Wilkies selections 2010-2011

[_wilkie_]

Hi All, I have recently been working on a computer system that predicts the outcome of football macthes, using a database of previous. The program has been tested with previous years in all english leagues upto and including the conference. Below are my picks for this weekend, split according to the confidence rating of the system(Team in caps & underlined is the chosen team). Strong Confidence: NOTTS COUNTY vs Yeovil SOUTHAMPTON vs Rochdale WALSALL vs Colchester BURY vs Gillingham CHESTERFIELD vs Lincoln Southend vs TORQUAY Altrincham vs KIDDERMINSTER CRAWLEY vs FleetWood Town Hayes & Yeading vs Histon - Draw KETTERING vs AFC Wimbledon MANSFIEDL vs Tamworth Minor Confidence: Cambridge vs Gateshead - Draw Stevenage vs Crewe - Draw Oxford vs Morecambe - Draw Macclesfield vs Stockport - Draw Barnet vs CHELTENHAM Brentford vs SHEFFIELD WEDNESDAY Dag & Red vs Leyton Orient - Draw Any feedback on these 'picks' would be good, also, any input on the best way to place these bets i.e singles, trebles, accumulators. Generally, I place only accumulators, however, due to the amount of matches usually selected, this is not really the way I wish to proceed moving forward, once a decision has been made I will update accordingly. Enjoy, Wilkie

[Jenspm]

Re: Computer Program Predictions Does your program predict the % chance of a winner, and thus take the given odds into consideration?

[_wilkie_]

Re: Computer Program Predictions The program does predict the % of home draw away, which is one factor I consider when chosing which section to place the pick in dependant on the outcome given, which is given twice, from two different methods. If the outcomes are the same from both methods, I will use the pick. No, I do not use odds or calculate odds currently, do you recommend this to enhance the system? If so, in what way would this be best used? Wilkie

[Datapunter]

Re: Computer Program Predictions

I have recently been working on a computer system that predicts the outcome of football macthes, using a database of previous.

That description would cover just about any "computer" system, it says nothing.

NOTTS COUNTY vs Yeovil

Unless you have some method if defining of the availabe odds represent value a "pick" has no meaning whatsoever. Would you still bet on NOTTS COUNTY to win the match if the odds where 1.01 ? I think not, so at what point is it worth betting ?

The program does predict the % of home draw away, which is one factor I consider when chosing which section to place the pick in dependant on the outcome given, which is given twice, from two different methods. If the outcomes are the same from both methods, I will use the pick.

Here i have absolutely no idea what you are talking about. I'd love to give you some feedback but it's all pretty vague and mysterious.

[Jenspm]

Re: Computer Program Predictions

The program does predict the % of home draw away, which is one factor I consider when chosing which section to place the pick in dependant on the outcome given, which is given twice, from two different methods. If the outcomes are the same from both methods, I will use the pick. No, I do not use odds or calculate odds currently, do you recommend this to enhance the system? If so, in what way would this be best used? Wilkie

I hate to plug my own threads, but this gives a very brief run-down on value, have a look. The discussion going on there is also quite interesting :ok http://www.punterslounge.com/forum/f21/how-beat-bookies-value-betting-guide-inc-spreadsheet-102626/ In short, if the product of your percentage and the odds you can get is over 100, take the bet, else, leave it alone.

[FleetFanatic]

[_wilkie_]

Re: Computer Program Predictions

That description would cover just about any "computer" system, it says nothing. Unless you have some method if defining of the availabe odds represent value a "pick" has no meaning whatsoever. Would you still bet on NOTTS COUNTY to win the match if the odds where 1.01 ? I think not, so at what point is it worth betting ? Here i have absolutely no idea what you are talking about. I'd love to give you some feedback but it's all pretty vague and mysterious.

the computer program is coded in unibasic with a unidata backend. it uses previous matches data to see how well the teams did from their making a prediction for the match at hand. i am unsure currently as to how to pick the matches with value, although i will dig into the link posted above to get a better understanding. The reason for me not needing to know the odds ia because I have always bet on accumulators and not singles, hence the odds always count.

[_wilkie_]

Re: Computer Program Predictions jenspm thanks for your link, I will be sure to take a look at it, much appreciated. Fleetfanatic, im interested as to how your predictions are calculated, as my system works of from previous 5 performances to find the team playing better. When I get a moment I will take a look at my own system as this may be a bug in calculation to say that Kettering will win, we shall see.

[FleetFanatic]

Re: Computer Program Predictions The tool I use isn't available at the request of my developer :lol but the basics are league table positions (difference)* home team's home form for season vs away team's away form for season** and last 6 form - weighted for bigger winning margins, losing at home more damaging than losing away, away wins rewarded more than home wins, etc etc - for home side only*** *this obviously fluctuates at the start of the season as one win can change league positions quite dramatically. Hence I feel my ratings get better as the season progresses. **again, after just 4 or 6 games, there is not a lot of "home data" or "away data" - so again this improves as the season progresses. ***this bit always puzzles people, but even the top sides tend to cock up away rather than at home. My ratings rate the chance of a home win, not the chance an away win. But technically, if the ratings say there's sod all chance of the home side winning, then that still implies that the away side is favoured.

[Datapunter]

Re: Computer Program Predictions

The reason for me not needing to know the odds ia because I have always bet on accumulators and not singles, hence the odds always count.

A serious misconception as the same applies as with singles. So you got 5 teams that you think will win, if the odds where 1.01 on each would you bet on 5 fold accumulator ? If you answer yes you got a big problem. You wouldn't as there would be no value whatsoever in a bet like that. If you don't understand this then please stop betting until you do as you are throwing money away. You must determine, based on your prediction, if the odds represent value or not and only bet when they do. In that case it makes little difference if you do singles or accumulators. If your perception of value is correct that is. Let's say you have estimated team A to have a 50% chance of winning the match. If you can get odds above 2.00 then you have value in the odds on offer according to your prediction model. If the available odds are below 2.00 there is no value in them. So how do you know if your model is accurate enough to make a profit ? If your model allows you to find value more often than not you'll make a profit. The only way to actually know if that is the case, is to apply the system to a fairly large group of bets, and see if you do end up in profit. On each individual bet you will never know if value is or was present as you can never determine the exact % chance of a selection winning. Some bets may have value and win or lose. Some bets may not have value and win or lose. The only thing you can measure is your profit over time and that tells you that on the past X bets you have indeed found, on average, some value. Or not in the case you got a loss.

[Jenspm]

Re: Computer Program Predictions To add on to what Datapunter said, Unless you have a system that is extremely accurate, singles are the way to go. Why? Because you only have to get the percentages of a single game right - do a five-fold and a single misread game ruins all the value in the other four bets. If your system is consistent and perfect, there will be no difference if you bet singles or five-folds. However, if your system is slightly off, singles will be better. So to conclude, go for singles. If you go to http://www.football-data.co.uk , you can download historic odds for all games in the English leagues for the past years, and then backtest your system to see how profitable and accurate it is. :ok

[_wilkie_]

Re: Computer Program Predictions

The tool I use isn't available at the request of my developer :lol but the basics are league table positions (difference)* home team's home form for season vs away team's away form for season** and last 6 form - weighted for bigger winning margins, losing at home more damaging than losing away, away wins rewarded more than home wins, etc etc - for home side only*** *this obviously fluctuates at the start of the season as one win can change league positions quite dramatically. Hence I feel my ratings get better as the season progresses. **again, after just 4 or 6 games, there is not a lot of "home data" or "away data" - so again this improves as the season progresses. ***this bit always puzzles people, but even the top sides tend to cock up away rather than at home. My ratings rate the chance of a home win, not the chance an away win. But technically, if the ratings say there's sod all chance of the home side winning, then that still implies that the away side is favoured.

cheers for that mate, my system has 2 separate selection processes in, I only go with the pick if both systems have the same outcome. One process has what I call 4 'cores', of which one core is a weighted goals system. Unlike you, I only use the results from upto the last 5 games, as I dont believe early matches have any relation to a game towards the end of the season, although I will be following your picks closely as it seems interesting.

[_wilkie_]

Re: Computer Program Predictions

A serious misconception as the same applies as with singles. So you got 5 teams that you think will win, if the odds where 1.01 on each would you bet on 5 fold accumulator ? If you answer yes you got a big problem. You wouldn't as there would be no value whatsoever in a bet like that. If you don't understand this then please stop betting until you do as you are throwing money away. You must determine, based on your prediction, if the odds represent value or not and only bet when they do. In that case it makes little difference if you do singles or accumulators. If your perception of value is correct that is. Let's say you have estimated team A to have a 50% chance of winning the match. If you can get odds above 2.00 then you have value in the odds on offer according to your prediction model. If the available odds are below 2.00 there is no value in them. So how do you know if your model is accurate enough to make a profit ? If your model allows you to find value more often than not you'll make a profit. The only way to actually know if that is the case, is to apply the system to a fairly large group of bets, and see if you do end up in profit. On each individual bet you will never know if value is or was present as you can never determine the exact % chance of a selection winning. Some bets may have value and win or lose. Some bets may not have value and win or lose. The only thing you can measure is your profit over time and that tells you that on the past X bets you have indeed found, on average, some value. Or not in the case you got a loss.

Some good advice boss, I do feel that I need to adjust my program to point of the value of a pick as well, even more so after reading your comments here, and the posts in http://www.punterslounge.com/forum/f21/how-beat-bookies-value-betting-guide-inc-spreadsheet-102626/.

[_wilkie_]

Re: Computer Program Predictions

To add on to what Datapunter said, Unless you have a system that is extremely accurate, singles are the way to go. Why? Because you only have to get the percentages of a single game right - do a five-fold and a single misread game ruins all the value in the other four bets. If your system is consistent and perfect, there will be no difference if you bet singles or five-folds. However, if your system is slightly off, singles will be better. So to conclude, go for singles. If you go to http://www.football-data.co.uk , you can download historic odds for all games in the English leagues for the past years, and then backtest your system to see how profitable and accurate it is. :ok

I think I will go with both folds and singles at first, see how I get on then chose the future method from there, although I appreciate the point of one misread corrupting the whole bet. Also, funny you mention www.football-data.co.uk , this is where I import data from to my database.

[FleetFanatic]

Re: Computer Program Predictions

Unlike you' date=' I only use the results from upto the last 5 games, as I dont believe early matches have any relation to a game towards the end of the season, although I will be following your picks closely as it seems interesting.[/quote']You misunderstood me, I said a combination of overall league data & current form. Maximum score for the league position can only be 23 (top home to bottom in 24 league division). The home home form vs away away form could be anything (right now it's only going to be about 4/5/6 at most). Maximum score on last 6 current form could be 39, it's lowest -18. So a team that starts well then falls away could, in theory have a good league difference over its opponent, but their current form if on a bad run will knock the overall rating right down anyway. In theory I could see teams rate over 70, I think the highest this method rating has seen is only the low 60s. If a team rates 30 or higher in my ratings, it's rare for them to lose, a lot of them win as expected, especially as the season progresses. It's just getting 3 together every week that aren't a ridiculously short price that's the challenge. :lol But as this is still early in the season, some teams fluctuate in form.... 10 games or so in & we normally start to see things bed in better. :ok

[Machine]

[FleetFanatic]

Re: Computer Program Predictions Odds aren't factored into my calculations either. If you got 2 teams with similar ratings, one priced 1/3 and the other priced 8/11, and you believe your ratings to be a reasonable indication, which team would you back? It's simply a question of value. Expecting the 30+ teams to win is from past seasons - although there is a weird correlation of "the shorter the price, the more likely they will be the one to balls up" :lol If you ignore the minor non-league teams in my list, do you notice how many teams we have in out top 5 or so that are the same? :) The only home you have as BACK is Carlisle - the only league team rated 30+? Also Carlisle.

[Machine]

Re: Computer Program Predictions Yes did notice that the top ratings were quite close, as are the bottom ones. As for the lower price ones, Im not against short odds on as I look at it if you can get at least 10% on your bet your doing better than the 1.5% at the bank. As the match data increases and things stabilise a little more Ill be taking the top 20% ( 80>100) as my home picks. Carlisle currently 2.15 @ betfred :clap If I had odds of 1/3 and 8/11 and trusted my ratings Id back both and dutch the stake as I only back singles. Odds are not an indication of how good the bet is, bookies dont give a flying fcuk what odds they give out as long as they manage their book to the overound. So if 1 of the 2 similar rated teams comes in to 1/3 it screams the bookies trying to get his overound back on track because the punters are hammering it or the opposite no one is interested in the 8/11.

[FleetFanatic]

Re: Computer Program Predictions I'm not saying odds are an indication of "a good bet" - what I mean is if the ratings are accurate, and then if you have 2 teams rated the same but at odds of 1/3 and 8/11, surely, based on the ratings alone, the 8/11 is value compared to the 1/3?

[Machine]

Re: Computer Program Predictions Thats the six million dollar question. I once read an article by a professor of economics called Leighton Vaughan Williams in it he states

The assumption, or at least the hypothesis, must be that the accuracy of the betting markets is created out of the information and intuition of many people rather than the conclusions of a few. Those myriad people feed in the best information and intuition they can because their own financial rewards depend directly upon them. And it really is a case of “follow the money” because those who know the most are likely to bet the most.

Thus going off that hypothesis 8/11 would not be the correct play, a case of false "value" so to speak. I can see the thinking behind going for the 8/11 purely on a return basis pro rata but that may not always be the best move to make. Which road to take is the hardest decision.

[FleetFanatic]

Re: Computer Program Predictions But either you believe your ratings or you don't. If you have two teams rated the same, then you are saying they both have the same chance of winning as the other..... and one is the bigger price, then in the eyes of your ratings, it must the better value. If it isn't, and is "false value" then surely that is an admission that you don't believe your ratings to be correct?

[Machine]

Re: Computer Program Predictions I answered your question before, If I believed in my ratings then I would back both as that is what my ratings tell me to do. Dont forget your ratings only give you the most probable winner, not the price at which they should be backed at As the "false value" is unknown dutching the pair (in this case) is the obvious answer, equalizing the risk over both prices.

If I had odds of 1/3 and 8/11 and trusted my ratings Id back both and dutch the stake as I only back singles.

[FleetFanatic]

Re: Computer Program Predictions Yes, your ratings are giving the most probable winner. Therefore it follows that if it rates 2 teams just as strong as one another to win their matches, then if the ratings are to be correct, the better priced team is clearly better value! Good luck dutching a 1/3 & an 8/11 shot; £10 example: 1.33 stake £5.68 1.75 stake £4.32 Returned if either selection wins: £7.56 So, unless they both win, you have lost money. I understand your theory of damage limitation in the event of only 1 of the 2 winning, but if they both win, you might as well have lumped it all on the bigger price. I guess it's down to individual preference. ;) My preference is not to back 1/2 or lower.

[Machine]

Re: Computer Program Predictions People get giddy when they see bigger prices I suppose. Say your ratings came from these variables :- 1/3w , 1/3w,1/3w, 1/3w,8/11L 8/11L 8/11L 8/11L Would you still back the 8/11 shot.

[FleetFanatic]

Re: Computer Program Predictions Ah but if that was the case, then what you would be saying is that your ratings aren't as accurate as you thought they were, as the 8/11 shots kept losing despite their ratings. From the performance I've seen in my ratings in previous seasons, an 8/11 shot would still be a solid bet, providing it rated high enough. I have every faith in the ratings, I know they will not always be right, but they are a very good indicator overall. Oh, and for the record, an 8/11 shot still isn't a price to get giddy over, as you put it. :lol

[Machine]

Re: Computer Program Predictions

Ah but if that was the case' date=' then what you would be saying is that your ratings aren't as accurate as you thought they were, as the 8/11 shots kept losing [b']despite their ratings.

You obviously cannot get what Im saying. Your ratings are made up of a multitude of variables,some positive some negative and collectively they make up the whole rating.(if they were all positive every rating would be 100%). Now if the 8/11 falls into the negative part of the the whole rating collective then no matter how many times you back it is doomed to failure and thats a fact (as rafa would say). The same also goes for the 1/3. You can call it better "value",you can tie it to a tree and call it Toby if you want. But if it is from the negative side of the ratings make up it will still be the wrong bet to make. If your ratings give you 2 teams you should back them both not just the higher price else it is you who do not believe the accuracy of your rating. :ok

[FleetFanatic]

Re: Computer Program Predictions My ratings aren't done on % chance, they are done on a score rating, and I simply list them from highest score to lowest score. The highest score isn't a % chance, it is simply the strongest rated home side. Theoretically, the highest rating I could see would be on the last day of the season, could be 106, if the team was top of the league at home, having won all it's home games against a side that had lost every match away from home all season. On top of this they would have had to have won their last 6 games all by 3 goals or more. Realistically, low 60's is about the highest I ever see. At this stage in the season I am already seeing teams in the lower non-league rate in the 40s, but their ratings will change as their current form does. Therefore the 1/3 & 8/11 shots I talk about don't fall into any "negative category". In my methods, if they are near the top of the ratings it is only positive stuff, sides with negative data get punished & have a low or negative rating. Hence to me the 8/11 is value compared to the 1/3, if they both rate the same. That is why we disagree, as our methods are different - it's just this week we've come up with similar "top teams". My top rated teams don't always win, but those rated 30+ win a lot more than lose overall. And most weeks I will have lots of teams rated 30+. So based with that knowledge, why would I back teams rated 30+ at 1/3 if I have teams I can back at 8/11 or better?

[Machine]

Re: Computer Program Predictions Well agree to differ then ;)

Good luck dutching a 1/3 & an 8/11 shot; £10 example: 1.33 stake £5.68 1.75 stake £4.32 Returned if either selection wins: £7.56 So, unless they both win, you have lost money. I understand your theory of damage limitation in the event of only 1 of the 2 winning, but if they both win, you might as well have lumped it all on the bigger price.

By backing both I am getting the one which is the correct price for its rating by backing the 8/11 only you are hoping its the correct price for its rating, if they both win at their respective %ages no one has gained or lost.

[_wilkie_]

Re: Computer Program Predictions Here are my predictions for this weekend stoke blackburn exeter huddersfield nottscounty accrington chesterfield oxford shrewsbury crawley eastbourne wrexham york Once i have decided on the method of betting I will be doing with these I will update this post.

[_wilkie_]

Re: Computer Program Predictions OK, so after posting my predictions earlier, I have now decided on the bets that will be put on for the weekend, and they are as follows: Bet 1 - Chesterfield & Exeter Bet 2 - Blackburn & Notts County Bet 3 - Stoke & Accrington Bet 4 - Eastbourne & Wrexham Bet 5 - Bet 1 + Bet 2 Bet 6 - Bet 1 + Bet 3 Bet 7 - Bet 1 + Bet 4 Bet 8 - Bet 2 + Bet 3 Bet 9 - Bet 2 + Bet 4 Bet 10 - Bet 3 + Bet 4 Bet 11 - Bet 5 + Bet 3 Bet 12 - Bet 5 + Bet 4 Bet 13 - Bet 6 + Bet 4 Bet 14 - Bet 8 + Bet 4 Bet 15 - Bet 1, 2, 3 & 4 Cheers,