Monday, 19 September 2011

Bumpy Roads

Next time I notice an anomaly, perhaps I shouldn’t question it. A tres bon weekend in France, as the 10 value over selections produced no less than 7 winners, a profit of 7.29 points. The Match Odds selections didn’t fare quite so well with 'only' 5 winners from the 10 selections, but a profit of 1.74 points. Football Elite had two Recommended Bets here, Auxerre drew, but Toulouse came from 0-2 down to win 3-2 and a 90’ goal went in our favour for once.

I hope a few of you followed me in on the Overs, but a quick word on what 'value' means'. Take a look at Peter Webb's excellent recent post What does value look like? on this subject. Peter writes:

I’ll often make suggestions around value. In my football ratings I point to matches that will have above or below value median results. What I am trying to say is this looks value, that is value, this isn’t. But what does value look like?

One thing it is not, is a tip. I’ll liberally scatter the words ‘on average’ into many posts as that is what value looks like. All sports and most events show a great deal of variability and that means that it’s nigh on impossible to get things right all the time. Therefore when you see something from me, I’m not saying back or lay this item. I am saying there is value in that. To understand how value manifests itself over time take a look at the following chart (see below).

This chart represents a lay value strategy. I think I have an edge, in fact I am pretty certain I do based on my research, but this graph shows you how variable the results are. After nearly 500 events using £10 stakes I am up £400. But, as you can see, the road is bumpy and this is because value plays out over the very long term. In that mix you can get very long runs where it’s very positive or very negative, but over the long term you should see your bank produce higher lows during each draw-down. That’s how you know you are achieving value. Once you are certain, you start looking at how scalable the strategy is.

So when I make recommendations or give ratings, be sure to interpret it correctly. It’s easy when recommending something to look foolish, but that will only occur if you interpret the information incorrectly. What I am saying when I post up information is that, based upon my analysis, it looks value or look for value here. I know from experience that I can generally find it, but often the biggest hurdle to making it pay is commission and associated costs, that’s critically important as well. So don’t forget to focus on that as well!
It's an excellent post, which reinforces that the selections here are not tips. Value does not always win. It simply means that over time, you will come out ahead, although as Peter cautions, commission and expenses can make the difference between a profitable strategy and a losing one.

My results in England weren’t bad either, and again, it was the Under / Over selections here that outperformed the Match Odds with a profit of 9.88 points. Nine selections, nine winners. The Everton v Wigan Athletic game was the one match where I found no value, and the only Under bet was Aston Villa v Newcastle United, so again I benefited from plenty of goals. The ten Match Odds selections produced a profit of 3.66, much a boost to the season running total which is still overall down in the EPL. The totals for this week are in the screenshot below.
The highlight selection in Germany was the lay of Bayer Leverkusen at 1.46 who then proceeded to lose at home to Koln 1-4. My spreadsheet recommended a lay of Borussia Dortmund, and Football Elite had Hannover ’96, and with two late goals, we both came out ahead.

In Italy's Serie A, Football Elite found their third winner of the weekend as Atalanta beat Palermo. I had Atalanta priced at 1.96, the market at 2.44 so this was a huge value bet for me.

And finally, in my least favourite of the top five leagues, La Liga, my value lay of Barcelona didn’t quite pan out, as the hosts beat Osasuna by 8-0, but only a week before, Barcelona failed to win at 1.2 so a Lay Barca strategy is still ahead on the season.

My spreadsheet doesn’t handle extremes too well. Barcelona went off at around 1.11 I believe, and to come up with this as a price, I would have needed to estimate Barcelona’s goal scoring probability 100% higher and Osasuna's 50% lower to have been close to that. As a result, the value in games involving big favourites (Barcelona, Real Madrid, Manchester City, Manchester United, Chelsea in particular) are often off the scale, and result in Lay bets for these teams. Whether this proves to be profitable or not remains to be seen.

The XX Draw selections went 0 for 3 on Saturday, and 1 for 3 on Sunday, (at last a winner, and as these things tend to work out, it was at 4.1, the longest price of any selection this season). Peter Nordsted’s Drawmaster system found Aston Villa v Newcastle United, raising his ROI to 27.14% and taking the lead as mine heads south to 26.11%. P had a few comments about the XX Draws
Before Fulham's unlikely comeback to equalise against Man City and end the diabolical XX losing streak, there were 16 consecutive losers.

I was wondering how this streak ranks as far back as your draw records go.
This is the worst run since I started tracking the sequence of bets. I had a losing run of 8 back in Jan / Feb but these runs will happen.
If we assume average odds of 3.5, then the probability that any single match will NOT be a draw is about 0.71 (not factoring in your edge, which means that the probability is actually a bit less). The probability of 16 consecutive independent events with a probability of 0.71 is 0.00416997, or about 1 in 240.
It would be nice to have our winners evenly spaced, but betting isn't like that. That's why it is so important to understand what value means. I started the season with 7 wins from 10 selections, which was nice, but the losing run looks even worse as a result! The overall picture is pretty good. 7 wins from 27 selections doesn't sound like a great strike rate, but a level stakes 7.05 points profit, and an ROI of 26.11% tell the true story. It's also worth noticing that of the 16 losing selections, 5 were winners at the 80' point or later, and could have been traded for a decent profit.
Secondly, would you consider offering a paid service that would get subscribers your XX ratings as well as the match odds and over/under value ratings for each of the five major European leagues by Thursday or Friday of each week?

You kindly drip feed your loyal readers a bit each week, but without all of the selections, it is very difficult to fullow a system for a particular league and market category over the course of the season.
Of course I would consider offering a paid service. Consider the Pros: It's not something I am actively looking to do, but I do it for myself anyway, it takes up a lot of time, and if I could recoup some of hidden cost of that time investment, then why not?The Cons add up though, mainly with the added responsibility. It's one thing me losing my peanuts on my system, but it's something else dragging others down with me. There would also be pressure to get the results in and the selections out in a timely manner.
Finally, if you are doing ratings for match odds and goals for the five major European leagues, does that mean that you are placing 50-100 value bets per week?

That's a lot to monitor, especially if you are always trying to get slightly better odds on an exchange...
In a typical 'Round' of the big leagues, there are 49 matches, so potentially there are 98 value bets. This weekend there were a lot, 80 in total, but last weekend there were 'only' 56. These are selections where I think the edge is 10% or more. It is a lot to monitor, but it's a lot easier to monitor when things are going well than when you have a poor weekend!

Finally, Mark Iverson hosted another NFL chat-room yesterday for the Minnesota Vikings v Tampa bay Buccaneers game. This turned out to be an excellent with a big comeback from the Bucs who trailed 0-17 at HT after making just two First Downs. Minnesota (1.08 at this point) couldn't do a thing wrong, and the Bucs couldn't do a thing right, but in the second half, it was a different game as the Bucs rallied to win 24-20. I was all set to call it a day at half-time, and find another game to trade, but Mark talked me into seeing how the second half started, and hanging in was a good call. I know better than to think a game is over at 17-0, but what was unusual in this game was that until the second half, Tampa Bay hadn't shown any sign that they would be able to make a game of it, but the beauty of the NFL is that you just never know.


George said...

Very useful to see your season-to-date stats. Lots of blogs focus only on the latest round of matches. Signal/Noise is so high that even a decent edge (pos or neg) is not obvious unless you have a large number of bets to study (250 is my rule of thumb). In your season-to-date stats, assuming zero edge, your results are 82nd percentile. Can't make it into an academic journal article, but practically very very good. What do your 10/11 season stats look like? If you can support a 5% roi edge over 1500+ bets, you are sitting on a gold mine, my friend ...

George said...

At what point in time and over how many bookmakers do you select the odds that go into your records? I have a bit of a problem with FE on this issue. If the target is good replication of the tipped bets, then a tipster (I am not saying you are one) should limit himself/herself to 4-5 big-limit, big-payback bookies (perhaps review them annually). Otherwise, there is a bias there, which in reality wipes out a significant part of the perceived/reported edge. Or one can simply report different edges for different contexts e.g. early-market (low-limit) edge, pinnacle-only edge etc. By edge, I always mean roi%.

George said...

Last comment here. Why are you not looking at the asian handicap markets? As far as I am concerned, if there is liquidity there, I will dig for an edge. Asian handicaps is where the liquidity is. I think I saw something from you about the mismatch of incentives of teams on spread/handicap markets, which you don't like. Don't see how that's relevant. If your system finds value, then obviously it overcomes the incentives problem. If one cannot find value, then yes, the funny incentives could be used as a reason for not digging further, I guess.

George said...

Meant to say that signal/noise is low, not high, in my first comment.

Average Guy said...

Hi Cassini, doed your stake vary with the calculated value. If the velue % is higher is the stake higher ?