Can Someone Confirm My Happiness?

[venu]

hi,i'm new to this site......................

[Flying Dutchman]

I don't know how you can say that without knowing anything about my model. My model doesn't tell me how much of an edge a team has, only if they have one or not. The unit size tells me the confidence that an edge exists, not the size of the edge.

So, armed with limited information, I believe my attempt to reduce vig is a valid one. My thought is that the oddsmakers will do a very good job (most of the time) converting the moneyline into a comparable runline, and vice versa.

Mathdotgrump says crap like that about most everything. Ignore him, as you'll never do it right in his eyes.

[mathdotcom]

Mathdotgrump says crap like that about most everything. Ignore him, as you'll never do it right in his eyes.

Guy is way off basis by choosing to go after the line with minimum juice instead of maximum value. Since he has not provided much detail into his model, this lapse in reasoning is enough for me to conclude he has probably not found anything and has instead data mined himself into a hole without realizing it.

[Pokerjoe]

My guess would be that if you made an error it would be using data from a date past the date of the game to find your bets.

For example, if you had data from April 1st 2009 to yesterday and then used all of that data to try to predict a game on May 1st, 2010 you would seemingly have a very successful model. It is important that you only use data up to and not including the date of the event to come up with your prediction.

It is also possible that you just have a really successful model. A win rate of 62.8% is not something I have ever heard of in baseball if your median line is +/-100 but I know it is possible in other sports.

LOL, my first guess, too. The results are TOO good. I actually did that once in NBA research, wherein I wrote an excel formula wrong such as to include, not the 6 game previous to the bet, but the 3 games previous and the 3 games following. Pretty easy to do, actually.

And there are other corruption possibilities, too. With results like this, I'd seriously double-check my work.

[Pokerjoe]

The idea is that I want to introduce as little juice as possible. There is also a ceiling of +200 for the same reason.

I'd rather bet a runline with lines of +120/-130 than a moneyline with lines of +240/-280.

This statement means you don't know how juice figures in. And it's hard to think that someone who doesn't understand such a fundamental concept has found the Holy Grail.

[MonkeyF0cker]

I don't know how you can say that without knowing anything about my model. My model doesn't tell me how much of an edge a team has, only if they have one or not. The unit size tells me the confidence that an edge exists, not the size of the edge.

So, armed with limited information, I believe my attempt to reduce vig is a valid one. My thought is that the oddsmakers will do a very good job (most of the time) converting the moneyline into a comparable runline, and vice versa.

Umm. I think what he's getting at is that it likely isn't mathematically sound, especially since you don't understand that you're not "reducing vig" in the slightest by betting lower odds.

What you've created sounds like a system rather than a model to me since you aren't capable of quantifying an edge with it.

[mathdotcom]

This statement means you don't know how juice figures in. And it's hard to think that someone who doesn't understand such a fundamental concept has found the Holy Grail.

Exactly my point but worded even better. Thank you.

[Grind-It-Out]

This statement means you don't know how juice figures in. And it's hard to think that someone who doesn't understand such a fundamental concept has found the Holy Grail.

I sort of just dismissed it when mathdotcom said it because he has a tendency of shooting down ideas for random reasons. Since other people are agreeing, I'm going to assume he's correct.

So, admitting I'm wrong, can you point out to me WHERE I am wrong?

Let's say, albeit it unrealistically, that I'm 100% sure that team A has an edge against the moneyline. I have ZERO knowledge of how large that edge is. Say the moneyline is -220 and the runline is -115. Knowing only this information, what should I do? Or, is it not possible to make a sound decision using only this information? Or is it not possible to use this as an example since it is impossible to have 100% certainty of anything?

[CrimsonQueen]

It's weird that you have a model that says you have an edge, but can't tell you how much of an edge.

Somewhere in your model, you'd think it would determine "The line is -120, and the line should be -125."
"You have an edge"

If it knows you have an edge, what number/s are determining that you do indeed have an edge? In order to help, you'd really have to show your work...

And as to my other point, someone else worded it better, it's about the value of the line, not the price you're getting. So if the ML is -250 but should be -350, that's way better than if the RL is -105 and should be -108. So they don't always directly correlate. Sure they do sometimes...but sometimes isn't a model........

[mathdotcom]

I sort of just dismissed it when mathdotcom said it because he has a tendency of shooting down ideas for random reasons. Since other people are agreeing, I'm going to assume he's correct.

So, admitting I'm wrong, can you point out to me WHERE I am wrong?

Let's say, albeit it unrealistically, that I'm 100% sure that team A has an edge against the moneyline. I have ZERO knowledge of how large that edge is. Say the moneyline is -220 and the runline is -115. Knowing only this information, what should I do? Or, is it not possible to make a sound decision using only this information? Or is it not possible to use this as an example since it is impossible to have 100% certainty of anything?

I am now going to pout and not answer your question in the second paragraph.

[jgilmartin]

It's weird that you have a model that says you have an edge, but can't tell you how much of an edge.

Somewhere in your model, you'd think it would determine "The line is -120, and the line should be -125."
"You have an edge"

Had the same thought. Also, if the model outputs number of units to wager based on a level of confidence, how is this level of confidence obtained if not due to some measure of edge?

[Pokerjoe]

I sort of just dismissed it when mathdotcom said it because he has a tendency of shooting down ideas for random reasons. Since other people are agreeing, I'm going to assume he's correct.

So, admitting I'm wrong, can you point out to me WHERE I am wrong?

Let's say, albeit it unrealistically, that I'm 100% sure that team A has an edge against the moneyline. I have ZERO knowledge of how large that edge is. Say the moneyline is -220 and the runline is -115. Knowing only this information, what should I do? Or, is it not possible to make a sound decision using only this information? Or is it not possible to use this as an example since it is impossible to have 100% certainty of anything?

It's certainly true that you don't know how large your edge is. In fact, no one ever does. The game isn't about deluding yourself into thinking you have precision, it's about working to gain relative advantage.

What people are talking about here is that at some point you have to be deciding that a team lined at +100 has, say, a 55% chance of winning by your estimate and is therefore a wager. That doesn't mean you have a 5% edge. If you build up a large enough sample size, you might find that when your model says 5% edge you actually have ~2% edge, which would still be great. But that tilde should always be there. There is no such thing as real precision in this game, only relative precision.

As far as the runline/moneyline question, understand that they are just different expressions of win chance, and neither is necessarily better than the other, and in fact there is an odds line that will make them equivalent, and that odds line is pretty much what the books offer. For one thing, moneylines ARE pointspread lines: the pointspread is 0. All sides bets are combinations of pointspreads and moneylines. All of them.

If you think you can calculate the runline expression of a moneyline/total/home-or-road fave combo better than Pinny, go for it. But you almost certainly can't.

The wisdom of which bet form to use depends only on the quality of the form's odds, not the form itself (with some consideration for bet sizing opportunities). Generally, MLB moneylines are better propositions because they have lower juice (and I'm talking about real juice, not the spread between the -/+ US-style offers), but each wager should be individually appraised.

This has all been a longwinded way of saying: use Ganchrow's half-point calculator. Actually I don't know if he even has MLB on there, but he probably does, and in any case, if he doesn't, you should have the math skills to make your own.

BTW, the part of your quote I boldprinted is an impossibility.

[MonkeyF0cker]

It sure sounds like a system to me...

[Flying Dutchman]

OK, enough yapping, here's how you test it.

Divide your db into two sections. Section 1 is "Training," you'll use it to train your model/system.

Second data set is "Validation." You'll test the model you found out of the Training section with the validation set.

The units/win% out of the validation set is going to be closer to your future units per time period/win% going forward given no modeling screw-ups or league changes.

As you have only 1 year of data, you don't have enough to get a large enough sample size to do either as there are seasonal changes in baseball and you won't even have 1 season in either set.

You probably want to get at least 3 seasons, two to train against (or at least 1) and 1 to validate against.

[pats3peat]

very interesting

[Maverick22]

if there is anyone you trust. that has a baseball database. give them ur model. let them evaluate it. and then return their results. if it matches. then ur onto something...

OR

create another model that isnt the "real thing" and ur model will come up with an answer. then give it to someone. let their model evaluate it. and if your models match...it COULD let you know ur onto something. if they differ...could be for any number of reasons(your fault or their fault)

[Indecent]

OK, enough yapping, here's how you test it.

Divide your db into two sections. Section 1 is "Training," you'll use it to train your model/system.

Second data set is "Validation." You'll test the model you found out of the Training section with the validation set.

Shuffle the order of the games before you assign train/validation sets to help the model generalize.

[mathdotcom]

I will never understand models that don't predict a fair line...

[jgilmartin]

It sure sounds like a system to me...

Think you may be right.

[DukeJohn]

I think I just created a frickin sweet model, but I'm paranoid that I did something wrong with testing that invalidates the data. If someone could verify that I tested correctly that would be hugely appreciated.

The process:
I created a model, backtested using data from April 1st, 2009 - April 30th, 2010, and then made repeated changes to the model until backtesting on that data set yielded the best result. Then, once I was confident in my model, I "forward tested" from May 1st, 2010 through yesterday.

The results (from May 1st onward):
494 - 292, +285.99 units, Z-Score = 6.05 (All bets were between 1 and 5 units)

Since I didn't backtest with the data I used for "forward testing" does that mean that these results are valid? They almost seem too good to be true, which has me worried.

I didn't read all the posts and didn't see what I am about to say, but I am sure someone had to have mentioned it, but just in case. You only "back tested" one year. You may even do good for a couple of years, but make sure you really do have something. You have the time now that MLB is coming to an end. Go back and use fresh data, meaning 2000-2008. If that is too much work, then at least 5 years. If you can make it work for all those years then you have something to look forward to in 2011.

BoL,

[slickeddie]

Give me the picks, I'll play'em.

[Flying Dutchman]

Shuffle the order of the games before you assign train/validation sets to help the model generalize.

This isn't going to work if he is using a some sort of time series method and he absolutely doesn't have enough data to see seasonality.

[Indecent]

This isn't going to work if he is using a some sort of time series method and he absolutely doesn't have enough data to see seasonality.

Both good points, it's certainly not appropriate for every model

I thought it was worthwhile to add even if it doesn't fit his model exactly. Hopefully someone else who hadn't tried that before (and uses a model appropriate for the technique, has enough data for it to be worthwhile, etc) can improve their results.