[chunk]

"Public" loses not b/c of relative sharpness of open vs close.
It loses b/c of juice.

Every uneducated bet made at any time is subject to either overcharging or underpaying.
And since farther line movements are random, -EV bet is doomed to lose regardless.

For once, I can totally agree with you.

[mcuni]

Oh, I see. But what if someone will bring 20k (or 30K) on a opposite side 4 min before line is closed. Line will move up and you will be losing to closer having much worse odds?

Ok, but as i told that was an extreme example. Reduce 5 minutes to 1. Increase 10k to whatever amount is possible. Someone definitely will come to exploit ineffectiveness i will create 50% of times (as i randomly choose side to bet on). But due to lack of time it won't gonna happen too often. Well, I will beat the closer 90% of time, not 100%. Still good, heh?

Actually I wasn't all that serious about this example. The main point was to argue on closer vs. opener sharpness and (therefore) perspectives to win if you don't beat closer. I told that I thought that beating closing price was pure randomness. Though example of Matthew919 who demonstrated ability to beat closers consistantly disproves this (well, it was only 100 games distance plus backtesting but still).

"Public" loses not b/c of relative sharpness of open vs close. It loses b/c of juice.
Every uneducated bet made at any time is subject to either overcharging or underpaying.
And since farther line movements are random, -EV bet is doomed to lose regardless.

Ok, public loses because of both. As if it was a random number generator on the bookies' side inventing all these lines every day, then we both would have beaten MLB line relatively easy. Everybody would.

But I hope I understood your point. Line is sharp (it's not random, accounts for lot of factors) -> you can be just as sharp but that is hard, and rare player can consistently win vs. pure line -> it's much more probable that you can win vs. public, not vs. line -> you should estimate chances better than average to have better price.

Generally when you beat the closing line you beat average player who drives the line, not bookmaker. Actually given you are as sharp as bookmaker, your win is made up of an average edge over the closing line you manage to get less margin of bookmaker.

My point was based on sureness that you can (and have to) be sharper than bookies' line. That's what matters. I just don't understand how you can influence getting better or worse price even if you always choose right side. You can't control those who will bet after you!

How the process of desicion making looks like? Let's say, you know that fair probability of home team to win as 50%. You know it from your model which you believe replects reality best (otherwise why to use it?). Now there are three possible variants.
1) Line says the same. Here you just keep tracking the line and if it drops below 50% minus juice you have to pay then you bet the home team. If it goes higher than 50% plus juice then you go with away team.
2) Line underestimates the home team and implies, say, 40% probability of its win. You just bet on home.
3) Line overestimates the home team (supposing 60% probability of win, for instance). You bet on away.
Notice that with all these variants you don't even know where the line will go. You can't predict it, but you know that you have an edge at present moment. What secures your win is the fact you know which side to choose and what price to pay.

Ok let's agree that market is efficient therefore price 60-70% of time will go correct way (though I can't see it from statistics). Then all you need to do is to detect profitable bets earlier than others do. But again, BTCL is a consequence rather than a cause of you winning. An indicator which you can't even control. If market efficiency changes (there are more stupid money on market) then your BTCL drops and vice-versa.

[hutennis]

Look. You are saying a lot.
Without getting in to every sentence, I'll summarize your position as follows:

Public is dumb.
Books are puppets at the mercy of dumb public.
Juice is a secondary issue.
You are smart. At least as smart as odds makers.
You know where probability of win/lose is.

None of it makes any sense to me, but I could be wrong.

So for me to get a better grip on it, please, answer this question.

The line is -120/+110 (right at the open, right before close or anywhere in between)
IP of these odds is 53.39%/46.61% or -115/+115.
If implied = actual then market is efficient and you are doomed.
But it does not have to be efficient for you not to be able to win.
It can be inefficient and you will still lose.

In order for you to make money on your bet true probability expressed in odds should be either <+110 (+109, +108 etc.) or >-120(-121, -122 etc.)

How do you determine that that's the case?
What makes you think that line is off by that much?

[theclutch7]

How important is it for a model to beat the closing line in baseball? If I were to bet whoever I happened to like at any given time of day are my odds 50/50 to beat the closing line? And lets say I'm winning enough to be ahead, is it good enough if I'm only beating the closer half the time? Or since I'm not making moves ahead of the market half the time will it eventually catch up to me? What is a good percentage to beat closing lines at if it is important? Again, only referring to baseball money lines specifically, I read that its really important in football, but not sure about importance in baseball.

it depends on the weather and where they are play.

[matthew919]

But again, BTCL is a consequence rather than a cause of you winning. An indicator which you can't even control. If market efficiency changes (there are more stupid money on market) then your BTCL drops and vice-versa.

1. BTCL absolutely is a consequence of winning, not a cause. Telling someone to BTCL on its own is no more meaningful that telling someone to "pick winners." Where BTCL becomes a useful and incredibly powerful tool is when benchmarking your own results, and determining whether you will continue to be successful against the market. Saying you can't "control" it is meaningless. Can you control whether you identify lines that are incorrect? If so, then you can control your BTCL rate.

2. Market efficiency will NOT change for the worse. If anything it will become better with time, given how much data is freely available to the marketplace. The only exception I could see is is online gambling were legalized in the states.

3. 99.9% of the time, it's sharp money that shapes lines, not public money. Books like Pinny are the biggest gamblers on the planet in that regard, as they essentially subcontract out their analysis to sharp bettors, and leverage that information against the public. Exceptions are few and far between. Again, if gambling were legal in the states, this might change overnight. I dream of that day.

[matthew919]

That being said, sharps are getting absolutely crushed this year in MLB, from what I can tell (myself included). If you have been beating up the closing line, then chances are the closing line has been beating up on your bankroll.

But personally, I'd rather be on the right side and losing than be on the wrong side and winning.

[MasterP10]

That being said, sharps are getting absolutely crushed this year in MLB, from what I can tell (myself included). If you have been beating up the closing line, then chances are the closing line has been beating up on your bankroll.

But personally, I'd rather be on the right side and losing than be on the wrong side and winning.

Couldn't agree more! My bankroll has taken a nice beating the last 6 weeks and I have been beating the closing line (pinnacle) by 20,30 even 40 cents for a lot of the games and all I have done is lose, lose and lose. According to my spreadsheet I have had 3 winning days out of the last 24 lol

But like Matt said, I would rather be beating the number and losing then winning -EV bets.

[mcuni]

Look. You are saying a lot.

Sorry for many-words. This derives from willingness to be better understood. Apparently to no avail, as I never stated being smart. Nowhere told I that I did not belong to dumb public or was an above average player.

In short words my position was:
- average player loses (cause lines are sharp and because of juice which is kind of a tax)
- nobody can control bets other players do after his bet
- beating the closing line is a consequence rather than a cause of winning

If someone wins he will most probably beat the closing lines automatically, but he doesn't win because of this. He wins because he estimates probabilities better than bookmakers.

The line is -120/+110 (right at the open, right before close or anywhere in between)
IP of these odds is 53.39%/46.61% or -115/+115.
If implied = actual then market is efficient and you are doomed.
But it does not have to be efficient for you not to be able to win.
It can be inefficient and you will still lose.

In order for you to make money on your bet true probability expressed in odds should be either <+110 (+109, +108 etc.) or >-120(-121, -122 etc.)

How do you determine that that's the case?
What makes you think that line is off by that much?

For me the process looks like this. There is a lot of statistics. And there are lots of variables describing the state of things. These are performance indicators like pitching strength estimators etc., series data, weather, park... It's not easy to determine what's more important, what is less and what is not important at all. So by trial and error you form some kind of a model, backtest it, validate it. Inputs are important data, output - probabilities of outcomes. Sorry for saying obvious things in many words again, but the question was just about this. So how do I know (*think, I know*) that line is off? Just comparing model outputs with line. Then I have some confidence window, say 10%. If model output differs from line by 10% or more, I can be rather confident in it and make a bet. Otherwise I wait for line movements and if line goes off that window, I will bet on it.
I thought that was how anyone who dealt with betting with models acted.

[mcuni]

2. Market efficiency will NOT change for the worse. If anything it will become better with time, given how much data is freely available to the marketplace. The only exception I could see is is online gambling were legalized in the states.

Fare.

3. 99.9% of the time, it's sharp money that shapes lines, not public money. Books like Pinny are the biggest gamblers on the planet in that regard, as they essentially subcontract out their analysis to sharp bettors, and leverage that information against the public. Exceptions are few and far between. Again, if gambling were legal in the states, this might change overnight. I dream of that day.

Yes, I believe that is why Pinnacle says it welcomes winning players.
Thanks for clarifying your position about BTCL. Using it as a model benchmark - interesting idea (which was probably obvious for everybody here but me).
I just found the source of data about opens and closes which I used to test line sharpness. All of a sudden it turned to be this site. Did you also use sbrodds for your backtesting? If so, I need to check all again.

[matthew919]

Yes, I use the Pinny open and closers from this site, and filter out the lines that were clearly rogue at open (there aren't too many).

Try using a simple RMSE measure of openers and closers. First, adjust the lines by how they are shaded by the juice- I do this very roughly, by adding 0.25 runs for every 10 cents on the no vig line (so a line of o7 -115/ u7 +105 would become an adjusted total of 7.25). You can do an even better job by considering the different cent value of even/odd integers. Then compute:

RMSE_open = sqrt( mean((Adjusted opening total - Actual Total)^2) )
RMSE_close = sqrt( mean((Adjusted closing total - Actual Total)^2) )

If your odds data is clean, you will see a lower RMSE for the closing lines.

[theclutch7]

just play who is home and the picture, of course the weather also!

[mcuni]

Yes, I use the Pinny open and closers from this site, and filter out the lines that were clearly rogue at open (there aren't too many).

Try using a simple RMSE measure of openers and closers. First, adjust the lines by how they are shaded by the juice- I do this very roughly, by adding 0.25 runs for every 10 cents on the no vig line (so a line of o7 -115/ u7 +105 would become an adjusted total of 7.25). You can do an even better job by considering the different cent value of even/odd integers. Then compute:

RMSE_open = sqrt( mean((Adjusted opening total - Actual Total)^2) )
RMSE_close = sqrt( mean((Adjusted closing total - Actual Total)^2) )

If your odds data is clean, you will see a lower RMSE for the closing lines.

I see. I was talking about run lines, it's interesting to discuss them as well.

For totals I did the following:
1) Took the data for totals from this site: regular seasons 2010-2012. Notice that opens total line contains odds for Under only; closes total - for both Over and Under. So for opens I calculated Over odds using the same juice as for closes. Juice is generally 20 cents (which is u7 being -105 for o7 -115). The worst odds were -125 and the best 105, otherwise bookies change the total.

2) I filtered out the games with less than 350 batters faced for any of starting pitchers. I just don't bet on these as I need reliable stats for starters and predictions work better in this case. So I got a subset of 2133 games for 3 seasons.

3) I tuned totals for opens and for closes just like you described. For opens I have 8.08 average, min 5.375, max 12.5. For closes - 8.01 average, min 5.25, max 12.25.

RMSE for opens: 4.220066.
RMSE for closes: 4.214342.

It's almost no difference! Actually I prefer to check correlations, so correlation coefficient opens vs. actuals is 24.47% and closes vs. actuals 25.51%. This shows that closes are slightly more accurate but just slightly.
I believe the data I used is clean - I compared several random cases with data from site manually and it matched.

Did you get lower RMSE for closes?

[matthew919]

Good work. Similar to what I've seen. Don't be fooled- the numbers you report may be small in magnitude, but they are very meaningful.

Edit: These are my four year numbers, for a total of 9410 games:
RMSE_open = 4.312172
RMSE_close = 4.305899
PearsonCorr_open = 0.2319904
PearsonCorr_close = 0.2405462

It's not surprising that your RMSE numbers were slightly lower for both open and close, while the correlation coefficients were higher. I'm sure you can figure out why that is.

[matthew919]

Now do the next logical thing, and test the ROI for games if you had bet the "correct" side at open, versus a randomly selected side. You might be surprised by the results.

You can also play with this by binning games by their line movement, and computing ROI for each bin. You should see a striking trend of increasing ROI as line movement increases.

[mcuni]

It's not surprising that your RMSE numbers were slightly lower for both open and close, while the correlation coefficients were higher. I'm sure you can figure out why that is.

Sure

Now do the next logical thing, and test the ROI for games if you had bet the "correct" side at open, versus a randomly selected side. You might be surprised by the results.

Today I did almost the same thing with money line. I used the same subset as above (2010-2012 regular season lines for games with both starting pitchers having >=350 batters faced during the current season; it’s slightly more than 2100 games). Then I tried to answer the question: what profit one would have got if he bet on BTCL opening line. That is betting all teams whose opening odds are better than closing ones. Off 1874 games I got profit of 26.76, ROI = 1.43%, winning 52.45% of bets. Then I considered betting on teams whose opening odds were 5% or more better than closing odds. Now the profit was 27.11 but for 465 games, ROI = 5.83, winning 49.89% of bets. For 8% edge it’s profit 29.96 for 133 games, ROI = 22.53, winning 55.64% of games. For 10% edge it’s profit 17.13 for just 60 games, ROI = 28.55, winning 60.00% of games.

So if you are able to spot games with weird line which is subject to big moves then you can definitely make profit of this. These games are rare and line moves can be easily based on urgent information that changes chances. In some cases it can be weather, pitcher replacement, injury information, which was unknown at the moment opening line appeared. So part of this edge just cannot be utilized in real-life conditions. Look: 29.96 profit for 133 games compared to 26.76 for 1874 means that whole lot of 1741 games gave 3.20 units of profit (were almost useless). So what really contribute into high ROI and profit are those 133 games for 3 seasons, all with more than 8% line movement*

* I use European odds so 8% of line movement is a movement from 1.8 to 1.656 or from 2.2 to 2.376. So this is 100/0.8 = 125 to 100/0.656 = 152 or 120 to 138 => it’s about 20+ cents moves.

But finding those 45 games per season... To me it looks like mission impossible.
Will be interesting to check the totals.

You can also play with this by binning games by their line movement, and computing ROI for each bin. You should see a striking trend of increasing ROI as line movement increases.

That's a nice idea with binning! I'll try this tomorrow.

[matthew919]

Solid post my friend. You are correct about large line movements. Anything over 7% or so would almost certainly be due to a pitching scratch, or some bad data. I filter these cases out entirely when I am doing these kinds of analyses.

A lot of line movement is due to lineup announcements as well. But a sharp bettor can still use all the information at hand to identify a line that is incorrect at the time of the wager. If nothing else changes about the game between then and closing, chances are the line will move in your favor simply because you were on the correct side. If something does change (injury, etc) it will work in your favor roughly half of the time (increasing your edge even more), and against you half of the time (negating your edge).

You might think that this implies you should wait until you have all of the information before betting, but I claim you'd be wrong. There are relatively few games where lineup changes will have a large effect, and for a sharp, the advantage you get by betting early far outweighs the risk of being on the wrong side of one of these.

Also, keep in mind you're filtering down to less than 1/3 of games based on your criteria- there are plenty of opportunities to be found when starting pitchers have small sample sizes. I'll take another look at my 4 year data set and report the numbers I get for ML line movement and ROI, to see if they're any different.

Keep the analyses coming. It's good info for people who haven't attempted this exercise, or don't have the data to do so.

[mcuni]

Anything over 7% or so would almost certainly be due to a pitching scratch, or some bad data. I filter these cases out entirely when I am doing these kinds of analyses.

That's what I'm aware of. If those high-yield lines are defective then there's little profit left for the rest of bets. I would even say there is no (at least with my reduced data set).

You might think that this implies you should wait until you have all of the information before betting, but I claim you'd be wrong. There are relatively few games where lineup changes will have a large effect, and for a sharp, the advantage you get by betting early far outweighs the risk of being on the wrong side of one of these.

Agree, I read some article where author simulated different lineups for Texas and checked how it could influence average runs expected. The influence of lineup shuffling was rather small and resulted in +/- 0.1 run expected (if I remember correctly), which is 2-3 cents at most.

Also, keep in mind you're filtering down to less than 1/3 of games based on your criteria- there are plenty of opportunities to be found when starting pitchers have small sample sizes. I'll take another look at my 4 year data set and report the numbers I get for ML line movement and ROI, to see if they're any different.

Keep the analyses coming. It's good info for people who haven't attempted this exercise, or don't have the data to do so.

Thanks for attention to these. I really appreciate the possibility to discuss this with someone speaking at the same language as it can help to find some new ideas. Being sceptical of line movements and their utility may be derived from excessive expectations. Desired ROI for me is at least 8-10% (while filtering out up to 3 of 4 games). It doesn't seem that being drived by the line could provide anything close to this (given most of high-yield games spotted are defective). But may be I just want too much.

As for totals they turned to be really surprising just like you told. Not at first, as simply betting on the correct side gave just a slight plus. But then I tried random side choosing and losses were shocking. Though when you think a bit it's a no wonder: with random betting you will lose juice which is about 45 units for every 1000 bets, or 100 units for my data set. So BTCL definitely makes sense in this case.

Then I tried sequentially cutting off small line movements. For:
- at least 0.2 change of corrected total (8 cents) ROI was already 3% for 1388 games (65% of my subset);
- at least 0.4 change of corrected total (16 cents) ROI was 4.6% for 784 games (37% of my subset);
- at least 0.6 change of corrected total (24 cents) ROI was 8.1% for 535 games (25% of my subset);
- at least 0.7 change of corrected total (28 cents) ROI was 11.7% for 365 games (17% of my subset).

Even larger line moves result in still increasing ROI. I've checked this for those games I filtered out (<350 BF) as well. ROI is also growing, just a bit slower. It's 3.4% for 16 cents, 5.7% for 28 cents and 12.8% for 32 cents. ROI also falls down for 40 cents moves and bigger.

Shares in subset are not that small; 535 or even 365 games is not all that little for 3 seasons, and ROI is acceptable. But here is again the issue i'm aware of: one might be unable to utilize these because as I suspect most of these moves were due to weather changes. Totals are very sensitive to rains, winds and temperature. So unless one is brilliant weather forecaster he won't be able to benefit from this situation. It would have been interesting to analyze it if there was weather forecast information at the moment of opening lines.

[MDRTYson]

Worthwhile experiment for everyone: log the closers on all your bets (at pinnacle or another large book), and set up your spreadsheet to auto-calculate how much you beat that closing line by. Split those into bandings, say quintiles, and you'll notice how profitability correlates, although it does take a good sample size - as sample size increases you can use septiles, deciles, whatever the 20-quantile is called, etc.

There's no great magic here, nor anything terribly useful for the prediction process. It's simply that, _on average_, the pressure of information results in money moving the line towards a more accurate closer. Not always, but most of the time.

This is one of the reasons why books like pinnacle and SBO can exist. They use that weight of money to sharpen their lines, and thus can accept professional action.

This is an interesting test. I want to check it out. Do have the spreadsheet modeled out already? If so, I'd love to look at my historical data but have no idea how to create this.

[HeeeHAWWWW]

It would have been interesting to analyze it if there was weather forecast information at the moment of opening lines.

Shouldn't be that difficult. You have a limited number of venues, and weather data within the USA is excellent.

Unfortunately, it's also logistically plausible enough that I'd assume someone else has already done it. There won't be any massive edges to be found there, just a little refining.

[mcuni]

Shouldn't be that difficult. You have a limited number of venues, and weather data within the USA is excellent.

Maybe, someone can advice on the most accurate and informative site about weather in the US? It would also be interesting to have a site that provides access to data via xml, for example.

Actually I checked those 365 profitable games with 28+ cents line moves. I expected to see something special about them, like lower temperature than average or more rainy games. But nothing like that. Those were also equally distributed between AL and NL, percentage of night games was just as high as for total sample. And average temperature was just a little bit higher than total sample average, its distribution was also normal. Retrosheet data lacks a lot of information about precipitation (lots of 'Unknowns') but those games with drizzle or rain have similar shares in those 365 games and total sample. Same for Sunny/Cloudy games. So maybe it has nothing to do with weather.

[liuyelian]

I don't know, but you're making your life harder if there are better odds all day long and you take the worse odds later in the day because you were too lazy to run your analysis a few hours earlier.






________________________________________ _____________________________
? Born on 28.10.08 ?
avvie - sunakonakahara
siggie - artosan
Diablo 3 Gold
Buy Runescape Gold Buy RS Gold
How are you!I'm new!

[HeeeHAWWWW]

Actually I checked those 365 profitable games with 28+ cents line moves. I expected to see something special about them, like lower temperature than average or more rainy games. But nothing like that. Those were also equally distributed between AL and NL, percentage of night games was just as high as for total sample. And average temperature was just a little bit higher than total sample average, its distribution was also normal. Retrosheet data lacks a lot of information about precipitation (lots of 'Unknowns') but those games with drizzle or rain have similar shares in those 365 games and total sample. Same for Sunny/Cloudy games. So maybe it has nothing to do with weather.

I don't know enough about baseball to know whether it's relevant, but it may be worth checking correlations with weather in the days beforehand (eg rain-soaked, sun-baked etc etc). The speed of the ground will surely alter batter effectiveness - although as always with betting, it may be so marginal as to provide no useful edge :-)

[MDRTYson]

I've never paid much attention to this as it pertains to my win/loss. I always like to see a line moving in a way that leads me to believe I may have gotten some value but that's it. After reading this, decided to make my plays for today (Monday) last night (Sunday night) and pay attention to the moves. In all of the games I bet last night, there has been some movement in my favor (if we're calling it that).

Astros - Moved from +175 to +168 (+7)
Pirates - Moved from +130 to +110 (+20!)
Indians - Moved from +111 to +101 (+10)
Rockies - Moved from +138 to +123 (+15)

So in 4 games, in one night, total of +52. My book puts up the next days games pretty early. I realize this is a small sample size of one day and it's a weekend going into Monday but given this kind of movement, seems beneficial to get these in early so far.

[EXhoosier10]

I've never paid much attention to this as it pertains to my win/loss. I always like to see a line moving in a way that leads me to believe I may have gotten some value but that's it. After reading this, decided to make my plays for today (Monday) last night (Sunday night) and pay attention to the moves. In all of the games I bet last night, there has been some movement in my favor (if we're calling it that).

Astros - Moved from +175 to +168 (+7)
Pirates - Moved from +130 to +110 (+20!)
Indians - Moved from +111 to +101 (+10)
Rockies - Moved from +138 to +123 (+15)

So in 4 games, in one night, total of +52. My book puts up the next days games pretty early. I realize this is a small sample size of one day and it's a weekend going into Monday but given this kind of movement, seems beneficial to get these in early so far.

I'll try providing data to back my next statement up, but bets with those kinds of line movement have been getting killed recently.


eh... pulled the wrong data. will have confirmation/refutation later today or tomorrow

[EXhoosier10]

I'll try providing data to back my next statement up, but bets with those kinds of line movement have been getting killed recently.


eh... pulled the wrong data. will have confirmation/refutation later today or tomorrow

data is from 5/12/13 - 6/2/13

>= 20 cent movement @ 5d open to close and betting the opener / closer using $100 bets
Away teams: 7-10 -$22 (-.43% ROI) betting the opener, -$316 (-5.76% ROI) betting the closer
Home teams: 14-14 -$763 (-8.3% ROI) betting the opener, -$1,051 (-9.99% ROI) betting the closer

>= 10 cent movement @ 5d open to close and betting the opener / closer using $100 bets
Away teams: 24-32 -$368, -$883
Home teams: 49-43 -$649, -$1,281

[matthew919]

data is from 5/12/13 - 6/2/13

>= 20 cent movement @ 5d open to close and betting the opener / closer using $100 bets
Away teams: 7-10 -$22 (-.43% ROI) betting the opener, -$316 (-5.76% ROI) betting the closer
Home teams: 14-14 -$763 (-8.3% ROI) betting the opener, -$1,051 (-9.99% ROI) betting the closer

>= 10 cent movement @ 5d open to close and betting the opener / closer using $100 bets
Away teams: 24-32 -$368, -$883
Home teams: 49-43 -$649, -$1,281

Thanks for doing this Hoosier. Confirms what I and others have suspected. I'd be willing to bet that dipping back even farther into April would be just as ugly.

I dunno if you can do the same for totals- line movement is a trickier animal for those- but based on my own wager history, I think it would look very similar. Long season still to go though. Sharps will be fine.

[MDRTYson]

Are the stats above specific to any line movement at 10 and 20 or only when the number favors be player at opening? In other words, if I would have bet the pirates at +130 and it moved to +150 the following day, would that be captured in your data? If so, then I don't know how telling those figures really are.

I'm assuming your talking line moves that "appear" in the players favor early an the number gets worse approaching game time?

[EXhoosier10]

Are the stats above specific to any line movement at 10 and 20 or only when the number favors be player at opening? In other words, if I would have bet the pirates at +130 and it moved to +150 the following day, would that be captured in your data? If so, then I don't know how telling those figures really are.

I'm assuming your talking line moves that "appear" in the players favor early an the number gets worse approaching game time?

+130 to +150 would not count towards this as i'm looking at line movement in the opposite direction. The opposing team's -140 open to a -160 close in your hypothetical game would count towards these records, though.

[hutennis]

Are the stats above specific to any line movement at 10 and 20 or only when the number favors be player at opening? In other words, if I would have bet the pirates at +130 and it moved to +150 the following day, would that be captured in your data? If so, then I don't know how telling those figures really are.

I'm assuming your talking line moves that "appear" in the players favor early an the number gets worse approaching game time?

It does not seem to me that effort spent on looking for line movement patterns has any meaningful chance to be productive. Lines move on news and money flow. Both news and money flow come in a random fashion thus making line movements random.
If it was not random and there was any kind of bias one way or another it would have been recognized already, acted upon and negated.

[matthew919]

It does not seem to me that effort spent on looking for line movement patterns has any meaningful chance to be productive. Lines move on news and money flow. Both news and money flow come in a random fashion thus making line movements random.
If it was not random and there was any kind of bias one way or another it would have been recognized already, acted upon and negated.

You are missing the entire point of this conversation.

[matthew919]

Sure



Today I did almost the same thing with money line. I used the same subset as above (2010-2012 regular season lines for games with both starting pitchers having >=350 batters faced during the current season; it’s slightly more than 2100 games). Then I tried to answer the question: what profit one would have got if he bet on BTCL opening line. That is betting all teams whose opening odds are better than closing ones. Off 1874 games I got profit of 26.76, ROI = 1.43%, winning 52.45% of bets. Then I considered betting on teams whose opening odds were 5% or more better than closing odds. Now the profit was 27.11 but for 465 games, ROI = 5.83, winning 49.89% of bets. For 8% edge it’s profit 29.96 for 133 games, ROI = 22.53, winning 55.64% of games. For 10% edge it’s profit 17.13 for just 60 games, ROI = 28.55, winning 60.00% of games.

So if you are able to spot games with weird line which is subject to big moves then you can definitely make profit of this. These games are rare and line moves can be easily based on urgent information that changes chances. In some cases it can be weather, pitcher replacement, injury information, which was unknown at the moment opening line appeared. So part of this edge just cannot be utilized in real-life conditions. Look: 29.96 profit for 133 games compared to 26.76 for 1874 means that whole lot of 1741 games gave 3.20 units of profit (were almost useless). So what really contribute into high ROI and profit are those 133 games for 3 seasons, all with more than 8% line movement*

* I use European odds so 8% of line movement is a movement from 1.8 to 1.656 or from 2.2 to 2.376. So this is 100/0.8 = 125 to 100/0.656 = 152 or 120 to 138 => it’s about 20+ cents moves.

But finding those 45 games per season... To me it looks like mission impossible.
Will be interesting to check the totals.



That's a nice idea with binning! I'll try this tomorrow.

My results are a bit different than yours for moneylines, and show a greater correlation between line movement and ROI. Below are two bar plots which show the trends, for both totals (where the units on the x axis are in my own "adjusted runs" statistic), and moneylines, where the binning was done by the movement of the no-vig implied probability from pinny.

Totals_ROIByLineMovement.jpg
ML_ROIByLineMovement.jpg


Also, weather reports are very accurate 24 hours in advance. You need to account for weather to have any chance of getting a sharp line. But in general there's not much that will change by game time. Doesn't surprise me that nothing weather-related jumped out from your outlier set.

[EXhoosier10]

It does not seem to me that effort spent on looking for line movement patterns has any meaningful chance to be productive. Lines move on news and money flow. Both news and money flow come in a random fashion thus making line movements random.
If it was not random and there was any kind of bias one way or another it would have been recognized already, acted upon and negated.

Hu, do you take the position that sports betting is a perfectly efficient market?
If no, do you take the position that there is no information to be gleaned from line shading/movement/where it opens/etc?
What about ERA, W/L record, SIERA, wRC+, pythagenport or any other public stat? All of these have been around for multiple years and smart people have surely taken advantage of them. Does looking at any of this information help a bettor find an edge? If so, why can't line movement/placement be just another data point to look at?

[hutennis]

You are missing the entire point of this conversation.

You know what, very well could be. I'm also missing some of the threads I've contributed to.
It looks like as soon as I start making too much sense, the whole thread is being moved to PT where it quickly gets lost.

Anyway, so what is the point? Which one is more efficient - open or close? is that it?
Gotta be close. It incorporates more relevant information.

[matthew919]

You know what, very well could be. I'm also missing some of the threads I've contributed to.
It looks like as soon as I start making too much sense, the whole thread is being to PT where it quickly gets lost.

Anyway, so what is the point? Which one is more efficient - open or close? is that it?
Gotta be close. It incorporates more relevant information.

In a nutshell:

1. Beating closing lines in MLB is possible.
2. Doing so consistently identifies you as a sharp, who will almost certainly win long term.
3. Not doing so identifies you as a losing long term bettor.
4. 2013 has so far been very generous to squares.

[hutennis]

Hu, do you take the position that sports betting is a perfectly efficient market?
If no, do you take the position that there is no information to be gleaned from line shading/movement/where it opens/etc?
What about ERA, W/L record, SIERA, wRC+, pythagenport or any other public stat? All of these have been around for multiple years and smart people have surely taken advantage of them. Does looking at any of this information help a bettor find an edge? If so, why can't line movement/placement be just another data point to look at?

Good questions. Lets see.

Hu, do you take the position that sports betting is a perfectly efficient market?

No. I take a position that this market maybe not perfectly efficient, but is efficient enough.
-105/-105. Is 100 true odds? I don't know. But it does not have to be. What if 100 is not perfectly efficient?
What if -104 is? Does it help you to win? not at all.
Another words, possible (and most likely probable) inefficiency is protected by what I call "margin for error", aka vig.
Any one who claims that market has larger inefficiency margin has a burden of prove to demonstrate it.
As far as I know, not even 1% of participants can do it. That, imo, makes market efficient enough.

If no, do you take the position that there is no information to be gleaned from line shading/movement/where it opens/etc?

Most likely, in 1% of the cases you can glean something useful out of it and that will result in gaining the edge.
If you can do it consistently, on average, you will be life long winner and member of 1% club.

What about ERA, W/L record, SIERA, wRC+, pythagenport or any other public stat? All of these have been around for multiple years and smart people have surely taken advantage of them. Does looking at any of this information help a bettor find an edge?

If those who set the odds are not aware of these stats and dont include them in their calculations, then yes, those stats are gold mine providing that you use them correctly.
But if they know everything you do about ERA, W/L record, SIERA, wRC+, pythagenport etc., then, even if you use stats as well as world class
statisticians employed by those who set the odds, you gain nothing. Your efforts and knowledge and their efforts and knowledge simply get
washed and all is left is juice. Unless you line shop of course.