- in Betting by Lucy
- |
- 25 comments
Get A Free Betting System… From The BBC?!
It has often struck me that most of the systems that we test here at Lay Back have been developed by one or two people as a maximum.
Indeed, I’d go so far as to say there is something touchingly amateurish about the vast majority of betting systems providers.
Many services really just start out as one punter’s own method, and often only come to market when the penny subsequently drops that further profit can be derived by promoting the concept to others.
The obvious exception to this rule, of course, is Goal Profits which is streets ahead of most other offerings, but then Steve and Kevin have been consistently adding to it since 2011 – and it shows.
Inevitably, one wonders whether more expensively funded approaches might yield better results than all these amateurs.
All of which brings me to Sam….
Does Sam really know the score?
Sam is an acronym for Sports Analytics Machine.
Which in turn is a fancy name for a computer system designed by Salford University to crunch large amounts of data in order to make predictions.
Sam came to public attention when Mark Lawrenson, who (attempts to!) predict Premier League football scores for the BBC, was asked to compete against it on Boxing Day 2015.
Now normally, Mark competes against celebrities. Which always seems to be a bit of a cheat to me because many of their selections are biased towards their favourite teams or indeed reveal a startling lack of knowledge about football in general, and therefore offer no real competition.
Sam, however, was a different level of opponent altogether.
Salford University’s website explains: “Developed by Professor Ian McHale and Tarak Kharrat, the machine takes into account multiple factors to come up with its predictions, such as recent results, which team is at home, the strength of teams that both sides have played as well as the quality and the form of the players on the pitch.”
All of which sounds fine and dandy… and no doubt leaves you wondering how Sam performed in practice.
Well, in truth, not much better than Mr. Lawrenson.
In terms of win/lose/draw outcomes, Mark got four out of ten right whilst Sam scored five.
And as for exact scorelines, well, Sam predicted just two correctly – unlike Mark, who got just one right.
So yes, Sam did beat Mark but it was still hard to make a case that this represented a giant leap forward.
Now, I recognise that ten games are scarcely enough to constitute a robust trial but the results were underwhelming enough for me to put the whole business quite out of my mind.
Until now that is,
And that’s because the BBC has started publishing Sam’s predictions every single week this season.
So there is now a real chance for us to test Sam out with a robust amount of data.
The only piece of additional information that we will need is the price for each outcome which I propose to lift from the exchange where we are likely to obtain slightly better value than at the bookmakers.
My Trial
The next round of Premier League matches takes place over the weekend of 15 October so I will begin the trial then.
Twenty weekends of ten matches each will be enough to bring us to our minimum 200 bet target, at which point it should be reasonably clear whether Sam’s predictions, care of the BBC licence fee-payer, really present a free road to riches.
It’s quite a thought… wish me luck!
Update: 2nd December 2016
It was in October that I finally decided I had to test an idea that’s been bothering me for ages.
Why not trial the dear ol’ BBC’s very own betting tips?
Even as I ask this question, I can hear gasps of amazement and shock emanating from the Home Counties.
Could Auntie seriously descend to such a level?
Well yes, indeed she could. And, in fact, does, on every Premier League match.
Of course, nobody quite dares mention the grubby subject of betting: but tips are tips and BBC Sport issues them regularly.
Every week in fact, the Beeb publishes predictions for the upcoming weekend’s PL fixtures, as supplied by the Sports Analytics Machine (SAM) at Salford University, which just happens to be conveniently located around the corner from the BBC’s Media City headquarters.
One can almost imagine John Motson popping round to the bookies with a couple of savvy maths students.
But of course, the real question is.. is SAM any good at picking winners?
Well I have to say, things are looking reasonably rosy so far.
But before I jump into the detail, I ought first to explain how the process works.
Each week, the BBC writes an article on every PL match to be played over the forthcoming weekend.
And, at the very end, they print the probabilities assigned, by the SAM computer, to each outcome in the Match Odds market. There is also a note of the scoreline that SAM thinks is most likely.
Here, for instance, is the information given for this Saturday’s big match between Manchester City and Chelsea:
I then do two things… firstly, the simple bit.
I place a single point back bet on the suggested Correct Score.
So, as I write, the 1-1 result is priced at 7.8 and I’ve backed that.
That’s me done for the Correct Score bit of my trial. Here it gets a bit more involved however.
Because now I come up with a Match Odds bet by comparing SAM’s probabilities for the three possible results with the prices at Betfair.
Of course, it helps to know that you can quickly convert percentage probabilities into decimal odds by dividing them into 1.
So.. taking the City v Chelsea game as an example, SAM believes there is a 26% chance of an away win.
Well, 1 divided by 0.26 works out at 3.85. Which means that SAM thinks a Chelsea win is slightly less likely to happen than the Betfair punters who priced it at 3.65.
SAM also believes City should be priced at 2.0 (as 1 divided by 0.50 equals 2.0) and the draw should be 4.17, because that’s what you get when you divide 1 by 0.24.
Now, if SAM is right, there are some value opportunities here: as Betfair’s odds are rather different from what the Salford super-computer has worked out for us.
But crucially, SAM doesn’t tell us which value opportunity to bet on. So this is where personal discretion comes in.
OK, so what I’m going to do here is go for a home win for City, who are priced at 2.2 on the exchange.
Because if SAM is right, I’m getting a 50% opportunity for the price of a 45% chance — which sounds like a good mix of value and likelihood to me.
So, to sum up:
- I’ve simply followed SAM’s instructions in the Correct Score market and bet on whatever scoreline it thinks most likely.
- I’ve then used the probability data it generates to work out where the value opportunities are in the Match Odds market.
It follows that my Correct Score bets are a case of ‘doing what I’m told’ whereas the Match Odds selections involve a degree of judgement.
Performance to date
Correct Score Bets
Following a very slow start, with 18 losing bets out of 18 over the first two weekends, matters have turned around impressively to leave us 7 points to the good overall.
The ROI is also a very satisfactory 15.4%, though the (inevitably low) strike rate does guarantee some serious ups and downs.
I am betting with a single point on every selection and allowing for 5% commission:
Match Odds Bets
Unfortunately, the Match Odds bets haven’t gone anywhere nearly so well.
Once again, I’m allowing for 5% winner’s commission but I am betting 2.5 points on each selection.
The reason why is that, at least when I started out, I thought the Match bets were more promising than the Correct Score tips:
Of course, if you were to recalculate my progress with the Match Odds bets at single point stakes, you would find I am only 3.35 points down overall. Which is hardly a disaster.
You can download my complete results log by clicking here.
My Trial So Far
I have mixed feelings about the trial so far.
The Match Odds selections have fared poorly even though they are based entirely on value betting principles.
Yet the much more encouraging Correct Score tips have done well in spite of the fact that they usually just involve backing the score with the shortest odds at Betfair.
Maybe the moral of the tale, therefore, is to leave discretion to one side and simply do what SAM tells you!
Still, I’ve not put on fifty bets yet with either method so it’s too early to jump to conclusions.
I’ll be back in a couple of months with my next report.
Update: 10th March 2017
Who doesn’t find the idea of discovering a completely free betting system irresistible?
And perhaps even more irresistible is the idea that such a system might be found in precisely the sort of place where you might expect to be looked down on for being interested in sports betting in the first place…
You see, I’m thinking of good old Auntie, traditionally the sobersides protector of the nation’s morality.
Surely it can’t be?!
Indeed, dear reader, it can.
For although the Beeb don’t explicitly state that their season-long experiment with the football predictions made by Salford University’s supercomputer, SAM, is to be used for betting purposes, there seems little other potential benefit from their service.
And so, having kicked the idea around for ages, I decided, back in October, I could no longer resist the idea of testing SAM’s forecasting skills.
So.. eleven weeks later (I’ve had a few weeks off for good behaviour or it would have been more), am I making any money yet?
Read on to find out…
Well, so far the answer is yes, a very small amount.
For every Premier League match that I’ve looked at, I’ve considered two bets:
- A bet on the Correct Score, as predicted by SAM
- A bet on the Match Odds market, depending on where SAM indicates that there is value to be had.
I always put on a Correct Score bet and I usually (but not always) put on a Match Odds bet: the exception being where the odds implied by SAM’s predictions are very similar to the live market prices, in which case I stick to Correct Score betting only.
It follows that I’ve put on slightly more Correct Score bets than I have Match Odds ones: 102 against 86. Which means that, overall, I’ve placed 188 wagers in total – for a combined profit of just over 3 points:
Overall Performance | 5% commission deducted from winning bets |
Total Profit | 3.36 |
Total No. of Bets | 188 |
ROI | 1.1% |
Strike Rate | 25.0% |
There has been little difference in the outcome of Match Betting and Correct Score wagers. However, as the latter inevitably has produced a much lower strike rate, its performance has been rather more volatile.
In practice, the two methods produce similar-looking performance graphs, which suggests there would be little point in following them both:
The complete results log is available for review here.
My Trial So Far
I’m not exactly blown away by Salford University’s supercomputer.
Results to date suggest that it’s hardly any better at predicting outcomes than the market itself.
Indeed, most of the time, the scoreline that the service selects as the most likely is simply the one with the shortest price at Betfair. Which does rather make you wonder what the point of all that expensive technology is.
Still, maybe as the season’s end approaches, there will be a pleasant twist to this tale. Let’s hope so anyway.
I’ll be back once the English domestic season concludes in May with my trial conclusion.
Conclusion: 21st May 2017
Last October, I finally kicked off a trial of an idea I’d been toying with for ages.
In short, I was wondering whether the BBC’s weekly football predictions could make me some serious money.
Now, before you throw up your hands in horror, let me point out that I was not considering following Mark Lawrenson’s regular selections column.
No, I was hoping to benefit from the Beeb’s generous habit of publishing Premier League correct score predictions that are generated each week by Salford University’s expert system.
The capabilities of this august supercomputer are written up in gushing prose on the university’s own website. So I thought that maybe herein lay opportunity.
Matters got off to a flying start over the first few weeks, only to be followed by a setback in the middle period of the trial.
Still, I remained in credit overall by the time I wrote my second progress report… So did we finish in the black?
Read on to find out…
Sadly, after 238 bets, I’ve finished well in the red.
The last four weeks were dire leaving me with a total loss for the trial of over 35 points.
The summary is as follows:
Overall Performance | 5% commission deducted from winning bets |
Total Profit | -36.5 |
Total No. of Bets | 238 |
ROI | -10.27% |
Strike Rate | 22.69% |
The trial was really two trials in one.
This is because I normally put two wagers on each match:
- A back bet on the final exact scoreline;
- A punt on whichever match outcome represented best value in the light of SAM’s pre-match assessment of the game.
Now, I always put on a Correct Score bet as SAM made a precise prediction of the most likely final score of every Premier League match.
However, I didn’t always put on a Match Odds bet as well. This is because SAM doesn’t predict exact result outcomes (such as, ‘the match will produce a win for the home team’) but instead calculates the probabilities, in percentage format, of it ending as a home win, an away win or a draw.
So, whenever SAM came up with probabilities for the three match outcomes that were very similar to the Betfair prices, I just didn’t bother with a Match Odds bet. But when it was clear that SAM was highlighting that Betfair offered value on a particular result, I bet on that outcome.
As a result, I have placed more Correct Score bets (131) than Match Odds ones (107).
Unfortunately, neither the Correct Score nor the Match Odds bets were successful.
The Correct Score bets were particularly dispiriting as the strike rate was so low:
Overall Performance | 5% commission deducted from winning bets |
Total Profit | -20.53 |
Total No. of Bets | 131 |
ROI | -15.7% |
Strike Rate | 11.5% |
Match Odds betting produced a better strike rate but a similar net loss:
Overall Performance | 5% commission deducted from winning bets |
Total Profit | -15.97 |
Total No. of Bets | 107 |
ROI | -7.1% |
Strike Rate | 36.4% |
My Trial Conclusion
I’ve put quite a bit of effort into this trial but really have very little to show for it at the end.
I was hoping for great things from Salford’s supercomputer but it proved no better than any of the many run-of-the-mill tipsters I’ve tested on my site.
I also noticed that, very often, the Correct Score predictions seemed to be little more than announcements that the scoreline with the lowest price on Betfair would most likely be the outcome.
Did we really need to fire up a supercomputer to learn that?
So, as you can tell, right now, I’m feeling underwhelmed by technology.
Anyway, the full results log, if you’d like to investigate further, is available here. And that’s it from me.
AI may be the future, but for now you’ll be much better off learning the skill of trading over at Goal Profits where a couple of seasoned professionals will do a better job of pointing you in the right direction.