Beating the bookies – how the online sports betting market is rigged

"Send enquires about this paper to our newly acquired private island in the Bahamas." :-)

One of the most depressing things I realized when I learned to count cards, and confer upon myself a small but meaningful advantage in the game of Blackjack, was that the casinos simply ask you to leave if you win too much. That put an upper limit on the rate at which one could win. The folks who figure out slot machines have a much better time of it because it takes longer for the casinos to figure out they are losing money.

There has to be more to this story. I would like to hear the bookmakers side of this. Perhaps the researchers were scripting or triggered some other kind of security tripwire.

Winning $900 split across several different bookmakers is absolutely nothing in the sports betting industry.

William Hill, one of the companies that the researchers claim restricted them is a multi billion dollar company. They aren't sweating small time bets like this.

EDIT: I noticed that the screenshots they used as proof their bets were restricted are for bets on very minor football leagues (Australian semi pro football), its common for betting limits to be lower for games that don't see a lot of betting action & is not proof enough to me that the bookmakers lowered their limits globally

I've been trying my luck at tennis betting ( https://matchstat.com/profile/soniman ) and the only strategy that seems to work consistently is finding injured players and betting against them. Good example of an injured player is a player that limps (Andy Murray at Wimbledon this year). The computers don't watch matches so injured players can be overvalued by the models. Also, any British tennis player tends to be overvalued because the books are based in the UK. For instance there is an overweight British player named Marcus Willis that last week was a 1.01 favorite to win at a tournament in Las Vegas; he lost, apparently due to dizziness aka being fat. There are some other patterns that can be exploited but the highest return strategy has to be betting against injured favorites. However, I think that would probably require more tennis watching than I want to do. Maybe I could hire somebody on Mechanical Turk to watch tennis for me and report on players who appear to be injured or take medical time outs.

> A few weeks after we started trading with actual money some bookmakers began to severely limit our accounts, forcing us to stop our betting strategy.

Isn't this like literally one of the oldest tricks in the book? I remember reading Reminiscences of a Stock Operator, which talks in part about early 1900's bucket shops, and the same stuff was there even then. Similar stuff is also mentioned in market microstructure textbooks with market makers on one side and informed traders on the other side.

Is rigged even the right word here? It might be, but did the bookmakers have a responsibility to keep accepting their bets? Is it different from claiming that casinos are rigged?

I always thought bookies just kept moving the lines until they were making money at the snap of the ball. Their initial line isn't as important as adjusting the line so that there is equal money on both sides of the line and therefore the bookie is guaranteed money due to the small fee they build into the bets. Their profits don't depend on accurately predicting the game's score, just moving the line strategically as more bets come in.

Yes, of course you can beat the "popular" bookmakers.

Once you start beating them ( being profitable in value prices ) they will simply close/ban your account. Nowadays, it happens extremely fast ( in a day or a few hours, depending on your moves ). It's a well known tactic, and in practice, you cannot do anything about it ( other than keep opening new accounts in new names ).

Try beating a betting exchange.

"Retail" bookmakers are only interested in mugs. If your betting patterns indicate someone who is wise to the market, you will be limited or shown the door. Be prepared to have arbitrage positions pulled out from under you. The game is rigged insofar as the book decides if it wants to entertain your position.

If you want to make money you have to bet against, and be able to beat the books that know what they are doing - The high limit, low margin books like Pinnacle, SBO, IBC et al will happily take you on.

Not mentioned in the abstract is that they also used real money:

> During that period we obtained an accuracy of 47.% [sic] and a profit of $957.50 across 265 bets, equivalent to a 8.5% return (Table 1, Figure 3).

For some reason the "ok but how much did you ACTUALLY MAKE?" is always my favorite part of this kind of business or economics literature.

> A strategy intended to beat the bookmakers at predicting the outcome of sports games requires a more accurate model than the ones bookmakers have developed over many years of data collection and analysis.

I disagree with this assumption and I think they have painted themselves into a corner because of it. To illustrate, imagine charting win rates against bins of price-implied-chances. $3 horses win roughly 33% of the time, $4 horses 25% for example. It resembles a noisy 1:1 linear relationship. Do the same for your selections and your line will be noisier, but crucially you're not taking bets where the price is worse than your estimate. This can leave a window of profitibility when you subtract the two, even when you are less 'accurate' as measured by win rate or KLD or other measures.

The goal is profitibility, not accuracy. The problem with including the odds you are betting against as a feature for your ensemble is that it dampens that window. If you're right about your selections, you'll bet less and win less. * If you're concerned about the volitility that comes with being less accurate, there are better ways to address that.

I've been doing this for a couple of years and in many ways it's a dream side-project. Location independent, no customers, automatable, and in some jurisdictions tax-free. It can be a little lonely at times though. I would love to chat with anyone else applying tech/math to beat the bookies. Sorry for the throwaway, I'll put a contact in my profile.

What the authors are doing is “chasing steam” and most books will ban or limit you if you try this. This should come as no surprise to people with experience in the industry. The books aren’t limiting/banning them for winning - it’s the way in which they were winning. They are betting slow moving books’ lines. Their strategy only works at poorly managed books as long as they can bet quickly. I doubt they made any bets at a sharp friendly book like pinnacle.

If anyone would like to collaborate with some model building get in touch. I already have a large db of most of the stats you’d ever need and some okay but not amazing models for most major sports.

A couple of suggestions to improve on this method.

The authors' regression left an intercept or 'adjustment term' of 3.4% - 5.7%. For a perfect bookmaker, this intercept term would be equal to the overround. The number calculated unfortunately averages that overround between different bookmakers and at different times (overrounds often decrease over time). It might be more effective to adjust for the actual overround of each market sampled, i.e. divide each price by the sum of the inverse of the prospects.

They appear to use a flat betting strategy, and the threshold to bet or not was selected based on profitibility. I was simplifying in another comment when I said profitibility should be the goal. In reality it's utility you should be optimizing for. Nobody wants a ultimately profitable system that reads like an EKG, they want a high sharpe ratio. The paper's results are actually very good here, but the trend could be lifted and stabilized further by betting proportionally to expectation, or by explicitly optimizing for such.

Link to the code used from the PDF:

https://github.com/Lisandro79/BeatTheBookie

They claim the market is inefficient. I worked for a company that to some extend fixed that. Can't remember all the details (I worked on a different part), but something like this:

Most big betting companies were customers. They all continuously sent their updated odds to us, and we would broadcast to the other companies. They would react to the change based on certain rules and send new updated odds back to us. This would then converge.

The inefficacy comes from promotions, company X always wanting to have odds .1 better than company Y etc.

Edit: Not sure how it works now, but: https://www.betradar.com/ and https://mts.betradar.com/

Just like the rake or the edge the house advantage of making the rules has always been the reality of betting. Online or offline. Forever. Get too successful and you’re no longer invited to bet.

Asymmetry of information has never been the bookmaker’s most powerful weapon. The book is.

Those who are successful at it accept this reality. They grumble and make peace with it - paying the super taxes and liquidising markets where they’re asked to.

Ultimately however, while it’s interesting to see how they do some of this (and there are plenty of practices not covered in the paper, I assure you) it’s a bit like complaining the DM won’t let you do something in dungeons and dragons - you’re dicing with the god of your domain so the rules can change at any minute.

I used to work for a popular UK online bookmaker. The thing that a lot of these comments are missing is that bookies aren’t going to sell a product at a loss. They’re also not even selling the product you think they are (something akin to an investment).

Bookmakers sell excitement / entertainment - the thrill of the potential win is the product, and costs approximately 10% of what you can afford to stake.

The most interesting thing was left out: how did they find this data? Both historical and realtime is pretty hard to find. Ten years of odds from a dozen bookies looks like a massive task.

Next: how do you mask this behavior to not be obvious. Once you have a betting stratetgy the real difficulty is turning it into one that isn't obvious.

I don't see why we cannot have completely distributed betting platform using para-mutual betting strategy that collects and distributes bets and winnings on the basis of publicly reported sports results with no take-out.

I prefer para-mutual rather than a house deciding the odds. It is a more free-market approach. It has been used in horse racing, but the takeout has been too large which makes it hard to be profitable.

They could have reduced a lot of the work (calculating the mean across 32 bookmakers and applying a constant for the margin) by just taking the price from a highly liquid exchange like betfair, which is pretty close to a 100% efficient de-marginated line.

Although it still wouldn't have prevented their accounts from being limited.

Shouldn't the goal of a bookie simply be to balance his book and have no position on the actual odds of the match? Of course it makes sense to deny action to known sharps, this should come as no surprise.

I don't totally understand why the bookmakers would limit their accounts.

The bookmaker wants to balance his book for each game to make sure he makes a profit no matter what the outcome is. To balance their books they might give better odds for an outcome than what a statistical model might suggest.

But what difference does it make if the bettor who helps them balance their books is a consistent winner or not?

Do they prefer to give these "good" odds to people who are losing money long term?

For anyone interested in running data analysis I have probably the biggest and most comprehensive dataset in the world, bigger that what this team had available.

Personally I've come to find poker is the only reliable way to "beat the house". But that's because you're not beating the house, but other players - the house gets their rake.

If you play a tight-aggressive game in venues the pros avoid (anyplace with less than $2/hour comps in Vegas) you can do decently. Not get rich, but make a few hundred in an afternoon.

Courtesy of slashdot:

https://science.slashdot.org/story/17/10/21/1744218/data-sci...

TL;DR: Researchers made $950 (8.5%) but say it wasn't worth the effort [1].

[1]: https://github.com/Lisandro79/BeatTheBookie

Wanted to post this a few days ago, but life got away on me.

I worked for a monopoly bookmaker and spent a fair amount of time looking at how they work. Our turnover was $2.6bil/yr with $150mil profit in a country with less than 10 million people.

So, Some things to give you guys a bit more context. 1. There are two types of bets. Fixed Odds and Tote. - Tote is a pool based betting system where the odds can change after you have placed your bet. The odds are calculated automatically based on the distribution of bets on the options available. Typically the house will keep 50%+ of the total pool as profit and distribute the remainder among the winning punters. This is a very high profit betting system that the book keepers are trying to keep alive. It's dying off at a pretty rapid rate though. - Fixed Odds Betting (FOB) is where you get payment on the odds you lock in at the time of placing your bet. Most betting now is FOB.

For the sake of responding to various points other respondents have made I will focus only on Fixed Odds; especially as Tote is only used for horse/dog racing.

2. How do the odds work? For us, we had university students who'd manage the books. They had software that showed them how much risk/leverage they had and what the guaranteed profit was. They can set "bet limits" and manually approve (or deny) any bet that was greater than the bet limit. Most of the time they would have open websites from other bookies and copy the odds from theirs as they change. It's quite popular for bookies to just copy each other manually.

For Live/In-Play betting the book keepers will watch the event and manipulate the odds as things occurred. Either using their own knowledge or copying from other gambling sites. Again, the process is completely manual at the back end.

There is a move for organisations around the world to consolidate on their sources of odds (e.g. using a common back-end odds distribution platform); but ultimately there is still a large manual component to changing the odds, especially during live play.

3. How do they make money? On Tote, they take 50%+ of the total pool before creating dividends.

For Fixed Odds, they balance the books. They change the odds to always ensure it's in the houses favour. We always aimed for 10-15% profit on events with fixed odds bets. Home players/athletes will always have much lower odds because of people's tendency to bet with the heart.

They deny bets. The bookie doesn't have to take your bet. For large bets they will often push back an offer to you at a lower rate than advertised to ensure their books stay balanced. For live/in-play bets they'll delay your bet until that odd is no longer available ensuring your bet is not accepted.

They have A LOT of different betting options where only a few will actually win. People tend to bet with their hearts and the number of options are setup to basically ensuring the bookie is profitable.

If you win too much, they shut down your account. They have no obligation to deal with you. Their goal is to make money and they see your gambling as a way you "enhance your enjoyment of the event", not an attempt to make money. So there isn't a large tolerance for people who do make money.

4. How do I (the punter) make money? Surprisingly, you can consistently make money gambling.

Don't bet on Dogs/Horses. Even the top 1% of punters barely break even. They're profitable because of the kick-backs the bookies give them for having high turnover (>$1mil/yr).

Find a sport you know a lot about that supports in-play betting. It's going to be you vs a person. So if you have indepth knowledge of the sport you'll be able to see changes in flow and make winning bets before the bookie notices.

FWIW, I bet on League of Legends. During the LOL Worlds I can make 2500% with >90% win rate. Now, I'm only winning a few $k total so nothing significant.

That's all I can think of at the moment. Happy to answer any questions you have.

How long do you think the authors would have waited to publish this paper if their accounts were not restricted and they would continue to profit? My guess would be a long time.

"Of course the game is rigged. Don't let that stop you—if you don't play, you can't win." -- Robert A. Heinlein, "Time Enough for Love"

A bookmaker's tissue prices (the initial market prices they offer up), are not intended to be a measure of probability. They are meant to be a measure of expected Weight of Money (WoM).

A bookmaker's job is not to accurately reflect odds of occurrence, but to ensure a balanced book of liabilities. There are many books covering this going back hundreds of years and is the principle discovery of those who gathered at Tattersalls coffee shop on the Strand and invented modern bookmaking (via horse racing), and for whom there is named an enclosure on all 55 British racecourses to this day.

The tissue has to be "over-round", (i.e. the probabilities they represent have to add up to over 1.0, or 100%) because sometimes a market will look at the prices, see that the odds are very much in the favour of a selection and act accordingly. As such, the WoM causes a market to move.

This is not news. If you have a reasonable idea of true odds and you are being offered different odds, Kelly (who worked with Shannon - the creator of Information Theory), established the optimal stake to bet at each stage. A substantial amount of research has been done on Kelly Criterion and its application because it underpins many a hedge fund strategy: it works for fixed odds games, fiscal markets and bookmaker markets just as well.

There are trading opportunities here, and there is a wide community of people who look to exploit inevitable market moves using exactly this technique: establish average industry odds, look to where betting exchanges are and bet accordingly, moving out of the market when a profit comes to you.

The bookmakers don't care - you've helping get turnover up, and they know increasing turnover through the market is the best way to get balanced liabilities.

On a horse race, they might offer prices that give up an over-round of 110%-130% most days, but on Premier League Football which has a much, much higher turnover, and WoM is far more predictable (due to team loyalties coming into play), over-rounds can be as low as 102%.

These markets are therefore more likely to provide value to the authors' methods, however I note they are making the majority of their bets between 5 and 1 hours before a game, and therefore to some extent are able to factor in team selection and some of their results might be the result of a market inefficiency: team line-ups aren't announced until 60 minutes before kick-off.

Now, here's two major downsides:

1. Bookmakers will eventually end up closing their accounts, because winners are never welcome long-term.

2. There is a reason why successful gamblers don't publish. Even Thorpe who invented card counting and blackjack basic strategy realised publishing (which was his academic need) ultimately caused him to need to do something else: he ended up privately and quietly running a hedge fund.

These guys are probably finished within 2 weeks of this paper going around, and what's more because now a whole ton of people will go to implement this method, bookmakers will adapt and simply move from tissue to industry average as quickly as possible, whilst limiting even more players to reduce liability exposure (as has been the style in recent years).

EDIT: I only skim-read the paper when I wrote the above. Now I've read it a little more closely I am even more convinced there is nothing to note here, and also, their accounts have already been limited or closed.

I am confused. If the betting is rigged, then how were the authors able to beat it consistently? What did I miss?