Beating the closing line in baseball: How important is it?

[chunk]

This Hutennis guy is one sharp cookie. There was also a study done in the Outback down in Australia on head to head kangaroo boxing. It left no doubt that very few can win at any sport long term.

[Carl-Haakon]

I see, I'm looking for it now. What chapter?

Section 4.2: "The Model used for Prediction and Evaluation of the Strategy"

[EXhoosier10]

This Hutennis guy is one sharp cookie. There was also a study done in the Outback down in Australia on head to head kangaroo boxing. It left no doubt that very few can win at any sport long term.

Hahah this guy is a clown

[hutennis]

Here's a valuable hint that most people do not appreciate until they have attempted this work, and have been successful at it: your line does not need to be sharper than the CRIS opener to be profitable. Another way to think about it is that books must set lines for every game. The book's lines, on average, may be more accurate than mine. In fact, they are more accurate than my lines, by a small margin, on average. And the closing lines are even sharper than mine. But that doesn't matter, because all I have to do is find the ones that are wrong. And I can. In fact, despite my lines not being as sharp (on average), than the market, I can turn a profit even on razor sharp closers.

This sounds like some kind of voodoo magic to me.
If you are constantly behind, by what means are you getting ahead at the end?
Transcendent meditation? Imagining reality?
Do you connect to the power of the Universe using "quantum isolation" to help your imagined "sharp lines" to become truly sharp?
A bit too much of Dr. Chopra?

If you take that attitude that you are a failure, then you will fail. More likely, you will never even try, which is far worse.

This must be too much of late night Tony Robinson bullshit.

[Carl-Haakon]

This sounds like some kind of voodoo magic to me.
If you are constantly behind, by what means are you getting ahead at the end?
Transcendent meditation? Imagining reality?
Do you connect to the power of the Universe using "quantum isolation" to help your imagined "sharp lines" to become truly sharp?

That, or he could just be orthogonalising information to the oddsmakers lines to see if it's properly accounted for. For someone who always reminds others about the dangers of cognitive biases, you're pretty barreled in.

You also made the exact same mistake that you accused me of doing: you misinterpreted the study you quoted to seem like it agreed with your point of view, when it in fact opposed it.

[hutennis]

Hutennis, please forget the words used in the study (since they are the source of our miscommunication here) and just focus on what the authors are actually trying to convey: they managed to make returns of 7.7 percent per bet over a sample of 744 bets. Would you consider this to be evidence for your definition of weak form, or evidence against it?

We can not pay attention only this part

they managed to make returns of 7.7 percent per bet over a sample of 744 bets.

without also paying attention to this part

Although we have some support for the alternative hypothesis we
cannot exclude the possibility that the excess return is due to chance.

Again, good old conformation bias at work.

[hutennis]

That, or he could just be orthogonalising information to the oddsmakers lines to see if it's properly accounted for. For someone who always reminds others about the dangers of cognitive biases, you're pretty barreled in.

That or he could be just getting lucky for a while. "Orthogonalising information" sounds really sophisticated but I kind of lean toward a simpler explanation as a rule of thumb.
I don't see a any bias here. Could you point it out?

You also made the exact same mistake that you accused me of doing: you misinterpreted the study you quoted to seem like it agreed with your point of view, when it in fact opposed it.

I already answered that in above post.

[Carl-Haakon]

We can not pay attention only this part



without also paying attention to this part



Again, good old conformation bias at work.

You're misconstruing my position; let me reiterate from one of my previous posts:

"Anyway, if you kept reading that paper (available here, for anyone interested), you'd see that the researchers easily found selection criteria which produced profits over their small sample, and that they found no reason to reject the idea that it is possible to make a profit by using publicly available information."

They had a small sample, which wasn't fully conclusive. But the point is not that the sample is conclusive evidence for market inefficiency (it's not), but rather that if it should be interpreted as anything, then it probably shouldn't be as evidence for your position.

[matthew919]

Beating closers by 7-8 cents on average over hundreds of plays this year using nothing but historical data. Exactly the same same BTCL rate on validation set consisting of almost 10,000 samples, with corresponding ROI of 5-6%. All using statistically rigorous validation methods.

Clearly nothing to do with my mathematics and methodology. Good thing I'm so good at voodoo.

To quote one of my favorite internet comment sections I've ever read (see: http://baseballanalysts.com/archives...ou_shouldn.php), "there is a bit of the 13 year old virgin lecturing on sex going on here."

Here's a pop quiz: I have two estimates of an unseen parameter, call the estimates x1 and x2. Each estimate was drawn from a separate distribution centered on the true value of the unseen parameter. One of the distributions, however, has a slightly higher variance than the other. Lets say v1 = 1.1*v2, for the sake of argument. The two estimates that you draw differ by a "large" amount (for the sake of stupidity, lets just say |x1-x2| = 2*(sqrt(v1)+sqrt(v2) ). In which segment of the number line would you guess the true parameter lies?

Also: Do you believe that when an MLB totals line jumps a full run in the course of 24 hours, that the movement is attributable solely and exclusively to some "new" information that was not present before the line was released? Or could it be that the original opener was awful, and got hammered by people like me?

[Carl-Haakon]

That or he could be just getting lucky for a while. "Orthogonalising information" sounds really sophisticated but I kind of lean toward a simpler explanation as a rule of thumb.

Funny, seeing as how you quoted a study that orthogonalizes information to oddsmakers lines in order to see whether market inefficiencies exist. Please answer me this, and it's really the last question I have for you: did you read that study before you quoted it?

[hutennis]

You're misconstruing my position; let me reiterate from one of my previous posts:

"Anyway, if you kept reading that paper (available here, for anyone interested), you'd see that the researchers easily found selection criteria which produced profits over their small sample, and that they found no reason to reject the idea that it is possible to make a profit by using publicly available information."

They had a small sample, which wasn't fully conclusive. But the point is not that the sample is conclusive evidence for market inefficiency (it's not), but rather that if it should be interpreted as anything, then it probably shouldn't be as evidence for your position.

It can not be interpreted as anything. It should only be interpreted as no evidence for market inefficiency. Period.
"Not guilty" can not be interpreted as "innocent". It means only one thing - not guilty.

[hutennis]

Funny, seeing as how you quoted a study that orthogonalizes information to oddsmakers lines in order to see whether market inefficiencies exist. Please answer me this, and it's really the last question I have for you: did you read that study before you quoted it?

For the purpose of this discussion it is irrelevant. After finishing whatever orthogonalizing they had they drew a conclusion.
And conclusion is weak form stays efficient. And that all that matters.

[Carl-Haakon]

It can not be interpreted as anything. It should only be interpreted as no evidence for market inefficiency. Period.
"Not guilty" can not be interpreted as "innocent". It means only one thing - not guilty.

In this case, it's as if you're on a jury and interpret "Not guilty" as "Guilty" after the defense has presented a mostly but not fully conclusive alibi. I'm done with this discussion, but I would still like to hear whether or not you actually read that study from Örebro U

[hutennis]

Beating closers by 7-8 cents on average over hundreds of plays this year using nothing but historical data. Exactly the same same BTCL rate on validation set consisting of almost 10,000 samples, with corresponding ROI of 5-6%. All using statistically rigorous validation methods.

Clearly nothing to do with my mathematics and methodology. Good thing I'm so good at voodoo.

And I have a pet pocket dragon breezing green fire.

In an absence of very hard evidence our claims are almost equally serious.
A said almost b/c your claim has an explanation that does not require any evidence. LUCK

[hutennis]

In this case, it's as if you're on a jury and interpret "Not guilty" as "Guilty" after the defense has presented a mostly but not fully conclusive alibi.

Bad, twisted analogy.
The right one would be

Prosecution claims "gulty" of one count of inefficiency but can not prove it beyond a reasonable doubt.
Therefore the verdict by jury is "not guilty" of inefficiency. Whether it is "innocent" on efficiency is not relevant.

[matthew919]

The p value for this year alone for rate of line movement accordance is 1.795615e-07, under a binomial test for significance. For a sample size of 10,000 plays on my validation set it would be infinitesimal. But there I go with all that voodoo again.

[hutennis]

The p value for this year alone for rate of line movement accordance is 1.795615e-07, under a binomial test for significance. For a sample size of 10,000 plays on my validation set it would be infinitesimal. But there I go with all that voodoo again.

No. you are simply making unsubstantiated claims again.
And since what needs to be done to substantiate them can not be done, you should not make claims

[EXhoosier10]

The p value for this year alone for rate of line movement accordance is 1.795615e-07, under a binomial test for significance. For a sample size of 10,000 plays on my validation set it would be infinitesimal. But there I go with all that voodoo again.

No. you are simply making unsubstantiated claims again.
And since what needs to be done to substantiate them can not be done, you should not make claims

Great response, Hu. Guy provides actual statistics that show that his methodology is almost surely not due to luck, and you say his numbers are unsubstantiated.

Please end this conversation as Hu is nothing but a huge troll

[hutennis]

Great response, Hu. Guy provides actual statistics that show that his methodology is almost surely not due to luck, and you say his numbers are unsubstantiated.

Please end this conversation as Hu is nothing but a huge troll

land I provide actual description of my pet pocket dragon who breezes green fire with exactly the same substantiation - post on a public Internet forum.

[MonkeyF0cker]

The entire quant finance community is a fire-breathing dragon living off of magic pixie dust.

Because hutennis says so.

[hutennis]

The entire quant finance community is a fire-breathing dragon living off of magic pixie dust.

Because hutennis says so.

LOL. Thats funny.

First. I have Never said so. It is not a conformation bias for a change, but another good old staple of irrationality - "straw man"
If you can beat the argument - dream up a distorted version of an argument a attack that.

Second, We are not talking financial markets here, but if you wish...

The entire quant finance community is rushing to move offices around the world closer and closer to where exchanges are not b/c they are counting additional milliseconds it takes them to get NEW information and their edge depends on it, But b/c they want to be closer to public libraries where they can get hold of 10 year S&P charts.

Yeah, right. Nice try.

Look, It does not take that much, really.

Instead of this

The p value for this year alone for rate of line movement accordance is 1.795615e-07, under a binomial test for significance.

just start posting your openers well ahead of of pinny openers and this whole efficiency - weak - strong - analytically - prediction - past - new -
- significance bullshit will be sorted out at no time.

Otherwise, I do have a pet pocket dragon who is breathing green fire. Not just simple fire. Green fire.

[matthew919]

LOL. Thats funny.

First. I have Never said so. It is not a conformation bias for a change, but another good old staple of irrationality - "straw man"
If you can beat the argument - dream up a distorted version of an argument a attack that.

Second, We are not talking financial markets here, but if you wish...

The entire quant finance community is rushing to move offices around the world closer and closer to where exchanges are not b/c they are counting additional milliseconds it takes them to get NEW information and their edge depends on it, But b/c they want to be closer to public libraries where they can get hold of 10 year S&P charts.

Yeah, right. Nice try.

Look, It does not take that much, really.

Instead of this



just start posting your openers well ahead of of pinny openers and this whole efficiency - weak - strong - analytically - prediction - past - new -
- significance bullshit will be sorted out at no time.

Otherwise, I do have a pet pocket dragon who is breathing green fire. Not just simple fire. Green fire.

I do in fact generate my plays in advance, and yes, I do post them on the internets. I record both my WL and line movement stats as the season progresses. I did it at this forum for about 80 plays this year, but moved it elsewhere, mainly because I'd rather share my plays with people who I respect, and who also actually win at gambling (i.e. not you). I'm sure you can find my thread at this site if you try hard enough. Only 80 plays or so, so the p value for that would "only" be in the realm of 0.0001. I just got lucky I guess.

[chunk]

Considering that Hu maintains that only 1% of sports bettors make it, I think that they should get jackets and fly colors like the bikers do. After a large enough sample size with documented success, you could get a 1%er patch. That way when Hu or anyone else causes trouble, all the 1%ers come in and kick the shit out of 'em.

[hutennis]

I do in fact generate my plays in advance, and yes, I do post them on the internets..

WOW.

If you claim being able to PREDICT, you must make PREDICTION IN ADVANCE.
Who cares about PAST records. Any record can be generated in minutes.
I feel embarrassed even mentioning such an obvious thing.

Start posting your exact criteria for taking a bet in each game BEFORE lines open.
For example
ABC vs XYZ. IF O/U is <=9 and Over is dog or Par (or you can specify the odds) I take Over.

That's the only way to have a controlled experiment.

Otherwise, any conversation about successful record is simply intelligence insulting nonsense.

I understand, there are plenty of gullible, easily impressionable people out there who can eat this kind of crap,
but not everyone.

[matthew919]

I will do this, for one day only. You'll notice that the lines are NOT up for tomorrow- not at CRIS, and not at BetOnline (which serves up godawful openers on the regular). I will post the model output for tomorrow's plays, in condensed format. And not to shut you up, because clearly nothing will make a believer out of you. But rather for others who intend to follow in these tracks and are looking for guidance and assurance that it is possible.

For me, posts from the likes of Ganchrow and mathdotcom on this forum were very helpful as I learned. So here we go. You ready, hu???

Model is running now...

[hutennis]

I will do this, for one day only. You'll notice that the lines are NOT up for tomorrow- not at CRIS, and not at BetOnline (which serves up godawful openers on the regular). I will post the model output for tomorrow's plays, in condensed format. And not to shut you up, because clearly nothing will make a believer out of you. But rather for others who intend to follow in these tracks and are looking for guidance and assurance that it is possible.

For me, posts from the likes of Ganchrow and mathdotcom on this forum were very helpful as I learned. So here we go. You ready, hu???

Model is running now...

For one day only?
Do you think it is going to be enough to draw statistically significant conclusions?
Is that a joke?
Never mind me, but is it supposed to make a believer out of anybody who is in a right mind?

What is the point, really?

[matthew919]

Whoops, nevermind. BOL just opened before my model finished running. Oh well.
Edit: I'll post it anyway in a few minutes though, once it's done running.

Also, since when do statistically significant results mean anything to you? If you find my thread, you can independently verify that the plays I made beat closers at a rate that is far beyond any statistical doubt. But for some reason, you also need the output that the model generates before the lines open, or else those results are invalid? That doesn't even make sense.

[hutennis]

Also, since when do statistically significant results mean anything to you? If you find my thread, you can independently verify that the plays I made beat closers at a rate that is far beyond any statistical doubt. But for some reason, you also need the output that the model generates before the lines open, or else those results are invalid? That doesn't even make sense.

Statistically significant results mean anything to anybody who is in right state of mind ONLY when results are obtained by a CONTROLLED experiment.

Independently verify? You mean take you word for it? Take your word that you posted actual results without any alterations and the methodology you used is free of biases, manipulations, mistakes etc?
Since when "taking word" became "independent verification"? Maybe in a world of internet forums it did, but not in a real word, my friend.
Imaging scientist calling a press conference and announcing "I discovered cold fusion! Take my word for it!" Silly, is not it?

But for some reason, you also need the output that the model generates before the lines open, or else those results are invalid? That doesn't even make sense.

Please, think about what you are saying. It is beyond ridiculous.

[matthew919]

Here's the data for tomorrow. The right columns give the confidence intervals I use to bet. Just adjust the O/U by 0.25 runs for each 10 cents- so o8.5 -110 becomes 8.75. If this falls outside the interval, bet the over. I voided the CHC game due to wind, and I highlighted the games I have my eye on once the "actual" market opens, based on the horrible openers that BOL is currently give. Feel free to check where they are at right now, and then watch where all of those lines end up by the time the market closes. Cheers!

Date	HOME_TEAM_ID	AWAY_TEAM_ID	BET_OVER_IF_LINE_IS_BELOW	BET_UNDER_IF_LINE_IS_ABOVE
20130607	CHC	PIT	6.907413	8.207413
20130607	WSH	MIN	7.193659	8.493659
20130607	TOR	TEX	8.2875	9.5875
20130607	DET	CLE	7.885863	9.185863
20130607	BOS	LAA	9.787777	11.08778
20130607	TB	BAL	6.0785	7.3785
20130607	NYM	MIA	6.019413	7.319413
20130607	CIN	STL	7.977821	9.277821
20130607	KC	HOU	6.797712	8.097712
20130607	CWS	OAK	7.919839	9.219839
20130607	MIL	PHI	7.0795	8.3795
20130607	COL	SD	10.26722	11.56722
20130607	ARI	SF	7.995217	9.295217
20130607	LAD	ATL	6.00375	7.30375
20130607	SEA	NYY	5.358701	6.658701

[matthew919]

You can verify by the timestamp the date and time of posts, as well as the lines which were offered at that time. So there's no issue of manipulation, or mistakes. I don't even understand what you're saying there. It's just digging you in deeper at this point.

Also, methodology isn't even relevant when performing tests for significance- results are either significant or not. If poor methodology is used, the results of a statistical test should reflect that. That's what hypothesis testing is. Your last comment is especially funny, because I am in fact a scientist.

[hutennis]

Anything posted after the lines are open is dirty and not worth even looking at.

Please post whatever your model does for the June 8 games.

This way nobody needs time stamps, assumptions, tracking your posts and other nonsense like that.
It also takes care of methodology issues.

Nice, clean, crisp, controlled.

[matthew919]

Hu- do me a favor- check the opening lines for EVERY GAME I HIGHLIGHTED. All but one just moved in response to the CRIS opener, by a huge amount. These are the games my model output indicated were out of whack. Good day sir. Your intellectual burial is now 100% complete.

For good measure, I'll post 6/8 too, as you requested.

[matthew919]

Here, I just did all of the work for you:

OVERS:
LAA/BOS - was o9.5 +100, now o10.5 -110
STL/CIN - was o7.5 -105, now o8 -105
CWS/OAK - was o7 +105, now o7.5 -110

UNDERS:
TEX/TOR- was u10 -115, now u9.5 -115
BAL/TB - was u8.5 -115, now u8.5 +100
NYY/SEA - was u8 -115, now u7.5 -115

[hutennis]

Hu- do me a favor- check the opening lines for EVERY GAME I HIGHLIGHTED. All but one just moved in response to the CRIS opener, by a huge amount. These are the games my model output indicated were out of whack. Good day sir. Your intellectual burial is now 100% complete.

How do I know you highlighted it before line moved hugely and not after that and, thus simply stealing the credit for that move?
Am I supposed to take your word for it, or spend time checking time stamps?Why do I need to do it.
That's your responsibility to make sure that question like that don't even arise.
That what CONTROLLED means. Clean of doubts and ambiguities

For good measure, I'll post 6/8 too, as you requested.

Thats more like it!

[matthew919]

You can tell because when someone edits a post, it is clearly labelled underneath the post.

See? I just edited this post five minutes after posting it, and you can tell below. What's it like living on a flat earth? Is the weather nice there?