[Bsims]

I have been working on a system designed to leverage the sports books lines against them. It is based on the following concepts.

For any given game, there are many previous games very similar to today’s game. If you look at the results of these “like” games, you will get distribution of final scores. These can be used with today’s odds to identify potential wagers.

The sports book odds on any given game can be compared to the odds of previous games to identify many “like” games. The distribution of the results of these can be used to identify potential wagers for today.

Any thoughts on this approach?

[PAULYPOKER]

This might help?

Click on free match=up under each MLB game

Shit loads of data...........

http://www.scoresandodds.com/index.html

[HUY]

Bsims, I like this approach. You describe a system that learns from the odds. That's a much better system than the usual ones that only learn from results. Results have very large variance so trying to learn from them is pointless.

[chilidog]

I have been working on a system designed to leverage the sports books lines against them. It is based on the following concepts.
For any given game, there are many previous games very similar to today’s game. If you look at the results of these “like” games, you will get distribution of final scores. These can be used with today’s odds to identify potential wagers.
The sports book odds on any given game can be compared to the odds of previous games to identify many “like” games. The distribution of the results of these can be used to identify potential wagers for today.
Any thoughts on this approach?

This was the basis of a very successful betting system that worked quite well for many years. It's no longer profitable, though (which is why the system seller quit selling it, and has since moved on to other ventures).

[Waterstpub87]

I have been working on a system designed to leverage the sports books lines against them. It is based on the following concepts.

For any given game, there are many previous games very similar to today’s game. If you look at the results of these “like” games, you will get distribution of final scores. These can be used with today’s odds to identify potential wagers.

The sports book odds on any given game can be compared to the odds of previous games to identify many “like” games. The distribution of the results of these can be used to identify potential wagers for today.

Any thoughts on this approach?

I disagree on this. I don't think much can be gleened from past odds.

1. You are not counting injuries
2. Past results did not incorporate new information from the games that already happened, IE you didnt know how well a team with mesh together at a season, or how players would be faring in the new.
3. Your definition of like games is not a clear one, do we mean same teams, or a -3 favorite playing on a road with a 48 total in NFL.

This seems like one of these gimmick things that people find through backtesting and then get killed when they put into practice.

[Bsims]

3. Your definition of like games is not a clear one, do we mean same teams, or a -3 favorite playing on a road with a 48 total in NFL.

For sports where the lines are primarily spreads (football and basketball), I use the spread and total. These give you an expected score for each team. In your example, the home team is +3 and total is 48. The expected score for the visitors is 25.5 and for the home team 22.5. I then search for previous NFL games with expected scores equal to this or near these. Ideally, you want a few hundred. This gives you the distribution of potential scores to use to evaluate today's odds.

[Bsims]

This was the basis of a very successful betting system that worked quite well for many years. It's no longer profitable, though (which is why the system seller quit selling it, and has since moved on to other ventures).

I'd be interested in any information about this previous approach and why it failed.

[Optional]

Books adjusting lines over time, especially to account for rule changes, might be a potential issue to look at.

[ChuckyTheGoat]

Bump. Bsims is a great poster. Guy is always looking for an angle.

[ferndog]

I have been working on a system designed to leverage the sports books lines against them. It is based on the following concepts.

For any given game, there are many previous games very similar to today’s game. If you look at the results of these “like” games, you will get distribution of final scores. These can be used with today’s odds to identify potential wagers.

The sports book odds on any given game can be compared to the odds of previous games to identify many “like” games. The distribution of the results of these can be used to identify potential wagers for today.

Any thoughts on this approach?

So i looked at favs in MLB 2018 between -130 & -140 with a total of at least 9 or higher & they are

SU:	46-28 ( 62.2%)

for a return of +$810 ROI 8.1% and

OU:	40-29-5 (58.0%)

for a return of +$831 ROI 10.1%
If you just look at these moneylines without the totals the record is

SU:	122-81 (60.1%)

+$1,153

ROI: +4.2%

I got these from killersports. Hope this helps.

[Rich Boy]

Bsims I applaud you sir for another insightful and thought provoking thread

[Rich Boy]

My 2 cents on the issue:

1. Sample size with a model like this will be an issue. "a few hundred" data points isnt enough to really be considered significant, it can provide guidance for sure but not enough confidence to risk money on.

2. Leagues are constantly changing rules and styles of game play. Totals are much higher in NHL, NBA and NFL games than in years past. The number of rule changes is increasing year after year as leagues try to tweak their product to maximize scoring and entertainment value. That alone has completely ruined my NHL UNDER system I ran profitably for years before getting wrecked this year. Using historical data is worthless now and new data sets need to be tracked, analyzed and backtested.

I like the path you are on Bsims and I think you will find something to use against the books, just factor those things in if you havent already sir.

[Bsims]

A couple of posters have mentioned the issue of rule changes. These certainly would impact any system that uses past result data as input to a model. The model would have to somehow incorporate the changes. Potentially a difficult thing to do (depends on the construct of the model).

However the lines would rapidly adjust for these changes. This would imply that the final scores projected by the lines would remain accurate irrespective of rule changes.

[Bsims]

2. Leagues are constantly changing rules and styles of game play. Totals are much higher in NHL, NBA and NFL games than in years past. The number of rule changes is increasing year after year as leagues try to tweak their product to maximize scoring and entertainment value. That alone has completely ruined my NHL UNDER system I ran profitably for years before getting wrecked this year. Using historical data is worthless now and new data sets need to be tracked, analyzed and backtested.

This raises an important point. The system based on lines alone has a major drawback. It doesn't produce many wager opportunities on the major bet types. For instance, in a spread environment it will indicate that the probability of the dog or favorite covering is about 50-50, hence no bet. Same for overs and unders. However, there are two opportunities where the distribution of the previous scores could be useful.

1. The off primary markets. In the spread environment above it could suggest good wagers on the money lines, parlays, buying points , etc.

2. In support of other models. In Rich Boy's case, applying the distribution of the past scores would probably not have supported his NHL under model. That would suggest he take another look at his model.

[ChuckyTheGoat]

Theoretical Question for u, Bsims:

* I'm assuming that you'd have some ALLOWANCE for side/total before grouping similar games. IE, -3/48 w/ 1pt allowance of side/total would group -2 to -4 w/ 47 to 49 as "similar games."

Let's say we do that. How do we proceed w/ the information? Most leagues aren't going to have systematic error. So, average line of 3/48 is going to produce output CLOSE to those numbers.

Are u saying that one angle might be parlays? IE, side/total in that ballpark might play out as close/tight lending to dog/under? Call me pessimistic...but i HOPE that u prove me wrong.

[Bsims]

Theoretical Question for u, Bsims:

* I'm assuming that you'd have some ALLOWANCE for side/total before grouping similar games. IE, -3/48 w/ 1pt allowance of side/total would group -2 to -4 w/ 47 to 49 as "similar games."

Let's say we do that. How do we proceed w/ the information? Most leagues aren't going to have systematic error. So, average line of 3/48 is going to produce output CLOSE to those numbers.

Are u saying that one angle might be parlays? IE, side/total in that ballpark might play out as close/tight lending to dog/under? Call me pessimistic...but i HOPE that u prove me wrong.

There is some allowance for games near the lines. But I don't use the approach you described. By selecting a minimum number of games we end up with some exact hits plus other nearby games. I do measure the "distance" from today's game but haven't yet incorporated that in selecting the "like" games. I'll do a major review before the football season to fine tune the selection process.

You are correct, the resulting distribution of scores is so close to the lines that it doesn't produce any betable situations. Last year in the NFL it would always be within 1-2% of the spread. For instance, a line of -5.5/48 would generate a distribution of 400 games near there. When you look at these results you would have visitors covering 51.3% of the time and home team 48.7%. At -110, neither would have an expected return over 1.00.

But you can use the distributions to evaluate other types of wagers. Money lines, parlays, buying points, first half scores, and so on. Football in particular is rife with key numbers and this approach would incorporate those.

[ChuckyTheGoat]

OK, Bsims. Good luck w/ the research.

If u want to work together on data-analysis, let me know. I'll talk to u next week.

[phinfan27615]

First the guy who was going to bet off of madden sims now this? Just go with your gut or fade some of sbr's big name posters, I feel like utilizing a system would take the fun out of it and also would end up losing

[Bsims]

First the guy who was going to bet off of madden sims now this?

Bigredfyah was the poster who suggested using Madden to simulate games. I suggested that might be fun (I wouldn't know since I've never seen the game) but not useful. I did spend a lot of time explaining the complexity of trying to use simulators.

[JoeCool20]

Yeah, I applaud the effort. But there ain't no "short cuts" or "winning systems" or 99% of us would have already found it out and made a killing on it!
There is no use driving yourself crazy delving into random figures or situations like this.

Why not take an arbitrary thing like going back and looking at every game ever played with the line opening at....
say 6 flat.

Then see what the overall win percentage of dogs vs favs is on that line. Then the next line that is 6 flat, you bet it.

Well, as soon as you put your money down, you will know that all previous results don't make a flying damn on the outcome of the game that is about to be played!

[chilidog]

Yeah, I applaud the effort. But there ain't no "short cuts" or "winning systems" or 99% of us would have already found it out and made a killing on it!
There is no use driving yourself crazy delving into random figures or situations like this.
Why not take an arbitrary thing like going back and looking at every game ever played with the line opening at....
say 6 flat.
Then see what the overall win percentage of dogs vs favs is on that line. Then the next line that is 6 flat, you bet it.
Well, as soon as you put your money down, you will know that all previous results don't make a flying damn on the outcome of the game that is about to be played!

That system was actually incredibly popular for a long time, but died out 2 seasons ago. Books adjusted, just like they always do when a winning system is found.

[SsgKen]

ESPN ran the Madden (latest model) last summer and the KC Chiefs never beat the Pats in 1000 trials. I think using computers for the NFL is not helpful. There are only a few games in a season compared to MLB. I like to play in-game bets because there are so many events to bet on you know a computer is putting up the lines and they are beatable.

[TKIRK2483]

I have a excel workbook built up with something similar to what you're aiming at and has decent success, running 22% ROI since June 1st.

I pull oddsshark data for a team's home/away for last 30 games and have a macro that removes everything but the moneyline and W/L columns, removes 100 out of the moneyline value (Ex. -120 turns into -20 for linear equation purposes, otherwise there's a huge gap in-between 100 and -100), turns the W/L into 1/0 values, then uses the forecast function to turn the data into percentage chances to win according to the moneyline value. Chart the forecast data into a linear equation (y=ax+b) and setup the equation so you can enter x and it spits out the y-value.

Ex: The linear equation for Washington at Home comes out as y=-0.001418x+0.441554. If Washington is a -147 moneyline, put in x as -47 and it spits out y=50.82%.

Have a sheet with the table built up of all formula values for each team/location that's a data reference for a dropdown menu on the daily worksheet.

Made a daily template worksheet I'll enter in the matchups with the predicted values and then scale the sum of the two predictions into a 100% total.

Example for today's games, Oakland @ San Francisco

Oakland on road: y=-0.001150x+0.730584 (+137) = 68.80%
San Francisco at home: y=-0.003634x+0.579927 (-147) = 75.07%

The totals equal 143.87%, scaled down to 100% converts to:
Oakland: (+137) 46.87%
San Francisco: (-147) 53.14%

I have betting methods I'm still playing around with from that point, kelly criterion, flat bets over a certain percentage, etc. but overall I've been pretty pleased with it. You're essentially picking up the odds that the book is leaving on the table.

[Bsims]

Welcome TKIRK2483, good 1^st post. My initial reaction was another nutty data mining system. But after more thought, I think it might require more thought on my part. Since you referred to starting June 1^st, I assume you have not back tested. This is a system that could be back tested.

My 1^st concern was impact of pitchers. But then I realized the past and current ML’s would account for them. Interesting way of handling odds. I always convert those to European Odds (-120 becomes 1.833). I think I might take a look at back testing. If so, I may have some questions. Good luck.

[ChuckyTheGoat]

Tkirk, I agree w/ Bsims. Outstanding 1st post. Good luck, keep bangin it out.

[gojetsgomoxies]

good topic.

i think the title doesn't quite fit the subject..... to a very high degree, the bookies are just balancing action so you aren't taking advantage of them. you are taking advantage of the betting public.

basically, i think you are taking advantage of betting biases by the general public...

i have talked about this before but i think one of the most important things most of us should do is constantly evaluate betting systems and try to determine whether it's real or not, how much data mining is involved and what could change (team play changes or public betting changes)...

here are 2 examples i'm aware of amongst many others,

BigDaddy (whose analytical acumen i respect) loved small road favourites in conference games.

and here's how it worked 2011 to now, -2 to -7 inclusive...... 2011-17: 42, 60, 59, 61, 58, 50, 50... so it worked really well for 4 years. break-even last 2 years.... not a huge trend in # of games but definitely fewer the last 2 years (numbers not shown)

what about conferences? all big winners except conference USA and SEC which are 35% and 47%..... why is USA so bad? WKY, UTSA, Rice, UAB are terrible.... we defintely didn't learn much from looking at USA

so that's the kind of analysis i like on this type of stuff

[JoeCool20]

LOL Trying to figure out an "edge" in a sporting event, or in games of chance, is such a waste of time.

Because there are no such "advantages" or "edges" to be found.

[jts1207]

LOL Trying to figure out an "edge" in a sporting event, or in games of chance, is such a waste of time.

Because there are no such "advantages" or "edges" to be found.

Congrats on one of the dumbest things I have ever read.

[chilidog]

LOL Trying to figure out an "edge" in a sporting event, or in games of chance, is such a waste of time.
Because there are no such "advantages" or "edges" to be found.

Math says you're wrong.

[Bsims]

Example for today's games, Oakland @ San Francisco

Oakland on road: y=-0.001150x+0.730584 (+137) = 68.80%
San Francisco at home: y=-0.003634x+0.579927 (-147) = 75.07%

The totals equal 143.87%, scaled down to 100% converts to:
Oakland: (+137) 46.87%
San Francisco: (-147) 53.14%

I'm not clear on how you got Oakland 46.87% and San Francisco 53.14%. If you divide each of their original percentages by the combined 143.87% you get Oakland 47.82% and San Francisco 52.18%. Have you considered Bill James Log5 approach to combining the percentages which would give you Oakland 42.27% and San Francisco 57.73%?

Also, do you just look at the last 30 games and only keep the home and away games for your regression. This would leave about 15 games in each regression?

[TheMoneyShot]

No offense but there's no formula that would allow you to have any leverage against a sportsbook.

Just do your best to pick winners. Stay away from double/triple juiced LIVE LINES....

Only thing that will ever allow you to gain momentum... and a winning streak.... great money management skills... and don't be so greedy on EXTRA PLAYS when you're + for the day.

There's nothing else you can do. You have to operate your mind... and your own system... like a MACHINE.

[chilidog]

No offense but there's no formula that would allow you to have any leverage against a sportsbook.
Just do your best to pick winners. Stay away from double/triple juiced LIVE LINES....
Only thing that will ever allow you to gain momentum... and a winning streak.... great money management skills... and don't be so greedy on EXTRA PLAYS when you're + for the day.
There's nothing else you can do. You have to operate your mind... and your own system... like a MACHINE.

There definitely is, and the 2 best ways are:

1. setting your own lines and comparing them against sharp books like pinny and sbobet
2. comparing soft lines at rec books against sharp books

once you find bets that give you a decent edge (3-5% if you're not a fan of longshots, increase to 15% if you can handle the long losing streaks from the longshots), hammer them

[dante1]

I disagree on this. I don't think much can be gleened from past odds.

1. You are not counting injuries
2. Past results did not incorporate new information from the games that already happened, IE you didnt know how well a team with mesh together at a season, or how players would be faring in the new.
3. Your definition of like games is not a clear one, do we mean same teams, or a -3 favorite playing on a road with a 48 total in NFL.

This seems like one of these gimmick things that people find through backtesting and then get killed when they put into practice.

Agree.

I hate to piss on another person's thoughts concerning gambling because often they are valid. The thoughts are valid but placing them into action just doesn't seem to work. I have been through these systems, like many of you, they usually fall apart in the long run. For example, I had some limited success with line movement, and then it exploded.

However, I will never stop listening to these ideas, because who knows. The bottom line is bookies always win and that has been the history of sports gambling since day one. Maybe, I am too free with the word always, let us say nearly always. I booked for two decades and I bitched along with all the other bookies. It was BS. We never had a losing year, we never had a losing season, we seldom had a losing week.
And at the end of the year the small book we had cleared many tens of thousands. During football season weeks were often in the many thousands, and this was with less than 40 players. Often on a Sunday 7 or 8 pages of book only represented a couple dozen players. I can't tell you how many times not one net winner for the day. Hard to believe but true.

[TheMoneyShot]

There definitely is, and the 2 best ways are:

1. setting your own lines and comparing them against sharp books like pinny and sbobet
2. comparing soft lines at rec books against sharp books

once you find bets that give you a decent edge (3-5% if you're not a fan of longshots, increase to 15% if you can handle the long losing streaks from the longshots), hammer them

All I'm saying is... you'd need that robot in the movie War Games.... (what was his name?) Joshua? To simulate all of these percentages/outcomes. There's not enough time (in a day) to figure out every game... in every sport on the board. How? Furthermore, the only chance you'd have is to center on one game. Even then.... you're still not putting the odds in your favor.

People thought of this for the last 2 decades...

To humanly accomplish this.... is impossible.

[chilidog]

All I'm saying is... you'd need that robot in the movie War Games.... (what was his name?) Joshua? To simulate all of these percentages/outcomes. There's not enough time (in a day) to figure out every game... in every sport on the board. How? Furthermore, the only chance you'd have is to center on one game. Even then.... you're still not putting the odds in your favor.
People thought of this for the last 2 decades...
To humanly accomplish this.... is impossible.

No doubt; it would suck if somebody tried to do that manually. Way too many games, lines, betting options, and books. But if somebody was serious about it, spend a few grand on an overseas developer and quietly rake in your profits until you've been limited just about everywhere. People should keep betting for entertainment; for the thrill. You're not going to beat the books long-term, especially as an American. It's just not going to happen. Way more money on the other side of the table.

However, if you have access to global accounts, you can make a literal killing doing exactly this. There are thousands of global rec books out there.