The misunderstanding of "beating the closing line."

[MonkeyF0cker]

There might be some small amount of value at close, but not nearly as much (from a percentage perspective) as hitting openers, even if you fail to BTCL 30% of the time.

Wait. What? Are you backtracking here? How can there be ANY value at the close? You were just saying how every market is strong form efficient at the close last night.

[Shonner]

I want to point out a lot of people are saying that you need a superior model to beat the closing line or you are basically SOL. I know everyone already understands this concept, but the easiest and best way to do it (with no risk really) is to just shop books and try to find weak lines near close time. Use the half point calculator to see if the odds @ your book have +EV when bumped up against the true value (Pinny/MB).

Another great way to beat the closing line is to do the same thing with money lines. You can take the ML at MB or Pinny and then figure out the "no vig" ML by using ganch's spreadsheet or doing simple math. Once you get the no vig ML, if your book beats it, take it, it's +EV. I have had a lot of luck at finding +EV in money lines, I can def. grab a handful every weekend.

Well I know there are some really bright math guys here (e.g., Justin, Data) that are successful in creating models and using complex math/stats, they are def. the minority, and I really think most of us (including myself) would benefit from taking these basic approaches as opposed to trying to create models and beating the closing line on our own. As Justin pointed out earlier, there are very few models that actually work and most bettors fail miserably at trying to beat the closing line with their model.

While this thread was all over the place for a while and there was a lot of bickering, it was all for the better because some really bright guys contributed. I understood most of it, but not all of it, and took a lot out of it.

Lastly, apologies to trixtrix for going after him, regardless of who started it (I don't even remember). As much as I may disagree it is never acceptable to put someone down, we are all here to help each other, learn, and make money and sometimes it is easy to forget that.

[Flight]

Shonner: Playing half points for free is akin to basic strategy in blackjack, where the edge is near 0 but still not good enough to make money. You need a new strategy to make money long term, like counting cards in black jack, or beating the closing line in sports.

But you are correct, players would do much better if they could just evaluate a line's position instead of picking random games on a gut feel. Just like players would do better if they played basic strategy at blackjack. I have observed public sports bettors and public black jack players, and more than 90% do not follow basic strategy at either game.

[Justin7]

Wait. What? Are you backtracking here? How can there be ANY value at the close? You were just saying how every market is strong form efficient at the close last night.

I don't recall saying every market is strong form efficient. It is very efficient, however.

If you look at the last 2 minutes before post, you might see a game move 30 cents. One of those numbers -- before or after the move obviously have value. There could be late breaking news -- i.e. who won the coin toss, that changes the "best information" price. There might even be a tiny bit of value due to market inefficiency... although this has gone down a lot due to Matchbook.

[MonkeyF0cker]

I don't recall saying every market is strong form efficient. It is very efficient, however.

If you look at the last 2 minutes before post, you might see a game move 30 cents. One of those numbers -- before or after the move obviously have value. There could be late breaking news -- i.e. who won the coin toss, that changes the "best information" price. There might even be a tiny bit of value due to market inefficiency... although this has gone down a lot due to Matchbook.

Are you saying that every line move is the direct result of new game information? How is that not strong form efficiency exactly? If you were to randomly select 10 economists, they wouldn't agree that the financial markets, where trillions of dollars are traded every day, are nearly as efficient as you make out a simple sports market to be.

[Justin7]

Are you saying that every line move is the direct result of new game information? How is that not strong form efficiency exactly? If you were to randomly select 10 economists, they wouldn't agree that the financial markets where trillions of dollars are traded every day are nearly as efficient as you make out a simple sports market to be.

In some ways, sports markets are more efficient than financial markets. There is much greater transparency of information. There is the magnified yield due to the short turn-around of investments in sports markets (which tends to increase efficiency). The down side is that there are higher transaction costs. Even if you are getting matched at 0.4%, that's $400 on a 100k order, whereas financial markets might charge $8 (or less) on a 100k order.

I don't think all markets are ultra-efficient, but NFL, MLB and NBA are IMO.

[Data]

Since JustinBieber is unqualified to be the score-keeper I have to continue monitoring and providing cliff notes on what's going on in this epic thread. Well, apparently we have a Round 3 going on. In the 1st Round we had a discussion on what to call BTCL, in round 2 Justin7 took a jab at hedging but simply left the ring having met some strong opposition. Now we are discussing how efficient the betting markets are. I find the subject of this round akin discussing personal tastes. Nonetheless, or may be due to the above, this is a fine opportunity to have some fun.

Justin7, I was wondering, how do you fit the right-before-the-close Asian money in your paradigm?

[Justin7]

Since JustinBieber is unqualified to be the score-keeper I have to continue to monitor and provide cliff notes on what's going on in this epic thread. Well, apparently we have a Round 3 going on. In the 1st Round we had a discussion on what to call BTCL, in round 2 Justin7 took a jab at hedging but simply left the ring having met some strong opposition. Now we are discussing how efficient the betting markets are. I find the subject of this round akin discussing personal tastes. Nonetheless, or may be due to the above, this is a fine opportunity to have some fun.

Justin7, I was wondering, how do you fit the right-before-the-close Asian money in your paradigm?

It's either sharp, or inefficient. I'm not sure which.

[Data]

It's either sharp, or inefficient. I'm not sure which.

Of course you are unsure. It is an artifact that you just cannot fit in.

[Justin7]

Of course you are unsure. It is an artifact that you just cannot fit in.

And how do you handle this artifact?

[MonkeyF0cker]

In some ways, sports markets are more efficient than financial markets. There is much greater transparency of information. There is the magnified yield due to the short turn-around of investments in sports markets (which tends to increase efficiency). The down side is that there are higher transaction costs. Even if you are getting matched at 0.4%, that's $400 on a 100k order, whereas financial markets might charge $8 (or less) on a 100k order.

I don't think all markets are ultra-efficient, but NFL, MLB and NBA are IMO.

Why do you think the transaction costs are higher, Justin? Could it possibly be because of an incredibly smaller amount of volume? What did your Finance 101 book tell you about the relationship between volume and efficiency? Anything?

[MonkeyF0cker]

It's either sharp, or inefficient. I'm not sure which.

How could you possibly consider it inefficient? That is entirely incongruent with your whole market premise.

[Flying Dutchman]

Since 2000, the NFL market has gotten LESS accurate in predicting the eventual winner judging from RMS of line to score.

Efficient Market it is not.

[Data]

And how do you handle this artifact?

I do not subscribe to EMH, so it is not an artifact for me.

[donjuan]

Since 2000, the NFL market has gotten LESS accurate in predicting the eventual winner judging from RMS of line to score.

Efficient Market it is not.

Are the totals for the 2000s higher or lower than previous years?

[RockyV]

The earlier you bet, the greater the chance that new information will come out. Ergo, it is almost impossible to BTCL 100% of the time if you are betting a week in advance.

I don't get this. Doesn't this suggest BTCL is not necessary? Like, the new information that comes out (assuming I've modeled things properly) should on average have zero impact on the value of my plays. So if the expected value of my plays doesn't change, then of what importance is BTCL?

[JustinBieber]

Since 2000, the NFL market has gotten LESS accurate in predicting the eventual winner judging from RMS of line to score.

Efficient Market it is not.

lol

[Justin7]

Why do you think the transaction costs are higher, Justin? Could it possibly be because of an incredibly smaller amount of volume? What did your Finance 101 book tell you about the relationship between volume and efficiency? Anything?

I never took a class on finance, so I can't tell you that.

But it's obvious that generally, the higher the volume, the less transaction costs are as a % of the deal. But that's not the end-all for efficiency.

Transaction costs. Let's say this distorts the true price by 1%. If I could lay -104/-104 by shopping, and I thought a team would win 51% of the time, I would not play. At 52%, I would. Due to transparency (at least in major sports), you can in theory get a very accurate price on matchups. The information is all there and mostly honest.

Stock betting. The transaction cost is less than 1%. Let's call it 0%. But how accurate is your information? Both markets are pretty efficient, but each has its own barriers. I think that for non-investment brokers (like me, that can rely only rely on published reports), the stock market is no where near as efficient as sports markets. The information just isn't there for us, and it is often wrong or manipulated.

[djiddish98]

Are the totals for the 2000s higher or lower than previous years?

Based on dubious sunshine forecast data - average total by year.

1979 38.52232143
1980 40.87946429
1981 41.41964286
1982 39.08928571
1983 42.30882353
1984 42.5
1985 42.36160714
1986 41.06919643
1987 42.99107143
1988 41.13392857
1989 41.234375
1990 40.18303571
1991 37.90625
1992 37.63616071
1993 37.01990521
1994 39.1814346
1995 40.62083333
1996 40.69791667
1997 40.88958333
1998 40.73541667
1999 40.11373391
2000 41.40874525
2001 39.82403433
2002 41.36162362
2003 41.01367188
2004 41.90208333
2005 41.18410853
2006 40.47592593
2007 41.60546875
2008 42.46289063
2009 42.86875

[bztips]

Transaction costs. Let's say this distorts the true price by 1%. If I could lay -104/-104 by shopping, and I thought a team would win 51% of the time, I would not play. At 52%, I would. Due to transparency (at least in major sports), you can in theory get a very accurate price on matchups. The information is all there and mostly honest.

Stock betting. The transaction cost is less than 1%. Let's call it 0%. But how accurate is your information? Both markets are pretty efficient, but each has its own barriers. I think that for non-investment brokers (like me, that can rely only rely on published reports), the stock market is no where near as efficient as sports markets. The information just isn't there for us, and it is often wrong or manipulated.

Well this just about explains in its entirety why people like Justin and me will never agree.

There is no way I would ever be confident enough in either the market or my models to be able to accurately distinguish between 51% and 52% probability. EVER.

I guess if others have models that have even 80% prediction confidence intervals of less than 1 percentage point, then you're a better model builder than me. But frankly, I don't believe such models exist.

[Thremp]

How do you know you have an edge if you can't build an accurate model?

[bztips]

How do you know you have an edge if you can't build an accurate model?

It's all a matter of degree, but your point is spot on. Edge is uncertain with any model prediction; that's why I would only bet when my model shows a "significant" difference from the line. Example: The line says 52%, my model says 53%, that's not enough difference to bet because a reasonably sized confidence interval around my 53% prediction encompasses the line. But if my model says 57% with a confidence interval of +/- 3%, then I'll bet.

As an aside, this is also why I don't care nearly as much about line movement (and BTCL) as others apparently do. In this example, the line would have to move 25 cents to completely wipe out my EV.

[Peregrine Stoop]

If the cl outperforms the opener 55% of the time that still leaves 45% where the opener ended up closer to reality than the cl did. Tough to chalk 100% of that up to noise... sometimes the market simply moves the line in the wrong direction. Is it more important then to be able to forecast which direction the market will move the line or whether or not the line is accurate? Doing both together is ideal so you can always get the best line on the correct side whether that be an opener or a closer. Most important is to be on the correct side though.

yes, if the possible outcomes fall across some sort of distribution like a bell curve and the closing line beats the opening line 55% of the time, it is very likely just noise and the closing line (or closing line t-1) is the true estimate and the opening line meant diddly poo other than an anchoring spot for the line to adjust from.

[MonkeyF0cker]

I never took a class on finance, so I can't tell you that.

But it's obvious that generally, the higher the volume, the less transaction costs are as a % of the deal. But that's not the end-all for efficiency.

Transaction costs. Let's say this distorts the true price by 1%. If I could lay -104/-104 by shopping, and I thought a team would win 51% of the time, I would not play. At 52%, I would. Due to transparency (at least in major sports), you can in theory get a very accurate price on matchups. The information is all there and mostly honest.

Stock betting. The transaction cost is less than 1%. Let's call it 0%. But how accurate is your information? Both markets are pretty efficient, but each has its own barriers. I think that for non-investment brokers (like me, that can rely only rely on published reports), the stock market is no where near as efficient as sports markets. The information just isn't there for us, and it is often wrong or manipulated.

I'm curious, Justin. How do you justify the very last line move of a half point before the close? Let's say the total for a Packers/Bears game moves from 40.5 -105/-105 to 41 -105/-105 an hour before the close at a book like Pinny and settles there. This has to be sharp action by definition, right? But it has no value against the closer. What does this say? Shouldn't people be pounding any 40.5 overs available? Obviously, some sharp(s) saw value in it at 40.5. How couldn't there be value in it according to your market premise?

The problem is that you contradict yourself in this situation. The last half point move can't have value with how you calculate it. So how could it be sharp movement?

[Shonner]

I
The markets are not completely efficient. There is juice. This effect means a 5% delta between the opening value and the theoretical value might be reflected as only a 3% line move. BTCL will over-perform your expectation, again and again.

Justin - I think I get your point but I am having trouble understanding what you mean when you are using the 3% vs. 5% example because of the Vig. I guess I am having trouble understanding what you are referring to with regard to the %'s (e.g., EV, win %).

Would you mind giving an example of this using showing in numbers how the 5% turns into 3% because of Vig?

It would be very helpful - thanks.

Thanks.

[Justin7]

Justin - I think I get your point but I am having trouble understanding what you mean when you are using the 3% vs. 5% example because of the Vig. I guess I am having trouble understanding what you are referring to with regard to the %'s (e.g., EV, win %).

Would you mind giving an example of this using showing in numbers how the 5% turns into 3% because of Vig?

It would be very helpful - thanks.

Thanks.

Assume you have a perfect model with a huge bankroll (much larger than the sports market you are attacking anyway). You will make any bet where you have any edge.

The if the market price is Pick -110 both ways, the market suggests it is a 50/50 chance of either team winning.

If your model "knew" there was a 52% chance of a team winning, you still wouldn't bet. The vigorish creates an inefficiency in the market. With perfect models and -110 pricing, the market will often close up to 2.4% off the "perfect theoretical price".

[Justin7]

I'm curious, Justin. How do you justify the very last line move of a half point before the close? Let's say the total for a Packers/Bears game moves from 40.5 -105/-105 to 41 -105/-105 an hour before the close at a book like Pinny and settles there. This has to be sharp action by definition, right? But it has no value against the closer. What does this say? Shouldn't people be pounding any 40.5 overs available? Obviously, some sharp(s) saw value in it at 40.5. How couldn't there be value in it according to your market premise?

The problem is that you contradict yourself in this situation. The last half point move can't have value with how you calculate it. So how could it be sharp movement?

It could be sharp; it probably is. In rare circumstances, it might be public (or big square) money. The sports market is not perfectly efficient, only very.

If the over 40.5 -105 were a sharp move, then obviously chasing slow-moving over 40.5 -105s would be a good play.

[Shonner]

Assume you have a perfect model with a huge bankroll (much larger than the sports market you are attacking anyway). You will make any bet where you have any edge.

The if the market price is Pick -110 both ways, the market suggests it is a 50/50 chance of either team winning.

If your model "knew" there was a 52% chance of a team winning, you still wouldn't bet. The vigorish creates an inefficiency in the market. With perfect models and -110 pricing, the market will often close up to 2.4% off the "perfect theoretical price".

Thank you. Basically, in a market with perfect models and -110 pricing, IN THEORY - assuming an efficient market, bets won't close at 3%+ or greater off the "perfect theoretical price" because that would indicate +EV and in a perfectly efficient market with vigorish, there's not +EV.

[Justin7]

Shonner,

Yes. In an efficient market, big players won't leave +EV bets on the table.

[Flying Dutchman]

Are the totals for the 2000s higher or lower than previous years?

Totals market has gotten better since 2000. The worst year overall was 1988 (totals started being WA in 1986), best year was 1992. Since 2000: worst year was 2002, best was 2008.

I'd post a few graphs but not sure how to do it in this interface.

[uva3021]

I wrote a dirge about my beliefs here, but I'll post the graphs below just as a graphical supposition (some descriptive measures between closing spread from Vegas and final score)







I figure most here would understand what I'm getting at, and what the graphs portray, if not I'll be happy to share my thoughts on the matter, though they are pretty counter productive and overly cynical

[Flight]

At first glance I thought you were trying to imply there is a trend, perhaps a trend of decreasing market efficiency.

I was going to post that you are wrong and instead you have observed randomness in it's raw form.

Then I read your blog and see that we in fact concur that we see randomness. The only quantity we see fluctuating is variance. Good luck with your MS.

[Shonner]

I wrote a dirge about my beliefs here, but I'll post the graphs below just as a graphical supposition (some descriptive measures between closing spread from Vegas and final score)

I figure most here would understand what I'm getting at, and what the graphs portray, if not I'll be happy to share my thoughts on the matter, though they are pretty counter productive and overly cynical

Please share thoughts - because you feel the closing line is just a "variation of stoppage", at the end of the day you believe BTCL doesn't have much value?

[Flight]

1) Read his blog, it explains why he created the charts and the misleading power of data mining.

2) Don't quote the entire post, we already saw his charts once.

[Shonner]

1) Read his blog, it explains why he created the charts and the misleading power of data mining.

2) Don't quote the entire post, we already saw his charts once.

1 - i did read his blog. i was trying to tie it into the topics in this thread and responded to his open offer to request more info. not all of us are as good with stats and math as others.
2 - i edited my post to delete the charts but you added nearly as much empty space by creating that post