Friday 12 February 2010

Gold All Over


A couple more questions made their way across the wires to me today.

Anonymous said:

“Just wondered why you don’t put up the Scottish predictions in advance?”

A: I do sometimes. As I wrote on Tuesday “Right now I tend to just give the Premier League forecasts, since those are the games that have the highest profiles, and the most liquidity. When there's nothing at the top level, I sometimes dip down into the lower leagues, even as far down as Annan Athletic on occasion, and pick out a few that I think might be of interest.”

I’m not sure that too many people are interested beyond the Premier League to be honest, and when there are a whole slew of matches, it’s quite a lot of work to list them all. Agreed that on days like yesterday, there weren’t too many to list, but on Tuesday there were some 37 league games in England and Scotland, and on a Saturday there’s many more.

Dubaitrader had “a couple of questions, if you don't mind me asking”, not at all – go ahead...
Firstly, how do find Football Elite? Their website seems ok and fair ROI over the years.
A: I found them by entering http://www.football-elite.co.uk/ into my browser…

Seriously, Football Elite have boosted my balance in the three months that I have been with them, but not by a huge amount. Their Recommended Bets are actually slightly down, rather unluckily in my opinion as an unusually high number of games have finished drawn. Certainly a strategy of laying the team they are opposing would have reaped rich rewards, as would backing the Under 2.5 goals. Of 37 selections, 14 won, 4 lost, and 19 drew. Just 9 were Over 2.5.

The profit for me has come from their Short List selections, those that came close to ‘Recommended’ status, but fell short. These have fared well and are in profit for me. The record here is 48 selections, 20 winners, 10 losses and 18 draws. Here the Under – Over split is less remarkable at 28:20.

Matt’s selections are almost always at evens or better, and although three months is too short a period to really pass a valid judgment, I have no reason to doubt his figures over the past 3 or 4 years and I do believe that by being as selective as he is, he has found that all important edge.
Secondly, without giving your major secret away, what is your main basis for your ratings? Any info would be appreciated.
and tied in with this is a question from Costas
The ratings are from the ELO rating formula(right?),what i want to ask is if you keep the records and update them by yourself or there is an internet database or service that they provide the ratings. I think you should change the blog's name to gold all over...hahaha. Costas
A: My major secret is that I really don’t have a major secret. Basic Elo ratings are quite simple to maintain, and my original idea was to generate odds based on these, but for football the draw is problematic. I tried using the Draw No Bet markets but these markets are illiquid and are only found in a few matches so I gave that idea up.

As for generating my own odds, as some people do, and comparing them with the bookmakers or exchanges, I found that pricing the draw complicated things too much, so I gave up on that plan.

I then made some adjustments to the formula for recalculating Elo ratings after games. This is a manual, and time consuming, process using the computing power of Microsoft's Excel, and the calculations account for more than just whether the team won or lost.

A pleasant, and quite unplanned, side effect of this strategy is that I am now able to enter in the ratings of the two teams before a game, and have my spreadsheet tell me what result will keep the ratings closest to the status quo, i.e. my predicted / expected result.

While the strength of a team will fluctuate over time, it won’t (typically) change that much from game to game. The expectation is that teams will perform to their strength more often than not, but football is a low-scoring game involving those unpredictable animals called human beings, which ensures plenty of surprises. The trick is to not worry about these results (Dundee United v Hamilton last night for example) since they happen, but to look at the big picture. If a draw is predicted and someone wins 1-0, so be it. These things tend to even out in the long run. Losing bets become winners with a 96th minute goal or whatever, and if you have value on your side, the money will come.

I am still experimenting with the best way to take advantage of these results. The Correct Score market looks promising right now for home teams expected to win by 2 or more, and the Asian Handicap might also provide an opportunity. For the two goal plus teams, I’ve been backing relatively large on the Match Odds market, with smaller amounts on the Correct Score. Ideally both come in, but it’s not bad if the team ‘only’ wins.

Looking ahead to this weekend, and no games at the top level in England, but in Scotland, Celtic should win by 1 at Aberdeen and Hearts will beat Falkirk at home. The other three games at Hamilton, Kilmarnock and St Mirren will be draws.

On Sunday, Rangers are one of my 2 goal teams at home to Hibernian.

Tonight sees one goal home wins in League Two predicted for Burton Albion (2.16) v Crewe and Northampton Town (1.96) v Aldershot Town. The strike rate on these selections in this League is exactly 50%.

All one goal home wins in the Championship, except for draws at Doncaster Rovers and Swansea City. The top selections are Coventry City at 2.16 and Preston North End at 2.26. League One has a couple of 2 goal winners projected in Brentford (1.74) v Wycombe and Millwall (1.61) v Exeter City.

Good progress on Serie A - I'm up to December now, and with a quiet weekend ahead I should be caught up and it's off to La Liga.

2 comments:

Unknown said...

More Elo rating questions I'm afraid...

Firstly the adjustments to the formula, I'm guessing this is an extra co-efficient based on the margin of the win? In which case, how does your formula react to say a 1-0 win over Chelsea, compared to a 5-0 over Portsmouth?

Also, when you refer to the predicted score that keeps things closest to the status quo, would that status quo be the current ratings of the teams, or perhaps something like an average of the last 10 ratings?

If this is too detailed then don't worry. I'm just very intrigued!

Cheers, DJ

Cassini said...

A 1-0 win over Chelsea (rated at 2499) results in more points than a 5-0 win over Portsmouth (855)would. If the winning team was Manchester United, then they would gain 23 points for beating Chelsea 1-0 but only 15 for beating Portsmouth 5-0. While the final margin is a factor, I do not reward a 5 goal win by five times as much as a 1 goal win. I also differentiate between away and home wins.

As for what I mean by status quo, right now I use the most recent rating although you do raise an interesting idea with the moving average... More spreadsheet work!