Can NHL/MLB Be Handicapped based on Team Odds History and Current Team Odds Alone?

[oilcountry99]

I'm wondering if value selections can be made based on lines alone. Meaning games are capped strictly on historical and current odds only, no game stats. Injuries would be taken into consideration but that should be in the line. I'm going under the assumption that the bookmakers know what they are doing and everything is incorporated into the odds. Why research all these stats and numbers when it's already been done. The goal is to find the game with the most value...what team is priced right for a buy signal or sell signal.

Can it be as simple as converting a teams win% to odds and comparing it to the game odds?

Just trying to look at a different approach for myself personally. Anyone with any experience, advice, tips, articles etc. please share. Thanks.

[HeeeHAWWWW]

Unfortunately, if it was that easy, everyone would do it, and the edge would disappear quickly. Same with any widespread public info.

[oilcountry99]

HH thanks for the reply. Realistically I know it can't be that simple, but i don't think it has to be complicated either. I believe money management is more important than the selections themselves. We all win and lose that's a fact. The key is winning long term and finding value in the line. In some games there is probably no value in either team, we want to be on the teams that are priced right, regardless of who that team is. I personally can't make a line better than the books, who can? Probably someone who should be working for them LOL. So how can we use the books and past lines to our advantage? I'm just not sure how to go about it and if it has merit. There have been many more intelligent people that have come before me and I'd be nieve to think all of these strategies haven't bee studied somewhere along the line.

[HeeeHAWWWW]

HH thanks for the reply. Realistically I know it can't be that simple, but i don't think it has to be complicated either. I believe money management is more important than the selections themselves.

Well, money management is important, but all you can do with it is twiddle your variance vs bankroll growth. It's not adding anything though - ultimately your edge is what makes the money, and stake sizing can only hurt that, not improve it.


As for making a line better than the books: surprisingly, that's not always that difficult. If you look at the history of betting markets, the price often bounces around all over the place. With tight juice (which means you need multiple books and minimal transaction costs), that means at some stage in the market history, there is value.

For example, let's take one match yesterday from my sport (tennis): Pouille vs Struff. This opened at 1.549 2.64, and it closed at 1.433 3.06. Let's assume the closing price is accurate, for roughly a 69%/31% split. If you'd bet Pouille on the opener (the highest price it ever reached), you'd have a 7% edge. The opposite is also true, as Struff reached 3.59 at one point, for 11% edge.

Most markets are less volatile, but there still an enormous number of such spots .... and that's just one book. You just need a model able to price accurately, and notice when that value is available (for the latter, bots are often used now).

[oilcountry99]

You just need a model able to price accurately, and notice when that value is available (for the latter, bots are often used now).

How does one go about creating such model? How do you know when your model is pricing accurately? (by beating the closing line?) The model you run for tennis, is it complex? How many indicators do you incorporate?

I appreciate your information and posts.

[Waterstpub87]

How does one go about creating such model? How do you know when your model is pricing accurately? (by beating the closing line?) The model you run for tennis, is it complex? How many indicators do you incorporate?

I appreciate your information and posts.

You go about creating a model by doing research into statistics that are predictive towards game results. You know you model is accurate when:

1. For most games, you get a very small predicted edge against the closing lines
2. You beat the closing line generally when you look at your results (Make sure you are testing out of sample)

Example of model I built for baseball:

Baseball: Batters 7 Stats, Pitchers 7 Stats, fielding 1 stat plus building a game simulator and running 1000 simulated games per actual game bet

[HeeeHAWWWW]

How does one go about creating such model? How do you know when your model is pricing accurately? (by beating the closing line?)

As waterspub says above, plus there are resampling methods common to all forms of predictive modelling. You have a metric, eg auc or logloss, and you try to maximise (or minimise) it as appropriate.

The model you run for tennis, is it complex? How many indicators do you incorporate?

About 1200 now. Most are inactive most of the time, but occasionally they'll add a few % here or there, and thus make for value.

[oilcountry99]

You go about creating a model by doing research into statistics that are predictive towards game results. You know you model is accurate when:

1. For most games, you get a very small predicted edge against the closing lines
2. You beat the closing line generally when you look at your results (Make sure you are testing out of sample)

Example of model I built for baseball:

Baseball: Batters 7 Stats, Pitchers 7 Stats, fielding 1 stat plus building a game simulator and running 1000 simulated games per actual game bet

Thanks, and how does your baseball model perform? How long did it take to develop?

[Waterstpub87]

Thanks, and how does your baseball model perform? How long did it take to develop?

The season before every season, I am working on the next seasons models so Fall=Basketball, Winter= Baseball and now Tennis, Spring= Football.

My baseball model is built on cumulative experience over 3 years or so. The 1st version of the monte carlo model I posted above went into action last summer. Actual units wise, I lost roughly 20 units. Based on where I bet and where the line closed, I should have made roughly 10 units, this being counted on edge vs closing line times money bet. Variance comes into effect, can't really do much about it.

I am looking forward to putting the newest version into effect in 2 weeks.

[u21c3f6]

oc99, I think you have received some excellent replies. Let me add my 2 cents. My focus is I think a little different than most who wager on sports. I wrote the following in a different thread and of course it applies to ATS wagers but the same argument could be made for ML wagers which I will get into after the following:

All closers are approx 50/50. If each individual closer was also 50/50, then the only way to win would be to beat the closer enough times to cover the vig and profit. I don't see it that way.

I agree that all closers approximate 50/50 but each individual game's closing is not 50/50. My hypothesis is that there is a group of closers that are 65/35, 60/40, 55/45, 50/50, 45/55,40/60, 35/65 etc so that the overall closer is approx 50/50. The trick is to find that subset of games that are not 50/50 and wager on the appropriate side so that your winning % will be greater than 50/50.

The above could be applied to ML wagers the same way where a -155 +145 market should be approx 60/40 but I would hypothesize that there may be games that are 75/25, 70/30, 65/35, 60/40, 55,45, 50/50, 45/55 etc so that the overall is 60/40 but the individual games are not necessarily 60/40.

Now, how do we find those wagers? You begin by tracking some data that would apply to our hypothesis and analyze the results. Our hypothesis does not even have to be correct. We just need results that have a clear line of demarcation between expected results and results of our experiment. If there is no clear line of demarcation then it is assumed that the data we are using is fully included in the odds and therefore has no value.

As an experiment I will use odds history and current odds as stated in your first post. I will describe method a little later I just want to get this selection in before game time:

Columbus -113

[u21c3f6]

Rest of selections:

Winnepeg +152 (This one is interesting because it is a double signal, explanation to follow)
Pittsburgh -102
St Louis +150

Someone please let me know if I should continue to post the selections here or if I should move this to the NHL forum. Tx. Joe.

[oilcountry99]

Keep posting....very interested. Relates to topic at hand!

[u21c3f6]

Of course one should try to get the best odds available to them. Please note that this is an experiment. It is relatively simple but let us see what the results show. We are really looking for either a profit or a larger than expected loss which would mean that I may be looking at things "backward". The selection "process" and hypothesis is as follows:

1. Using Pinnacle's closing odds for a teams previous game, the team is a selection if they lost their last game to a + odds team (assumption being that they may be under estimated for their next game due to the loss to an underdog). If both teams qualify then there is no wager.

2. Using Pinnacle's closing odds for a teams previous game, if the team won their last game as a + odds underdog, then wager on the opposing team. (assumption being that they may be over estimated for their next game due to their win as an underdog). If both teams qualify then there is no wager.

That's it, let's see what happens!

I would not try this with MLB since the odds fluctuate wildly based on pitcher and is not indicative of overall team performance.

Joe.

[HeeeHAWWWW]

All closers are approx 50/50. If each individual closer was also 50/50, then the only way to win would be to beat the closer enough times to cover the vig and profit. I don't see it that way.

Yeah - this wasn't my point though, it was just to make an example. Here it doesn't matter if it's true or not tbh, just that the price movement shows that by definition there is value at some stage in the market, because the movement is greater than the juice.


Anyway, if you want to measure closer accuracy, there are good ways to do it: this is a classification task, you have the outcome and a predicted probability, so what's needed is a proper scoring rule (https://en.wikipedia.org/wiki/Scorin..._scoring_rules). There are a few of them, but since you're close to evens something like pseudo-spherical or Brier might be a good idea.

[jtoler]

I love threads like this, reminds me that sbr isnt so stupid after all.

[BigdaddyQH]

I believe that it is a matter of confidence. Many people in here have thrown out some very interesting theories and strategies, but the bottom line is that the only one that will work is the one you are comfortable with. Any strategy will work if it consistantly wins, but that simple des not happen. You have to be confident and comfortable with your strategies when they lose also. If not, hesitation creeps in and that can be a killer.

[LT Profits]

Of course one should try to get the best odds available to them. Please note that this is an experiment. It is relatively simple but let us see what the results show. We are really looking for either a profit or a larger than expected loss which would mean that I may be looking at things "backward". The selection "process" and hypothesis is as follows:

1. Using Pinnacle's closing odds for a teams previous game, the team is a selection if they lost their last game to a + odds team (assumption being that they may be under estimated for their next game due to the loss to an underdog). If both teams qualify then there is no wager.

2. Using Pinnacle's closing odds for a teams previous game, if the team won their last game as a + odds underdog, then wager on the opposing team. (assumption being that they may be over estimated for their next game due to their win as an underdog). If both teams qualify then there is no wager.

That's it, let's see what happens!

I would not try this with MLB since the odds fluctuate wildly based on pitcher and is not indicative of overall team performance.

Joe.

Not the same thing but another very simple system based on the practically the same theory:

Sub-.500 favorites coming off of a loss facing any team coming off of a win are 742-469 with an ROI of +6.7% since the 2005-06 season. This season alone is 69-42 with ROI of 8.1%. Clarification: to determine if team is sub-.500, regulation losses and overtime/shootout losses count the same, so a team that is 10-5-6 would actually be sub-.500 (10-11).

Keep in mind that I personally rely on models, but if model agrees with play this system supports, it makes me feel good.

System play for Tuesday would be Florida if they remain favored.

[oilcountry99]

Anyway, if you want to measure closer accuracy, there are good ways to do it: this is a classification task, you have the outcome and a predicted probability, so what's needed is a proper scoring rule (https://en.wikipedia.org/wiki/Scorin..._scoring_rules). There are a few of them, but since you're close to evens something like pseudo-spherical or Brier might be a good idea.

I appreciate your sharing...this however is over my head, but interesting. How would this be applied in a practical sense related to what we are doing? Sorry for my ignorance

[u21c3f6]

Today's (3/14) plays (discussion later as to what to do with the data):

NY Islanders 105 (double signal)
Ottawa -125
Chicago 109
Edmonton -160
San Jose -199
LA -240

Joe.

[u21c3f6]

No selections today (3/15).

OK, what to do with the data. A database needs to be created to capture the data which can then be queried to find any lines of demarcation in any of the various combinations of data. Let me explain this a little further. We are not looking to datamine. If you have to come up with too many rules and qualifiers and exceptions to come up with a selection, it is probably datamined. What you should see is a clear consistent pattern such as the extremes of the data either being profitable or unprofitable (which then means that the central data may be profitable). Or a clear path of profitable or unprofitable selections as you move from one end of the data to the other end. For example, I am already suspecting that huge money line favorites probably can't be undervalued by very much if at all. But that should begin to show when you have enough data.

What data fields do we have? Frankly, even with this relatively simple method there is a lot of data fields we could collect. Such as:

Whether the selection was based on criteria 1 or 2
What were the odds
Result of selection
Opening odds
Closing odds
For criteria 1 - opposing team's previous game result and odds
For criteria 2 - selection team's previous result and odds
Etc

Additional ideas may be: Home/Away, Goalie, etc

OC99 (and others), Hopefully this gives you some ideas of how to proceed to look at what you asked in your original post. Enough for now. Joe.

[TechnicalTrader]

Check your inbox Oil...

[TechnicalTrader]

And here is your simple proof:

http://killersports.com/mlb/query?su=1&ou=1&sdql=AF+and+season%3D2012%2C2013%2C2014%2C2015%2C2016&submit=++S+D+Q+L+%21++

By only playing away favorites over the last 5 seasons (straight up), you would be in the plus. Not by much but you would be in the green. Show this to anyone on Wall Street and they will give you a serious look. The problem is, in sports betting, 99% of bettors bet with emotions and lack self discipline.

If your relationship manager at Chase told you that he can offer you a product which is guaranteed to return 0.5% to 1%, from April until September, you would be a complete idiot to turn it down. That link above is proof that you can earn a GUARANTEED 0.5% to 1% ROI by risking only 1% of your BR per bet ($100 per bet on a $10K BR). Hell risk $500 per bet (5%) and walk away with and average of $3,750.00 per season). Most days will see 3-8 bets and that is why I would not recommend betting 5%, but with incremental performance, you are looking at numbers way higher than 0.5%....

Not that difficult, my advise is to keep betting simple and aim for 0.1% BR profit per day.

[TechnicalTrader]

Not the same thing but another very simple system based on the practically the same theory:

Sub-.500 favorites coming off of a loss facing any team coming off of a win are 742-469 with an ROI of +6.7% since the 2005-06 season. This season alone is 69-42 with ROI of 8.1%. Clarification: to determine if team is sub-.500, regulation losses and overtime/shootout losses count the same, so a team that is 10-5-6 would actually be sub-.500 (10-11).

Keep in mind that I personally rely on models, but if model agrees with play this system supports, it makes me feel good.

System play for Tuesday would be Florida if they remain favored.

Not sure how accurate your numbers are but games from the 2012 season to date are 764-912 with that criteria:

http://killersports.com/nhl/query?sdql=WP%3C50+and+p%3AL+and+op%3AW+and+season%3E%3D2012&submit=++S+D+Q+L+%21++

ROI -0.5%

http://killersports.com/nhl/query?sdql=WP%3C50+and+p%3AL+and+op%3AW+and+season%3D2016&submit=++S+D+Q+L+%21++

2016-17 season is 165-177, RoI = 2.2%

Let me know if I misunderstood anything...

[LT Profits]

Not sure how accurate your numbers are but games from the 2012 season to date are 764-912 with that criteria:

http://killersports.com/nhl/query?sdql=WP%3C50+and+p%3AL+and+op%3AW+and+season%3E%3D2012&submit=++S+D+Q+L+%21++

ROI -0.5%

http://killersports.com/nhl/query?sdql=WP%3C50+and+p%3AL+and+op%3AW+and+season%3D2016&submit=++S+D+Q+L+%21++

2016-17 season is 165-177, RoI = 2.2%

Let me know if I misunderstood anything...

I said sub-.500 FAVORITES. You need to add odds < -105.

[TechnicalTrader]

My bad, excuse me!

Here we go:

YTD:

http://killersports.com/nhl/query?sdql=F+and+WP%3C50+and+p%3AL+and+op%3AW+and+season%3D2016&submit=++S+D+Q+L+%21++

2010 season TD:

http://killersports.com/nhl/query?sdql=F+and+WP%3C50+and+p%3AL+and+op%3AW+and+season%3E%3D2010&submit=++S+D+Q+L+%21++

Very nice!

Add a day rest and you have yourself a sweet and simple system:

http://killersports.com/nhl/query?sdql=F+and+WP%3C50+and+p%3AL+and+op%3AW+and+season%3E%3D2010+and+rest%3E0.5&submit=++S+D+Q+L+%21++

[rkelly110]

Hi guys. The stuff you guys are talking are way above my pay grade.

I use the KISS rule. Keep It Simple Stupid.

I have at least 5 years of data to back up what I'm about to say.

ML fave of the day minimum odds -222 (1.45) or higher has produced over 65% win rate for the whole season
in both MLB and NHL. Matter of fact for every sport except NFL (Nov a good month).

Yes, you can still lose money betting fav's, but not if you slightly increase wagers after a loss/s.
It's a marathon, not a get rich quick scheme.

Let the books work for you. Over 60% of favs win.

[TechnicalTrader]

Hi guys. The stuff you guys are talking are way above my pay grade.

I use the KISS rule. Keep It Simple Stupid.

I have at least 5 years of data to back up what I'm about to say.

ML fave of the day minimum odds -222 (1.45) or higher has produced over 65% win rate for the whole season
in both MLB and NHL. Matter of fact for every sport except NFL (Nov a good month).

Yes, you can still lose money betting fav's, but not if you slightly increase wagers after a loss/s.
It's a marathon, not a get rich quick scheme.

Let the books work for you. Over 60% of favs win.

Have you had success playing that? Just taking a brief look at the 2016 MLB season and I see a losing streak of 5 games from July 4th until July 9th:

Nats -250
Cubs -255
Cubs -265
White Sox -295
Jays -230

Winner; July 10th, Giants -245

assuming you increase your wagers by (only) 0.5% per bet, and bet to win (only) 1% BR, this chase right here would cost you:

2.5%
3.83% (2.55*1.5)
5.1% (2.65 *2.0)
6.63%
8.85%
8.05%

35% of BR which would've completely wiped out your winnings...

[Waterstpub87]

Have you had success playing that? Just taking a brief look at the 2016 MLB season and I see a losing streak of 5 games from July 4th until July 9th:

Nats -250
Cubs -255
Cubs -265
White Sox -295
Jays -230

Winner; July 10th, Giants -245

assuming you increase your wagers by (only) 0.5% per bet, and bet to win (only) 1% BR, this chase right here would cost you:

2.5%
3.83% (2.55*1.5)
5.1% (2.65 *2.0)
6.63%
8.85%
8.05%

35% of BR which would've completely wiped out your winnings...

You've missed the great part: to profit even from a -222 you would need to hit ~69%. So even without that, he should be losing.

[TechnicalTrader]

You've missed the great part: to profit even from a -222 you would need to hit ~69%. So even without that, he should be losing.

I understand what he is trying to do and a three game chase "would" work, even with those odds, anything higher than a three game chase would hurt your BR.
Up until that stretch, the system only saw three 3 game chases and 31 either straight up winners or two game chases.

[rkelly110]

Have you had success playing that? Just taking a brief look at the 2016 MLB season and I see a losing streak of 5 games from July 4th until July 9th:

Nats -250
Cubs -255
Cubs -265
White Sox -295
Jays -230

Winner; July 10th, Giants -245

assuming you increase your wagers by (only) 0.5% per bet, and bet to win (only) 1% BR, this chase right here would cost you:

2.5%
3.83% (2.55*1.5)
5.1% (2.65 *2.0)
6.63%
8.85%
8.05%

35% of BR which would've completely wiped out your winnings...

Yes, there are some losing streaks, but also some long winning streaks.

Let's just say that I have been in the positive at the end of the season.

NHL to date for this year is 100-49 67%. March has only lost twice so far. Now tell me how much money you
could stuff in your account?

-222 (1.45) occur rarely, most are higher. You are looking for the ML fav of the day (one game) in each sport.
If there is more than one game with the same odds, either bet them all or pass. I pass.

[TechnicalTrader]

My bad, I was looking at odds lower than -222

[Waterstpub87]

Yes, there are some losing streaks, but also some long winning streaks.

Let's just say that I have been in the positive at the end of the season.

NHL to date for this year is 100-49 67%. March has only lost twice so far. Now tell me how much money you
could stuff in your account?

-222 (1.45) occur rarely, most are higher. You are looking for the ML fav of the day (one game) in each sport.
If there is more than one game with the same odds, either bet them all or pass. I pass.

Is it like your standard martingale? so if you lose a -300 favorite at 300, you would bet 1200 the next day to win it back plus 100 dollars?

[rkelly110]

Is it like your standard martingale? so if you lose a -300 favorite at 300, you would bet 1200 the next day to win it back plus 100 dollars?

Wouldn't be betting a -300. -222 is the minimum.

NHL not so much, MLB yes, using the minimum odds as a guide, bet the -RL at plus odds and martingale off of that. I have no info on this. Maybe TT can help?

I don't martingale high odds, that's a recipe for disaster. I just gradually increase my bets and make my money during a long winning streak.

[rkelly110]

I understand what he is trying to do and a three game chase "would" work, even with those odds, anything higher than a three game chase would hurt your BR.
Up until that stretch, the system only saw three 3 game chases and 31 either straight up winners or two game chases.

You were in my thread last year. Lost 3 SU and ATS, start a 3 game chase. What I'm talking here is different.

[Slanina]

My bad, excuse me!

Here we go:

YTD:

http://killersports.com/nhl/query?sdql=F+and+WP%3C50+and+p%3AL+and+op%3AW+and+season%3D2016&submit=++S+D+Q+L+%21++

2010 season TD:

http://killersports.com/nhl/query?sdql=F+and+WP%3C50+and+p%3AL+and+op%3AW+and+season%3E%3D2010&submit=++S+D+Q+L+%21++

Very nice!

Add a day rest and you have yourself a sweet and simple system:

http://killersports.com/nhl/query?sd...+S+D+Q+L+%21++

What the hell is this? Why are we all grinding away when this is sitting here. And where was this in October....