Total Bases Props

[rjp]

I've got no familiarity with Pecota, so I can't speak on the subject. [img]/images/graemlins/smile.gif[/img]

[Thremp]

[ QUOTE ]
I've got no familiarity with Pecota, so I can't speak on the subject. [img]/images/graemlins/smile.gif[/img]

[/ QUOTE ]

Its just a projection system.

[gtrunner57]

I think my projections come from something very similar to Pecota.

For tonight I like:
Cabrera -1/2 over Byrnes at +115
Wright +1/2 over Bay at -160
Kent -1/2 over Berkman at +130
Ordonez +1/2 over Rodriguez at -150
Guerrero -1/2 over Wells at +120

[bevo_stevo]

I remember looking at these types of bets several weeks ago and deciding that they were just too hard to attempt right now. The reason is that handicapping these bets is very similar to handicapping the entire game.

I'm curious what you're basing your stats on - are they just this seasons numbers? Do you account for lefty/righty splits?

You might also try using a stat other than WHIP since a walk would correspond to 0 bases and a hit to at least 1 base - maybe by seperating out the hits and walks, or using something like batted balls put in play.

Can you also post the record as we go along?

[gtrunner57]

I would not agree that handicapping these bets is very similar to handicapping the game (even though I do handicap games on my own). The thing is, it doesn't directly matter what you think the ERA of the opposing team is or how many runs you think the batter's team will score. Also, even if they do have a very small amount of correlation, I don't see how it matters, unless you're implying they are so related that it would be easy for 5dimes to post good total bases lines assuming that they have good moneyline lines, which I don't believe to be true (that it would be easy to post one line given another).

Like I said, these numbers are similar to Pecota's method.

I do not account for lefty/righty splits, although I would concede that this could have a significant impact. The one problem for this is that lefty/righty matchups do not affect every hitter equally or even have the same effect from every pitcher. I would need to create at least two new statistics for every player in MLB. I have just really not taken the effort of creating the statistics or the time of researching how and what affects these types of matchups.

Maybe I could have been more clear, but the WHIP is used exclusively for affecting the projected number of plate appearances, as it has a very high R squared with PA allowed (as you can imagine).

The record going into tonight's games is:
5-6 for -4.7u

[bevo_stevo]

[ QUOTE ]
I would not agree that handicapping these bets is very similar to handicapping the game (even though I do handicap games on my own).

[/ QUOTE ]

From my previous comment I just meant that if you did this analysis across all the batters you would have a pretty good projection for how the game should turn out (adding in fielding, walks, etc.)

[ QUOTE ]
Like I said, these numbers are similar to Pecota's method.

[/ QUOTE ]

The parameters (inputs) are similar right? Or did you mean that you are creating some sort of correlation between past matchups and the players you are handicapping, and combining them to get a total bases prediction?

[ QUOTE ]
Maybe I could have been more clear, but the WHIP is used exclusively for affecting the projected number of plate appearances, as it has a very high R squared with PA allowed (as you can imagine).

[/ QUOTE ]

Oh, OK that makes much more sense.

One other thing that I might point out is that with this particular type of bet that its difficult to shop around at other books, which can cut down on your edge quite a bit.

[Thremp]

bevo_stevo,

Not exactly. These are derived very easily from stats. You shouldn't be using past matchups at all. It'll give you terrible results.

I'd actually imagine you can get some very solid lineshopping done on this if you know the places to go and look around a bit.

[bevo_stevo]

[ QUOTE ]
Not exactly. These are derived very easily from stats. You shouldn't be using past matchups at all. It'll give you terrible results.

[/ QUOTE ]

The question was more about the process that the OP was using. PECOTA (from what I can tell) uses a correlation coeffecient in conjunction with previous player performances (not necessarily just the player you are handicapping) to create a prediction for future performances. I was just wondering if OP was using a similar method. Incidentally, Netflix uses a related method to predict which movies you like - so its a very well established way to go about things.

[ QUOTE ]
I'd actually imagine you can get some very solid lineshopping done on this if you know the places to go and look around a bit.

[/ QUOTE ]

Instead of flapping my gums I actually looked around a little bit, and you're right.

[calmB4storm]

[ QUOTE ]
I do not account for lefty/righty splits, although I would concede that this could have a significant impact.

[/ QUOTE ]
This is very important IMO.

Just a few examples...
http://tinyurl.com/yt9cel
http://tinyurl.com/37ymwl

I definitely think you need to account for this.