronto's game logs gives me hope that i may unearth some potentially profitable patterns. (not for toronto necessarily as all their relievers including the closer position are filled by pitchers who seem to be pretty much average across the board including their closer(s).)

Does the pitcher you're focusing on fit with this or does he have individual characteristics(such as being a soft tosser) that would make your data set less reliable when projecting it onto him?

very cool ideas, but a couple of steps down the road for me. it's very very easy to become overly stimulated in regard to baseball statistics because there's so f------- many of them and so many different directions to explore.

anyways, time for me to stop blabbin' and get to work. now that spring is here i have to split my research time with outside chores. bleh..... again, thanks for the ideas!

[runnershane14]

check out what covers does for reliever availability, might help

covers.com/pageLoader/pageLoader.aspx?page=/data/mlb/matchups/g4_bullpens_3.html

[Thremp]

How much effort and how much edge do you think this will afford you?

[Hounds]

How much effort and how much edge do you think this will afford you?

both yet to be determined...

[waiverwire]

More detailed info (including pitch counts for each day) than what Covers has is here: http://dailybaseballdata.com/cgi-bin/bullpen.pl

[Ian]

IMO most handicappers vastly overrate bullpens. It would be silly to completely ignore them, but if you did you wouldn't be all that far off. If I were you I would concentrate on more important things in your capping methodology. Here's why:

-There's a strong negative correlation between winning games and bullpen usage. If a team's starter throws a complete game, they are almost certainly going to win. If a team's bullpen has to throw 160 pitches, they're going to have to score a ton of runs to win. So while it helps a team to have a fresh pen, if the team has to take advantage of that fact they are going to have to climb uphill to win.

-There's no way to tell if a team's best relievers will be used in high leverage spots. You might know that a team will bring in its closer in the 9th in a save situation, but if the team is leading by 3 runs the difference in win EV between an elite closer and Joe Average is almost nothing. On the other hand, if it's a tie game in the 5th inning with 1 out and the bases loaded, the difference between an elite closer and Joe Average is huge. However, despite being a high leverage situation, no manager in baseball brings in his closer in this spot, but a good many of them will send out Joe Average.

-Even if you somehow did know how many high leverage situations a team would face, and you also somehow knew which relievers would be used in those situations, you probably wouldn't know how good most of the relievers in the pen actually are. Compared to starters, relievers have tiny sample sizes. As a result, fluke seasons are so common that they're almost the norm. (As an aside, as a SABR guy you've probably run across articles stating that teams shouldn't spend a lot of money on bullpens. It's for the same reason. GMs won't be able to tell how good guys are because of small sample sizes, so it's smarter to get somebody for cheap with decent short term results than to give a huge multi-year contract to a guy who had a good run over a very small sample.)

I used to cap bases full-time, but I quit because it was too much work for the return I was getting. I still make a wager from time to time though, and a good number of times my bets are on teams the public is fading because everyone knows their pen has been used a lot lately. Getting the Mets as a dog last Monday is an example of this.

[Doug]

good post

[Dark Horse]

Just play the first five innings. It could save you untold amounts of headache. Few things are more annoying than to see a perfectly fine win, that took a couple of hours to build, butchered in minutes by semi-amateurs who were never good enough to start at this level.

Or bet against a bullpen if a book offers the last 4 innings. That way you can keep starters and bullpen more or less separate. Fading the D-backs bullpen this season, for instance, is guaranteed money in the bank.

[Justin7]

Ian,

Bullpens are... about 1/2 as important as the starter. They are still important though.

If you are good at analyzing bullpens, you can do very well in live betting.

[Ian]

Ian,

Bullpens are... about 1/2 as important as the starter. They are still important though.

If you are good at analyzing bullpens, you can do very well in live betting.

I've never done live betting (I only bet sides, totals, and the occasional prop), but I would guess you're right about live betting.

For betting sides and totals, all I can say is that when I was capping hardcore I weighted bullpens much less than 50% of the starter's value in the large majority of cases. There's more than one way to skin a cat in handicapping, so its certainly possible that some people gain an edge by heavily weighting bullpens, but generally speaking I still think the typical bettor greatly overrates their importance.

And to reiterate a point I made in my first post, it would be silly to totally ignore bullpens.