Sunday, 26 February 2012

Spreading the risk, and a favour to ask

An interesting post by Peter Webb on the nature of the mature football markets, part of which I quote:
Lots of people focus on just one match, research it to death and try and make a judgement on value; that’s fair enough. But often the market is ‘hyper efficient’ so value is tricky to find. Even if you can find it, it can take ages for the system to mature to a profit. Therefore one of the things I tend to do is spread my risk across many matches. This speeds up the maturity of the system I am deploying and gives me a better chance of a regular payoff.
I couldn't agree more with this assessment which is why my screen looks like some kind of Jackson Pollock painting with markets and chatrooms etc open all over the place as I try to multi-task. Peter's post wasn't totally altruistic though, as he goes on to extol the virtues of trading with BA linked to an Excel spreadsheet (which I suppose he has every right to do) .

I thought this therefore would be an ideal opportunity to bring anyone who is interested up to speed with my amateurish attempts to develop an SG bot as discussed a while ago.

Painfully slow progress, to be honest. With the first signs of spring and slightly warmer weather business is starting to pick up again, and I'm often late home and tired. Notwithstanding that I seem to have run into a couple of 'logical' walls, one in particular relating in how to establish the current score. I don't think it's sensible to rely on the stats sites for this - I've seen far too many errors over the years. So, as mentioned I intend to poll the correct score markets looking for 1000 odds to back, indicating that score is no longer a 'live' or available one. My thinking here is to implement some kind of binary tree with 0-0 as the primary node and going from there...

If anyone feels able to point me in the right direction I'd be grateful....

4 comments:

  1. Hi Dave

    You may disagree, but I always feel that it is better to check for zero money available on the lay side rather than checking for odds of 1000 on the back side. For me, it's just a little bit safer.

    Your binary tree method is absolutely fine and will work. It may be a little arduous to code however. One other way you might consider
    is to assign each score-line a value thus:

    0 - 0 = 1
    0 - 1 = 2
    0 - 2 = 4
    0 - 3 = 8
    1 - 0 = 16
    1 - 1 = 32
    1 - 2 = 64
    1 - 3 = 128
    2 - 0 = 256
    2 - 1 = 512
    2 - 2 = 1024
    2 - 3 = 2048
    3 - 0 = 4096
    3 - 1 = 8192
    3 - 2 = 16384
    3 - 3 = 32768
    AU = 65536

    The beauty of using binary numbers in this way is that each value can be summed with any other (or others) to create a new unique number.

    So what you would do is, if each score-line has lay money available, then their value is valid and you sum it. If there is no lay
    money available, then you skip it.

    For example, if you summed all the available values and it came to 131071, then you would know that the current scoreline is 0-0. If you summed all the available values, and the total comes to 131056, then you would know that the scoreline is 1-0, and so on and so forth.

    You could pre-define all the original values as I have done above and all the concluded scorelines thus:

    131071 0 - 0
    126702 0 - 1
    117964 0 - 2
    100488 0 - 3
    131056 1 - 0
    126688 1 - 1
    117952 1 - 2
    100480 1 - 3
    130816 2 - 0
    126464 2 - 1
    117760 2 - 2
    100352 2 - 3
    126976 3 - 0
    122880 3 - 1
    114688 3 - 2
    98304 3 - 3
    65536 Any Unquoted

    The advantage of doing it this way is that the lookup method is the same for each and every scoreline, so everything runs through the same small bit of code.

    Hopefully I've made myself understood, but it wouldn't be surprising if I haven't as I'm on the Jack Daniels!

    Besides, you may think the whole concept is crap. If so, please feel free to completely ignore me.

    Eddie.

    ReplyDelete
  2. BTW, I just quickly knocked-up that second set of figures, so there's a good chance they're wrong. Do double-check my pissed maths.

    ReplyDelete
  3. Cheers for that Eddie, haven't checked the maths but like the concept.

    I was thinking along the lines of holding the current score node of the tree in a variable and looping from there, but reckon your solution is a bit more elegant and probably considerably less coding.

    Agree about checking for no money on the layside - I hadn't thought of that! It occurs to me that as the game approaches the end the money more or less vanishes from all but the current score so will have to incorporate some logic into that somehow or the thing won't know where the fcuk it's at. That would not be good for a Correct Score bot!

    Watch this space....

    ReplyDelete
  4. You're right about scorelines disappearing as the game progresses, so you'll have to add-in some logic that determines which scores to ignore and at which time - they don't make it easy, do they?

    ReplyDelete