Baseball Total Bases

[DougOzzzz]

I decided to do some testing on MLB Total bases lines. I set up a spread sheet which is linked to PECOTA projections for all major league players this year.

Here's how it works:

First, you input the two hitters, and the opposing pitchers. The spreadsheet then looks up the expected 1B/PA, 2B/PA, 3B/PA, HR/PA for both hitters. It adjusts both numbers based on park effects - using essentially the same guideline that PECOTA used when generating the stats in the first place (PECOTA projects EqBA, EqOBP, EqSLG - park neutral stats that can be compared to the actual stats to see how it adjusts for park effects). After adjusting the stats so that they're park neutral, it then adds in the park effect of the home team (doubling the effect that PECOTA used, since half a players games are not at home), adjusts for the opposing pitcher's projected Eq H/9 And Eq HR/9, and adjusts for batter pitcher matchups (i.e. L vs L, R vs. L, etc.). Unfortunately, I don't have access to hitters vs. R, vs. L splits in a nice spreadsheet or database form, so I decided to use the totals for ALL players over the last 20 years.

After the numbers are generated, you can then run the macro which calculates the likelihood of 0 to 20 bases in an entire game. It then multiplies the probability of each potential matchup to calculate the overall likelihood of each batter getting more total bases, or a tie. For example, P(Batter1 > Batter2) = B1(1)*B2(0)+B1(2)*B2(0)+B1(2)*B2(1)+B1(3)*B2(0) etc.

There are a few flaws with the system. First of all, I currently don't have any system to project PAs/game. For now I am just giving slightly more PAs/game for the hitter on the visiting team - but adjustments based on the rest of the team and lineup position SHOULD be made as well. Also, the projections are based on stats accumulated BEFORE this season. While I think they are much more accurate than using current season stats at this point - some adjustment should also be made. I already mentioned the failure to adjust for each players splits.

It's pretty much just in the testing phases now, but with the proper adjustments I think it could be valuable. I did a test run today on all of the total bases matchups at WSEX. Here is a chart generated by my spreadsheet

Hitter Spread WinPct Need Actual Opposing Pitcher
Pujols, Albert -0.5 0.472 112 120 Harang, Aaron
Dunn, Adam 0.5 0.528 -112 -150 Marquis, Jason

Abreu, Bob 0.5 0.592 -146 -150 Myers, Brett
Beltran, Carlos -0.5 0.408 146 120 Glavine, Tom

Cabrera, Miguel -0.5 0.417 140 130 Leiter, Al
Jones, Andruw 0.5 0.583 -140 -160 Hampton, Mike

Lee, Derrek -0.5 0.38 163 120 Zambrano, Carlos
Lee, Carlos 0.5 0.62 -163 -150 Capuano, Chris

Bay, Jason 0.5 0.603 -152 -160 Wells, Kip
Bagwell, Jeff -0.5 0.397 152 130 Astacio, Ezequiel

Alfonzo, Edgard 0.5 0.552 -124 -160 Rueter, Kirk
Glaus, Troy -0.5 0.448 124 130 Halsey, Brad

Helton, Todd -0.5 0.427 134 130 Wright, Jamey
Giles, Brian 0.5 0.573 -134 -160 Redding, Tim

Vidro, Jose 0.5 0.571 -133 -160 Day, Zach
Kent, Jeff -0.5 0.429 133 130 Weaver, Jeff

Wells, Vernon 0.5 0.594 -147 -150 Towers, Josh
Tejada, Miguel -0.5 0.406 147 120 Cabrera, Daniel

Ortiz, David -0.5 0.404 148 NA Maroth, Mike
Young, Dmitri 0.5 0.596 -148 NA Halama, John

Jeter, Derek -0.5 0.397 147 100 Brown, Kevin
Crawford, Carl 0.5 0.603 -147 -130 Waechter, Doug

Sweeney, Mike 0.5 0.601 -151 -170 Anderson, Brian
Konerko, Paul -0.5 0.399 151 140 Buehrle, Mark

Crisp, Coco 0.5 0.558 -127 -160 Westbrook, Jake
Hunter, Torii -0.5 0.442 127 130 Mays, Joe

Teixeira, Mark 0.5 0.604 -153 -160 Young, Jason
Chavez, Eric -0.5 0.396 153 130 Saarloos, Kirk

Guerrero, Vlad -0.5 0.445 129 130 Lackey, John
Suzuki, Ichiro 0.5 0.555 -129 -160 Meche, Gil

For example, the first line says that Albert Pujols has a 47.2% chance of getting more total bases than Adam Dunn, the break even point for him is +112, and the actual line is +120.

The results are fairly consistent with the actual lines. Eight of the 15 matchups fall in between the 30 point spreads. Five of the other 7 identify edges of less than 10 points. The spreadsheet gives a solid edge to Carlos Lee (+0.5, -150) over Derrek Lee - an edge of about 5%. However, Derrek Lee has greatly outperformed his projection thus far, and Carlos has underperformed his - thus I wouldn't recommend this bet.

That leaves Carl Crawford (+0.5 -130) vs. Derek Jeter. This line is probably the one I would have singled out had I looked at the results myself. Crawford lacks home run power, but gets alot of doubles and an absurd amount of triples. He also walks infrequently. He's been worse than Jeter thus far, but Jeter's year hasn't been that good and it is being driven by batting average. The edge on this bet is just under 7% - and is probably slightly less than that. It might be worth a small sized bet, but looks to be the only matchup worth playing today. I'll keep running this all year and track the results in this thread.

[mrbaseball]

[ QUOTE ]
The results are fairly consistent with the actual lines

[/ QUOTE ]

I'd guess they set the lines very similarly. I found the same thing when I set up my totals system. I found how accurately my system predicted the actual line.

None the less this is very interesting stuff. I think the real key is to look for discrepancies based more on current form rather than projection like in the Lee/Lee instance. Then again you are a "projection" guy and I'm a "current form" guy.

I'd be real interested in seeing what kind of edges you uncover. I never make these kind of bets but wouldn't mind if I felt there was an exploitable edge available.

[DougOzzzz]

I just realized I had the pitching matchups reversed in several cases. The last column is supposed to show the OPPOSING pitcher. I don't have time to rerun the results, but I did test it on Jeter/Crawford again. Crawford's edge grows slightly to 7.02 percent.

[DougOzzzz]

Yesterday 1-0
YTD 1-0

Fixed a few minor bugs for today's test. Results should be somewhat better - still needs some work though.

Here's a link to a page with a nice table for today's output: Total Bases May 3, 2005.

Crawford is given an even nicer line (-110) than yesterday. He's struggles somewhat against lefties, hitting at around 100 points of OPS less during his career, a bit worse than average for the typical left-handed hitter. However, he's up against a rookie pitcher and PECOTA is not too friendly to Sean Henn. Jeter has been hot, but as I said before, it's mainly due to a high batting average and excellent plate discipline. The edge given by my spreadsheet is 15% - I'd guess it was closer to 10.

One other matchup that looks to be profitable:

Jason Bay (+0.5, -130) vs. Jeff Bagwell (-0.5, EVEN) at Houston, with Clemens going against Mark Redman.

There's no reason for Bay to be getting half a run and still be close to even odds. He's been better this year so far, his PECOTA projections are better. Both guys hit #3 in the order, but Bay is on the road thus should average more PAs/game. The line seems to be entirely based on the pitching matchup. Even adjusting for the pitching, Bay is given a 60.5% chance to tie or outscore Bagwell. The projected edge on this is 9.2%.

Vernon Wells was also given an edge against Tejada - but this is the type of hot/cold matchup that makes PECOTA's projections not too useful.

[DougOzzzz]

Yesterday (1-0)
YTD (2-0)

Crawford came through. Bay had 5 total bases, but Bagwell sat out so the bet was cancelled.


Here's a link to today's table: Total Bases 5/5/05
Today - wait, did they just steal my program to set the lines? Looks that way. In fact, the Jeter/Crawford match up went from -120/-110 to a whopping +120/-150 today! After 2 straight days of significant edges on Crawford, my spreadsheet says this line is almost exactly where it should be. Not only that, according to my spreadsheet not a single bet has any edge whatsoever - and only 1 bet (Abreu +0.5 against Beltran) is even close.

Oh well... no bets today.

[RaDiCaLReeD]

keep this going! I will enjoy following this!

[RaDiCaLReeD]

who offers total bases match-ups? I use sportsbetting.com and they dont seem to offer it. [img]/images/graemlins/mad.gif[/img]

[DougOzzzz]

I'm getting all my lines from WSEX . If anyone knows of other sites that offer lines on total bases matchups, please let me know.

Glad to see you're interested in this. Hopefully in the future there will be more exploitable lines than there are today, though I am skeptical.

[TheRover]

You inspired me to take a closer look at some prop bets...what do you make of this?

Carl Crawford (D-Rays) Will He Record a HIT - Must Start
05/05/05 Yes -300
19:15 ET No +220

[DougOzzzz]

[ QUOTE ]
You inspired me to take a closer look at some prop bets...what do you make of this?

Carl Crawford (D-Rays) Will He Record a HIT - Must Start
05/05/05 Yes -300
19:15 ET No +220

[/ QUOTE ]

Running this through my spreadsheet gives a 23.87% chance of 0 total bases (no hits). But it's really pretty simple: Crawford throughout his career has got a hit in 27.15% of plate appearances - he hasn't got a hit in 72.85%. He is a leadoff hitter - less than 4 PA's would be extremely rare. So raise .7285 to the 4th power and you get .2816. Add in the following factors:
1. there's a decent chance he'll get at least 5 PAs
2. He's facing a right handed pitcher (his BA is 34 points higher against rightys than lefties over his career)
3. The pitcher he's facing is a rookie, and a control pitcher who will probably give up a few more hits than normal.
4. I used his career stats - he's actually been better than that recently (last 2 seasons).
With all these factors, you get a slightly profitable bet on the "Yes" side.