Matched betting - who's done it and who's good at it? (No Referrals)

Associate
Joined
21 Nov 2007
Posts
1,064
Location
Fleet, Hampshire
OK. I understand the maths you are portraying now and accept that in your example you are making the point that it is not worth laying the bet. However you are rather loading the dice by assuming that Paddy's customers are a bit thick and BetFair customer paid attention in Probability lessons. The overround is unrealistic in your instance as the probability of each outcome is actually known and equal. More realistic is that both outcomes would have odds of 1.9 to provide an overround. This is the case with other scenarios where the odds are testable and repeatable such as Roulette.

If you assume that everyone missed probability 101 and that the lay bet can be placed at 1.95 (a more realistic match of back to lay odds I would argue) then the argument does not hold. In that case your liability is 9.25 and stake is 9.74

Scenario 1 - unlaid bet
Unlaid bet EV - (£9 * 0.5) + (-£10 * 0.5) = -£0.5

Scenario 2 - laid bet
Laid bet backing heads and heads comes up - you win your back bet +9, lose the lay bet liability -9.25 - net loss 0.25
Laid bet backing heads and tails comes up - you lose your back bet -10 and win the lay bet stake +9.74 - net loss 0.26

You don't combine those scenarios. The expected outcomes of scenario 1 are included in scenario 2.

Which has been my point all along. If you can get a match of back to lay below the overround then there is more long term value in matching. If you cannot get good matches your higher long term reward is in gambling.
 
Caporegime
Joined
29 Jan 2008
Posts
58,898
OK. I understand the maths you are portraying now and accept that in your example you are making the point that it is not worth laying the bet.

That isn't the point, the point is that laying has a cost associated with it, and also to illustrate how the QL is the EV of both bets combined. I'm not saying it isn't worth laying, laying has a purpose and that is to reduce variance. The point being argued is that you do get greater value in theory in the long run by not laying.

However you are rather loading the dice by assuming that Paddy's customers are a bit thick and BetFair customer paid attention in Probability lessons. The overround is unrealistic in your instance as the probability of each outcome is actually known and equal. More realistic is that both outcomes would have odds of 1.9 to provide an overround. This is the case with other scenarios where the odds are testable and repeatable such as Roulette.

The same can be shown with roulette, it is easier to demonstrate it with a coin flip. I think you're forgetting the previous example - I've provided two examples of the coin flip, the first one did indeed have the same odds for both heads and tails if you look back. There is an overround in both cases, the point here in this second example was to demonstrate why comparing the vig and the QL, in cases where the QL is lower, is potentially rather flawed. The reality is that the EV on the bookie leg is the same regardless, you get a lower QL when that particular bet at the bookie is priced rather more generously than the vig would suggest... if it is particularly generous then you get an arb even.

If you assume that everyone missed probability 101 and that the lay bet can be placed at 1.95 (a more realistic match of back to lay odds I would argue) then the argument does not hold. In that case your liability is 9.25 and stake is 9.74

Scenario 1 - unlaid bet
Unlaid bet EV - (£9 * 0.5) + (-£10 * 0.5) = -£0.5

Scenario 2 - laid bet
Laid bet backing heads and heads comes up - you win your back bet +9, lose the lay bet liability -9.25 - net loss 0.25
Laid bet backing heads and tails comes up - you lose your back bet -10 and win the lay bet stake +9.74 - net loss 0.26

You don't combine those scenarios. The expected outcomes of scenario 1 are included in scenario 2.

Which has been my point all along. If you can get a match of back to lay below the overround then there is more long term value in matching. If you cannot get good matches your higher long term reward is in gambling.

What you've done there is to make money from betfair, note you've omitted the EV calculation of the lay leg! if you can do that then you don't need matched betting in the first place!!!! I have made this point a few times and put it in bold at the end of the big post:

The only way this is not true is if you've got no cost at all from laying or indeed if your lays have a positive expected value... if they had a positive expected value then you have no need for matched betting and have found a system to make money from betfair.

That is exactly what you've done in the above example, you've assumed a lay leg where you're getting a positive EV.

Our EV of the bet at the bookie is as before (£9 * 0.5) + (-£10 * 0.5) = -£0.5

This time we lay £9.74 at the exchange at 1.95 for a liability of £9.25

Our EV from laying at the exchange is (£9.74 * 0.5) +(-£9.25 * 0.5) = +£0.245

You have a positive EV there from the exchange! If you were to just lay then you'd expect to make on average £0.245 per coin flip.

As I commented before, I'd happily lay 1.95 on a coin flip all day... forget matched betting you're just making money from betfair there.
 
Last edited:
Associate
Joined
21 Nov 2007
Posts
1,064
Location
Fleet, Hampshire
As I commented before, I'd happily lay 1.95 on a coin flip all day... forget matched betting you're just making money from betfair there.

Of course you would. I was trying to use your example to make the point.

Where you keep falling down is you are doing the EV calculations for both components of a matched bet in isolation. The maths is flawed. You cannot win both bets, and you cannot lose both bets. Your maths assumes you can.
 
Caporegime
Joined
29 Jan 2008
Posts
58,898
Of course you would. I was trying to use your example to make the point.

But you've not made the point there - you've just picked a scenario where you're making money from betfair. If you can do that then you'd not need to bother with matched betting in the first place.

Where you keep falling down is you are doing the EV calculations for both components of a matched bet in isolation. The maths is flawed. You cannot win both bets, and you cannot lose both bets. Your maths assumes you can.

There is no assumption in the calculations that you can win both bets, I've just presented straight forward EV calculations for each leg. We can consider what happens if we were to only bet on one of the legs and we can see that if we bet on both we essentially lock in the EV of both legs, that is what the above was demonstrating. I put in the second scenario as you have maintained a flawed comparison between the overround and the QL - this is especially misleading in the particular scenario you're looking at (low QL) as what you're most likely seeing there is a particular line at a bookie that is offering you a less bad or even a value bet - thus the second example with the skewed paddy power odds.

I'm not sure what part of the maths you think is flawed - are you disputing that someone betting £10 on heads at 1.90 will on average lose 50p per coin toss? This is trivial to simulate if you're unsure.

this might be useful background if you're questioning the ev calculations
https://en.wikipedia.org/wiki/Expected_value
 
Associate
Joined
21 Nov 2007
Posts
1,064
Location
Fleet, Hampshire
There is an assumption that you can win both because you are not linking the calculations. In the school parlance they are simultaneous equations.

You used this example
Our EV of the bet at the bookie is as before (£9 * 0.5) + (-£10 * 0.5) = -£0.5

Our EV from laying at the exchange is (£9.74 * 0.5) +(-£9.25 * 0.5) = +£0.245

There is nothing wrong in these calculations in isolation. However they fail to reflect the dependency. Both calculations average out two possible outcomes. The problem is that the outcome in 1 is the reverse of the outcome in 2 by definition. If you call the first part of each equation a and the second part b. You cannot have a and a occuring. You have to have an a and a b.
 
Caporegime
Joined
29 Jan 2008
Posts
58,898
There is an assumption that you can win both because you are not linking the calculations. In the school parlance they are simultaneous equations.

There is nothing wrong in these calculations in isolation. However they fail to reflect the dependency. Both calculations average out two possible outcomes. The problem is that the outcome in 1 is the reverse of the outcome in 2 by definition. If you call the first part of each equation a and the second part b. You cannot have a and a occuring. You have to have an a and a b.

There is no assumption to that effect there nor is there any claim to that effect. The calculations there are simply the EV of each leg. Though please note, the EV of either leg doesn't magically change or vanish just because you've made other bets.

Though you can see that the QL is equivalent to the ev combined.

For example, in the previous scenario, we can see the ev for the legs individually

Lets say we bet at the bookie:

Our EV of the bet at the bookie is (£9 * 0.5) + (-£10 * 0.5) = -£0.5

Lets say we layed at the exchange:

Our EV from laying at the exchange is (£9.27 * 0.5) +(-£9.73 * 0.5) = -£0.23

Hopefully that isn't too controversial. Note here that the combined ev of both sums to -£0.73

Now, so as not to confuse things, lets ignore the ev and just look at what happens if we were to back and lay together:

Like you said you can only have either heads or tails - that's fine:

If heads win then we have won £9 from the bookie and lost £9.73 at the exchange, the loss is £0.73
If tails win the we have lost £10 at the bookie and have won £9.27 at the exchange, the loss is £0.73

This is identical to the combined expected value, we've essentially locked this in, this is still true if you want to rig things on the lay leg so you're making money from betfair in that case you have a positive ev on the lay leg that somewhat offsets the negative ev on the at the bookies. Now I'd assumed this was intuitive but I think this is part of what you're getting a bit hung up on, there isn't any other "value" appearing from somewhere or disappearing anywhere just because you've matched/made other bets, all you've done by backing and laying is to lock in the value of the bets combined, you've essentially removed any variance, so whereas the guy just betting heads might have a series of wins of £9 and losses of £10 averaging out to a loss of -£0.5 per toss (and ditto to someone only laying and averaging out to -£0.23) when you back and lay together you've locked in exactly £0.73 (the individual legs are still winning/losing say £9, £10 etc.. just as if you only backed those legs) and if you were to repeatedly make that bet and look back at the back leg or the lay leg leg in isolation then in the long run they'd be giving you an average loss equivalent to their respective ev.
 
Last edited:
Associate
Joined
21 Nov 2007
Posts
1,064
Location
Fleet, Hampshire
There is no assumption to that effect there nor is there any claim to that effect. The calculations there are simply the EV of each leg. Though please note, the EV of either leg doesn't magically change or vanish just because you've made other bets.

Yes it does. EV includes probability within the formula. See your own Wikipedia link. If you have backed both sides of an outcome the probability of the combined enterprise is 1. Balanced correctly it makes no difference to the overall financial outcome if you win or lose the back bet as by definition the reserve outcome will occur on the lay to cancel it out.

If heads win then we have won £9 from the bookie and lost £9.73 at the exchange, the loss is £0.73
If tails win the we have lost £10 at the bookie and have won £9.27 at the exchange, the loss is £0.73

I conceded earlier that in this scenario as the gap between back and lay odd was greater than the overround it was not worth laying it.

I countered that if the reverse was true (lay odds of 1.95) you reduced the outcome to 0.27

If heads win then we have won £9 from the bookie and lost £9.27 at the exchange, the loss is £0.27
If tails win the we have lost £10 at the bookie and have won £9.73 at the exchange, the loss is £0.27

Please don't counter again that it would not be worth backing this, these are unrealistic odds in both cases. It proves the point though that in the case of close back and lay odds the outcome will be more profitable than the long term EV of gambling alone.
 
Caporegime
Joined
29 Jan 2008
Posts
58,898
Yes it does. EV includes probability within the formula. See your own Wikipedia link. If you have backed both sides of an outcome the probability of the combined enterprise is 1. Balanced correctly it makes no difference to the overall financial outcome if you win or lose the back bet as by definition the reserve outcome will occur on the lay to cancel it out.

The only probability contained in the EV calculation is the probability of flipping a head or a tail. That is unchanged, if you think otherwise then please do explain how it has changed?

Please don't counter again that it would not be worth backing this, these are unrealistic odds in both cases. It proves the point though that in the case of close back and lay odds the outcome will be more profitable than the long term EV of gambling alone.

No it doesn't. For a start in that scenario you've presented it is clearly more profitable in the long run to just lay at betfair - you'll have a positive ev there vs your guaranteed negative loss.

Though realistically if you're getting a +ev on a leg then it is coming from the bookies, the positive +ev benefits both the matched bettor and the person gambling, the ev doesn't change on the bookie leg just because you've laid (this is the bit you're still hung up on).

The whole point of the second scenario is you were making a flawed comparison between the overround (or really should be the vig) and the QL, if you're cherry picking matches where the QL is lower than the vig then what you're in effect doing is using a matched betting tool to find scenarios where some bookie's line is rather more generous for a particular bet than the vig is suggesting... in some cases you'll even be able to arb between a bookie and the exchange if the bookie is particularly slow. Again someone just betting at the bookie on that same bet is still getting that +ev too.
 
Associate
Joined
21 Nov 2007
Posts
1,064
Location
Fleet, Hampshire
The only probability contained in the EV calculation is the probability of flipping a head or a tail. That is unchanged, if you think otherwise then please do explain how it has changed?

The probability is changed because you are backing both outcomes. The formula is quite simple the probability of the lay bet winning is 1 minus the probability of the back bet winning. Not only that but we are not actually dealing in probability here we are dealing in certainty. If you back heads on your back bet you are backing tails on your lay bet. One of these is going to happen. The sole purpose of the lay bet is to take the risk out of the back bet.
 
Caporegime
Joined
29 Jan 2008
Posts
58,898
The probability is changed because you are backing both outcomes. The formula is quite simple the probability of the lay bet winning is 1 minus the probability of the back bet winning. Not only that but we are not actually dealing in probability here we are dealing in certainty. If you back heads on your back bet you are backing tails on your lay bet. One of these is going to happen. The sole purpose of the lay bet is to take the risk out of the back bet.

You're getting completely muddled now - if I flip a fair coin I have a 0.5 chance of heads and a 0.5 chance of tails. If I bet on heads there is still a 0.5 chance of heads and a 0.5 chance of tails.... if I bet on heads and lay heads... there is still a 0.5 chance of heads and a 0.5 chance of tails. The fact you've bet on a coin flip whether you've covered both possibilities or not, has no effect on the probability of that coin flip coming up heads or tails. The probabilities in the EV calculation are simply the probability of getting heads and the probability getting of tails. These probabilities are the same regardless of what you've bet on - your bets do not change the coin flip probabilities in the EV calculation!

Would it help if we write out the combined EV as well - I had assumed this was needless

Our EV of the bet at the bookie is (£9 * 0.5) + (-£10 * 0.5) = -£0.5

Our EV from laying at the exchange is (£9.27 * 0.5) +(-£9.73 * 0.5) = -£0.23

Our "combined EV" is ((£9 -£9.73 )* 0.5) + ((-£10 + £9.27) * 0.5) = -£0.73

This is the QL and is just the sum of the EV of the individual legs! The probabilities are unchanged but if you want the two legs combined then that's fine.

I think this is where you're still getting a bit of a block, you're getting hung up on the idea of being able to look at both legs.
Perhaps it would be clearer to write out say a series of coin flips, obviously in the long run we expect an even number of heads and tails so for the purpose of illustration this short sample will have equal numbers of each, though of course in reality it is random and the bettor who is not laying (or indeed layer who is not betting) is subject to plenty of variance - so just to be clear this is not a claim that you'd necessarily get an even number of heads and tails over a small sample of 10 flips, it is just a breakdown for the purpose of illustration of what happens in a series of flips in the long run... and to again demonstrate the link here between your realised/locked in loss and the combined EV.

say we get

HTHHTHTTHT

That's 10 coin flips, using the previous odds if we were backing and laying heads that would be a loss of 10* the QL of £0.73 = £7.30

We can still look at the individual legs though - what happened on the bookie leg:

HTHHTHTTHT
+£9 -£10 +£9 +£9 -£10 +£9 -£10 -£10 +£9 -£10 = -£5 and of course this is as per the ev, 10 * -£0.5 = -£5

What happened on the lay leg:

HTHHTHTTHT
-£9.73 +£9.27 -£9.73 -£9.73 +£9.27 -£9.73 +£9.27 +£9.27 -£9.73 +£9.27 = -£2.30 as per the ev, 10* -£0.23 = -£2.30

the QL is the same as the combined ev of both legs

------------------------------------------------------------------------------------------------------------------------------------------------------------------------
(If you like we can instead can insert your +ev betfair scenario for the lay leg instead if you like, same situation there)

The QL in that case varies slightly by a penny due to rounding... so the loss from backing both is 5 * -£0.25 + 5 * -£0.26 = £2.55

What happened on the lay leg with those coin flips:

HTHHTHTTHT
-£9.25 +£9.74 -£9.25 -£9.25 +£9.74 -£9.25 +£9.74 +£9.74 -£9.25 +£9.74 = £2.45 this is again as per the ev 10* £0.245 = £2.45

and of course combined ev over the 10 flips -£5 (from the bookie) and + £2.45 from the betfair leg = -£2.55 which is the QL

I don't know if the above helps at all but a coin flip is perhaps the simplest example to use here and the above is breaking it down further.
 
Last edited:
Caporegime
Joined
29 Jan 2008
Posts
58,898
:D

Hey, it is civilised and relevant to the thread :)

It's not like it is disrupting much else aside from posts along the lines of "made £XXX so far" and "has anyone got an FAQ/links/guide?" etc...

anyway off out for the evening, I'll perhaps pick this up tomorrow...
 
Caporegime
Joined
24 Oct 2012
Posts
25,020
Location
Godalming
:D

Hey, it is civilised and relevant to the thread :)

It's not like it is disrupting much else aside from posts along the lines of "made £XXX so far" and "has anyone got an FAQ/links/guide?" etc...

anyway off out for the evening, I'll perhaps pick this up tomorrow...


I know, it's almost frustrating to see you two being so civilized. Can't you glam it up a bit?
 
Caporegime
Joined
29 Jan 2008
Posts
58,898
@dowie in your example, the second one with the closer odds. Does your worked example not show that with a close match you have mitigated the loss over a series of bets by laying them? Which is my point all along.

I'll stick the really boring part of this in spoilers so as not to annoy @Diddums further (he's a bit tetchy at times!):

Sorry, which are you talking about? I've given two examples using coins flips in post #4999 but you also suggested one with a +EV on the exchange leg.

ref the two worked examples I gave:

I kept the overground and vig roughly equal in both - the second one is deliberately structured to show a bookie biasing things a bit (more on this below) - this is how you can get a less bad bet or even a value bet and therefore a closer match/smaller QL, when the bookie has priced things for a particular outcome more favourably than the overall vig would suggest. The main thing to highlight there was that the vig was approximately the same in both cases and isn't necessarily a good thing to compare the QL with - in the case of a closer match clearly the person only betting at the bookie is also getting the benefit of that better ev on the bookie leg.

ref the change you suggested:

Now onto the +ev on the exchange leg that you suggested, well this does indeed give you a closer match but it isn't really a reflection of what happens in matched betting - it is the wrong leg to adjust if you're trying to illustrate a closer match with an example! The exchanges are generally highly efficient, in fact if you can get a +ev on your lays in the long run then you don't need matched betting, you can just go make six figures a year from betfair as I've mentioned!

The main point was the poster asking about sometimes not laying and whether or not this offered better value - technically it does in the long run. In fact if you never layed you'd have an even better ev however your returns would bounce around all over the place and you'd need deep pockets!

The simple reason is laying has a cost, it involves crossing the spread and paying a commission if applicable, it is an inherently -ev thing to do but it is useful to reduce variance/swings in pnl. If the ev of the bets at the bookie are the same then adding a -ev series of bets is just lowering your overall expected value.

---------------------------------------------------------------------------

Now more matched betting commentary not so much about tedious ev coin flip examples etc....


Just a comment on what is actually happening when you get a close match (or even an arb) - this is inevitably coming from the bookie's leg. In theory a traditional bookies would just try to maintain their overround, they adjust their odds in response to the money flowing in on various selections and try to make sure that no matter what the outcome they earn their vig (this is the reason for the second scenario I constructed where I asserted the people in the town Paddy lives in prefer tails - thus paddy skews his odds to reflect that he gets more action on tails). Now these days bookies will try to pay a bit more attention to exchanges/wider market and will be better able to offset/hedge some of their exposure if they're getting a large amount of bets on say one side of a football game etc.. rather than say a local Liverpool bookie having to offer worse odds on Liverpool and better priced odds on Man U however their prices are not going to be as efficient as the exchanges - as a result of this you'll get the scenarios you're talking about with a closer match (which benefit both the person betting only at the bookie and the person matching) - in fact sometimes the legs will overlap and you'll get an arb - a guaranteed profit! This is great right? Not necessarily...

You're going to increase your chances of getting gubbed by the bookies if you take value from them (they don't know whether you layed or not they just know you made a value bet), if you've got to the point where you're getting a match with no QL (even despite you crossing the spread at the exchange) then you're basically getting a slightly +ev bet from the bookie. Beyond this you're getting an arb and you're almost certainly going to get gubbed quite quickly. The bookies aren't stupid, they do have a look at which customers ended up betting on certain outcomes etc...

See for example the experiences of people who've tried it over a large sample, I did a quick google search and sure enough note the observations from this guy from years ago:

http://community.betfair.com/general_betting/go/thread/view/94082/29234875/arbitrage

Done it for years.

1.)You need to open loads of accounts with lots of bookies.
2.)They will all limit you/shut your account pretty quickly.
3.) So you start opening accounts using friends/family, these also get shut down so you have to find more and more of em to use.
4.) If you use Betfair as the other half of the arb you will discover over time that the BF side of the bet will be losing long term and the Bookies bets will be winning over a period.(in the short run results will be all over the place).
5.) Then you realize that this is because Betfair is a more efficient mechanism for finding true odds and you see that it is more profitable to not arb anymore just do the bookie bets side and let them run.
6.) Unless of course you are paying the premium charge on here. In that case you want to book losing bets on your BF account so you arb as much as you can be bothered to try to avoid this appalling tax.

Given the above there is therefore a question of do you actually want to have really close QL bets if you're matched betting? Is the reduced loss but higher chance of getting gubbed worth risking vs the chance of getting further free bets if you were to make a more regular selection at the bookie? Perhaps overall better to just stick with popular bets rather than cherry picking particular close matches... but that is another debate altogether. :)
 
Last edited:
Back
Top Bottom