Thursday, 7 August 2014

FTL Attractions And Identifying Poison

Although several of the less followed leagues have already started, the 2014-15 party begins for real this weekend with the Football League and Conference back, along with the top league in France.

Twenty entries now make up the FTL and the selections should start flooding in any minute. A reminder that selections for matches played this weekend (i.e. Friday through Monday) should be in by midnight on Friday, with any Friday selections of course needing to be in before kick-off.

The twenty entries will now be competing for £1,400 in prizes, plus bonuses for finishing ahead of any of The Football Analyst, Football Investor, Skeeve or Football Elite. The Erskine Cup will also add some excitement (in my opinion) as the season unfolds.

The first Cassini Service email has also been sent out, with some selections this weekend, although there will no draws featuring promoted sides until the new year, and nor will there any longer be any ridiculously short priced draws from France and Italy. I will still be including these matches because they can usually be backed at higher prices, but backing these matches at Pinnacle Sports prices is just not viable. Backing at shorter than 3.15 in Ligue 1 cost us (the XX Draws) 13.78 points, while 12.28 points were lost backing at under 3.21 in Serie A.

The Draw price can vary by a lot - just this week the Draw price from the Ligue 2 game between Brest and Clermont Foot was recorded by Football Data at a best of 3.3, with an average of 3.06 (also Pinnacle's price). The low I saw was Ladbroke's 2.85. It is stating the obvious that there is a big difference between 2.85 and 3.3 - 15.8% value.

Even though my recorded prices are Pinnacle's, and they are chosen because betting with them is sustainable, it is worth mentioning again that their price can usually be beaten. The last time I checked, in the EPL of 2012-13, Pinnacle Sports had the top draw price in just 73 of 360 matches, although that 73 was higher than any other single book. I'll probably take a look at 2013-14 in the next few days.

Matches involving newly promoted teams in the first half of 2013-14 cost us 26.16 points before being profitable in the second half.

Football Elite's Matt sent out a very interesting preview of the Football League and some non-leagues too yesterday, and made some good points about the impact of the close season. I had a question on the Cassini Service email which was how am I making selections already when teams have yet to play. It's a fair question, but my thoughts are rather like matt's on this, and as Matt wrote:
And of course you also need to ensure that there’s no wholesale changes to the club otherwise last seasons performance is totally irrelevant. All clubs have some change, don’t be put off by individual player changes or even in some cases manager changes. Just as long as there’s not been a lot of changes to key men.
Southampton is one example of a team whose rating from the end of last season is basically redundant, but for most teams, as Matt also pointed out, the close season is relatively short. Three months, and not too much usually happens - it's not like expecting World Cup form to roll over from 2014 to 2018 - which would not be a good thing for England.

Peter Nordsted has entered his Drawmaster in the FTL, expected to debut selections next weekend, and
also sent out an email which included some of his services for next season, which includes "Pete Nordsted's Weekly Football Ratings and Analysis sheet".
An example of the Arsenal v Crystal Palace game can be found HERE
This match is of particular interest to both Peter (an Arsenal fan) and myself (not an Arsenal fan, although my Granddad was), and I was interested in comparing our numbers on this. Arsenal are a club who have made a number of moves this summer, and are likely (in my opinion) to be underrated by my ratings and thus suspended from any systems until their rating adjusts. Crystal Palace have made fewer moves and will struggle to beat last season's 11th place finish (in my opinion), but let's look at the numbers.

Peter has Arsenal expected to score 1.4 goals, Palace 0.31 while I have Arsenal scoring 2.03 and Palace 0.99.  A 2:1 home win if ever I saw one, although the market favours 2:0.

These numbers make the Over 2.5 goals 1.97 in my opinion, 4.11 in Peter's. Betfair's price has traded as low as 1.11(?) and as high as 1.89.

I also think (and I say this as a numbers man, rather than as a Palace fan) Peter's 0.31 goals expectation  is way too low. The market thinks so too. 1.32 for Under 2.5 goals is rather unusual (to say the least) in the EPL. The Crystal Palace v Aston Villa game was the shortest priced Under last season, at an average of 1.61.

In the 1,714 matches since I started recording goal expectations and actual goals, I only had two teams with an expectation as low as 0.3. Both did indeed fail to score, but a tiny sample size means little. Looking at other low expectancies, there were 85 teams expected to score 0.5 goals or less, but half (well, 42) managed to score, and 17 managed to score more than once.

It could be coincidence, but it's a promising sign that the closest expected to actuals is in the range where I have the highest number of predictions.
Teams        :          Variance        :         Actual            :         Expected
It's a good way to validate that the model is working close to 'expectations'; after all, the output is only useful if the data being entered is accurate. Poisson can quickly become poison.
Fortunately the markets are there to help - if our numbers are too far off the market's, well chances are we've messed up somewhere.  If it looks too good to be true, it probably is. 

No comments: