Geoff at full-time betting asked
Hi- fascinating reading as always- what are you basing the "strong draws" research on? is it your own stats or something that is widely available?I also had an e-mail from Steve who wrote
Thanks for the reply below my system is still going although the last week has seen a slight drop in the strike rate. On the blog you mention that you consider past form etc. Are you able to give any advice on how you incorporate past form into the ratings? Do you also consider past meetings between two teams?Since the questions overlap a little, I'll address them together.
The answer is that strong draws, along with all my football picks, are generated from my own statistics. The main component is each team's basic rating from an Elo based but personalised system, modified by recent form and historical home / away record.
As for how to incorporate form, to me it seems logical to put more weighting on the most recent result and reduce the weighting as you go back. I use the most recent six games for this. Why six? - it seems like this is a big enough number to smooth out any abnormal results, yet small enough to be meaningful. Besides, the ratings themselves are a record of long-term form.
Also, I don't just use the result, but I use the result relative to the expected result, so for example, Crystal Palace losing at Chelsea would penalise Palace a lot less than Palace losing at home to Brighton. When I see form shown in the manner of WwLDdl, to me, it's almost meaningless. I use the difference in ratings after each result so that I have something like +6, -3, +3, -5, +12, -9 and then weight them.
As for whether or not I consider previous meetings between the two teams, the answer is no. I don't believe that this adds anything to the ratings. Teams usually play each other just twice a season, (excluding the Scottish Leagues), and while it's mildly interesting to know that it's been x years since team A beat team B, with teams changing year by year, it really means nothing. There might be a case for a 'derby game' adjustment though.
Another nice winner last night with Real Sociedad (-1.1) comfortably beating Deportiva La Coruna at 2.14. Football Elite were on this one also, and as he commented in his update, the price on this game plunged late to around 2.0 at kick-off. 2.0 would have been close to taking this one off the value list, but getting on a couple of days ago at 2.14 meant this was comfortably value.
It was a good weekend for Matt's Football Elite picks. The three Recommended Bets all won with clean sheets. In addition to Real Sociedad's 3-0 win last night, Villareal won 2-0 on Sunday, and Sunderland beat Aston Villa 1-0 on Saturday.
The amount of people I have seen who say they use the last 6 games form in one way or another is uncanny. Some people base whole systems around it.
ReplyDeleteThere must be a sheep mentality at work. Was 6 the number of games the pools predictions used to have the WXLDXX for in the newspapers? I wonder if it's subconsciously derived from that.
To the best of my knowledge, the number six has no special significance when assessing football form.
In order to say that Real Sociedad @-1.1 is value or not, presumably, you compare your model's expected handicap for the match (-1.1) to the market's home/draw/away odds or to the asian handicap closest to evens. People often call "value" one of two situations. First: "My model says -1.1 and the market says -0.5. Therefore, this is value." Second: "Having tested my model (with fixed structure and parameters) by paper-trading every match, where my model's handicap is different from the market's handicap by a certain threshold value of x, the strategy has been comfortably profitable. Therefore, this is value." For easier reference, I would call the first, theoretical value and the second, historical value. Obviously, theoretical value is a lot easier to find than historical value. I guess my question is: when you talk about value, which one do you refer to?
ReplyDelete"Imaginary value"
ReplyDelete