Statistical Football Prediction Model 2008

[Rushian]

Re: Statistical Football Prediction Model 2008 A lot has been said about the robustness of a system. I am stil not sure whether one shoudl use the chi-sq measure or the probability of achieving a negative yield in the long run. I (think that I) have calculated both but I would be grateful if someone could check my calculations for the chi-sq, given the expected number of winners/losers.

Robustness Statistics				Monitoring Statistics
Expected Winners:		216.2		95% Upper Level:		16.4%
Expected Losers:		381.8		Av. Yield:		4.5%
χ2 Indicator:		68.6%		95% Lower Level:		-7.5%
				SE (Yield):		6.1%
				Prob (-ve Yield):		23.3%

So based on the expected winners/losers and the actual figures so far (i.e. 216.2 expected winners with 228 actual ones), there is a 68.6% that the system is doing well not because of chance (?). I prefer to focus on the probability of a negative yield in the long run which I estimate at 23.3%, a figure which is quite high! This partially explains my reservations in terms of the long-run profitability of the system. Note that the standard error of the yield is at 6.1% (down from 7.2% at bet number 463 as I mentioned in a previous post) which is very high when compared to the mean yield of 4.5% currently. A 95% confidence interval would put the yield between -7.5% and 16.4%.

[Rushian]

[Rushian]

[Rushian]

Re: Statistical Football Prediction Model 2008 One more breakdown that I did not include yesterday is the system's performance by bookmaker so far:

Bookmaker		Bets	Strike Rate	Yield
Bet365		265	33.6%	-1.8%
Bwin		35	42.9%	13.6%
Gamebookers		27	44.4%	-17.9%
Interwetten		14	85.7%	47.8%
Ladbrokes		101	34.7%	-12.7%
Sportingbet		8	37.5%	15.6%
William Hill		18	44.4%	0.8%
Stan James		38	39.5%	27.2%
VC Bet		35	34.3%	13.0%
Blue Square		57	47.4%	37.4%

The majority of bets have been with Bet365, although this is also skewed by the eay my spreadsheet is set to pick the first available bookmaker with the best price (so in cases of equal best prices, it follows the order above).

[Smokin Joe]

Re: Statistical Football Prediction Model 2008

The number of picks has increased since last year there were (1084 picks out of 2036 matches i.e. 53.2%) whereas so far this year' date=' the system picked out 598 out of 1058 matches i.e. 56.5%. This extrapolates to around 1150 picks for the whole year.[/quote'] Rushian, to me that is the most enlightening thing about your system. You have a very high percentage of picks - ie you place a bet in more than 50% of matches. I much prefer to see a system with many bets and a steady yield than a system which places bets in just 5% of games - though this is probably a personal preference. It is a pity that this thread hasn't received the input it deserves from other posters. I would guess that you have a much superior knowledge of statistics than the average poster on this forum, so unfortunately for you I don't think that you will be asked questions that you haven't though of yourself.

[Rushian]

Re: Statistical Football Prediction Model 2008 Thanks for the comments SJ! Maybe people are still celbrating the turn of the new year rather than replying to my statistical ramblings ... Anyway, the football-data website is still to be updated but I have to leave so I'll put the provisional picks up with a few rules for selections as usual. I've had a look at different bookmakers and from a quick scan the system picks are:

No	Div	Date	Home Team	Away Team	Bet	Stake	Bet Odds	Bookmaker	W/L/P	Profit/Loss
599	E2	3/1/2009	Brighton	Northampton	D	10	3.30	Bet365	P	0.00
600	E2	3/1/2009	Carlisle	Walsall	H	10	2.10	Bet365	P	0.00
601	E2	3/1/2009	Colchester	Tranmere	H	10	2.30	William Hill	P	0.00
602	E2	3/1/2009	Huddersfield	Oldham	H	10	2.60	William Hill	P	0.00
603	E3	3/1/2009	Barnet	Grimsby	A	10	3.40	Sportingbet	P	0.00
604	E3	3/1/2009	Bradford	Shrewsbury	H	10	2.40	William Hill	P	0.00
605	E3	3/1/2009	Morecambe	Lincoln	H	10	2.37	Bet365	P	0.00
606	E3	3/1/2009	Rochdale	Accrington	H	10	1.66	William Hill	P	0.00

However:

• Carlile is not a pick if it is available at 2.20 or more.
• Colchester is not a pick if it is available at 2.38 or more.
• Herefored becomes a pick if the best available price is less than 4.5
• Grimsby is not a pick if it is available at 3.57 or more.
• Bradford is not a pick if it is available at 2.44 or more.
• Wycombe-Bury will be a pick for a Draw if the Draw is available at 3.41 or more.

I hope this covers everything. If I am back by kick-off time, I'll try to ammend the list correctly. Fingers crossed as always :hope

[Smokin Joe]

Re: Statistical Football Prediction Model 2008 Rushian, what odds figure do you use from the football-data files? Is it the Betbrain average, maximum or just specific books? Presumably you used the same criteria during the testing period? The thing that gets me about doing that are the advent of the Exchanges. Say for example in your testing set (like the Arsenal bet we talked about last week) where they weren't a bet because their odds were greater than 3 (for instance). What would have happened if the best odds were 2.95 with the bookies that you use, and so would have been a bet, but you could actually have backed Arsenal at 3.25 on Betdaq (as I did). How does that work? I suppose my general question is, is there not something intrinsically wrong with using a best price ceiling which is generated by bookies odds, when in the vast majority of the times those odds will be beaten on the exchanges? For example, your yield at the moment is 4.5%. However, if you have used either best prices or the exchanges for every selection I have no doubt that your yield would be a fair bit higher, let's assume 6%ish. However, undoubtedly some of those bets would have been ruled out due to the larger odds being available on the exchanges which would have breached your price ceiling. How does that sit with you, as that just doesn't seem right to me?

[Rushian]

Re: Statistical Football Prediction Model 2008 I have just managed to get back, and got the files from football-data. From there, the only alteration is that Carlisle is no longer a bet as it is availabe at Ladbrokes at 2.20. Therefore the ammended list of selections are:

No	Div	Date	Home Team	Away Team	Bet	Stake	Bet Odds	Bookmaker	W/L/P	Profit/Loss
599	E2	3/1/2009	Brighton	Northampton	D	10	3.40	Ladbrokes	P	0.00
600	E2	3/1/2009	Colchester	Tranmere	H	10	2.30	Bwin	P	0.00
601	E2	3/1/2009	Huddersfield	Oldham	H	10	2.60	William Hill	P	0.00
602	E3	3/1/2009	Barnet	Grimsby	A	10	3.40	Sportingbet	P	0.00
603	E3	3/1/2009	Bradford	Shrewsbury	H	10	2.40	William Hill	P	0.00
604	E3	3/1/2009	Morecambe	Lincoln	H	10	2.37	Bet365	P	0.00
605	E3	3/1/2009	Rochdale	Accrington	H	10	1.73	Ladbrokes	P	0.00

SJ, I'll try to reply to your question shortly.

[Rushian]

Re: Statistical Football Prediction Model 2008 Smokin Joe (and anyone else interested) here goes:

Rushian, what odds figure do you use from the football-data files? Is it the Betbrain average, maximum or just specific books?

I use the best prices out of the 10 available bookmakers included in football-data. I did not use the Betbrain figures at all.

Presumably you used the same criteria during the testing period?

Yes.

The thing that gets me about doing that are the advent of the Exchanges. Say for example in your testing set (like the Arsenal bet we talked about last week) where they weren't a bet because their odds were greater than 3 (for instance). What would have happened if the best odds were 2.95 with the bookies that you use, and so would have been a bet, but you could actually have backed Arsenal at 3.25 on Betdaq (as I did). How does that work? I suppose my general question is, is there not something intrinsically wrong with using a best price ceiling which is generated by bookies odds, when in the vast majority of the times those odds will be beaten on the exchanges?

The way I see it, we are talking about two different things: On the one hand, it's "getting the best price for a particular selection". So if my system/model estimates that Arsenal should be priced at 2.50, and anything above that is considered as value, I have to get the best price possible... On the other hand, what if I decide to use the bookies odds as proxies to the true probabilities? So I say that I think that Arsenal should be priced at 2.50, but if any bookie offers Arsenal at more than 2.80 then I won't back them because there is a good chance that I have mispriced them. Because bookies know what they are doing, and if there is some value to be found, it's very, very little, and they won't be mispricing teams by a lot more.

For example, your yield at the moment is 4.5%. However, if you have used either best prices or the exchanges for every selection I have no doubt that your yield would be a fair bit higher, let's assume 6%ish. However, undoubtedly some of those bets would have been ruled out due to the larger odds being available on the exchanges which would have breached your price ceiling. How does that sit with you, as that just doesn't seem right to me?

I don't know what the yield would be if I had used the best prices for the 598 selections so far, but if I didn't include bookies probabilities in the selection procedure (i.e. if there was no upper limit for a bet to qualify as a value bet) there would have been 941 bets, returning a profit of 120 units which translates to 1.3% yield. Is this helpful at all or am I rambling again?

[Smokin Joe]

Re: Statistical Football Prediction Model 2008 Rushian, I understand what you are saying, and I am not advocating the abandoning of your upper limit (as it obviously serves its purpose). I think you said earlier that you aren't yet placing bets on your selections. If that is correct you will hypothetically have to answer this question: Say you were placing bets on your selections, and Carlise was only available at 2.1 on Ladbrokes - so it would have been a selection, but you could get 2.25 on Betfair or Betdaq (which is a reasonable assumption to make). What bet would you place (if any)? I assume (for your own sanity) that you would bet the 2.25. If that is the case, then I can understand where you are coming from. However, the fundamental problem that I would have when I am costructing a system is that when I look at the odds available in the historic data files on football-data, I assume that on the whole, if I had been actually placing bets I would have got better odds than the odds listed due to the fact that neither of the exchanges odds are listed. Hence, I would find it difficult to set a ceiling, as I would effectively be looking at "apples", but feeling that I would actually be placing bets at odds of "oranges". But, it obviously works for you.

[Smokin Joe]

Re: Statistical Football Prediction Model 2008 Thinking about it further, we just have a different approach. Mine is neccessitated by the fact that I have traditionally looked at the average Betbrain odds that is contained in the football-data files, whereas you have looked in detail at the actual odds being offered. Perhaps it is because of this shortcut I have taken that I don't feel comfortable about accurately setting a price ceiling - whereas you do have that luxury.

[Rushian]

Re: Statistical Football Prediction Model 2008

Rushian, I understand what you are saying, and I am not advocating the abandoning of your upper limit (as it obviously serves its purpose). I think you said earlier that you aren't yet placing bets on your selections. If that is correct you will hypothetically have to answer this question: Say you were placing bets on your selections, and Carlise was only available at 2.1 on Ladbrokes - so it would have been a selection, but you could get 2.25 on Betfair or Betdaq (which is a reasonable assumption to make). What bet would you place (if any)? I assume (for your own sanity) that you would bet the 2.25. If that is the case, then I can understand where you are coming from. However, the fundamental problem that I would have when I am costructing a system is that when I look at the odds available in the historic data files on football-data, I assume that on the whole, if I had been actually placing bets I would have got better odds than the odds listed due to the fact that neither of the exchanges odds are listed. Hence, I would find it difficult to set a ceiling, as I would effectively be looking at "apples", but feeling that I would actually be placing bets at odds of "oranges". But, it obviously works for you.

Yes, I said a few times tha this is simply a paper trail to check the system and hopefully get some feedback on how to further improve it. So I haven't been placing on a regular basis bets based on these picks. You mention the availability of exchanges but these do not necessarily change the philosophy of my model selections. Once you have removed the commission, they can be considered as yet another bookie for the system to track. Faced with the hypothetical scenario that you are setting, I would have taken Carlisle at 2.25 (minus any commissions) on Betfair (assuming that the price fell within my upper limit). The reason that I am paper-trailing with bookies odds and not exchanges is purely a techinical one, as the data are available in such a nice format. Before going live with actual money, I would need to seriously consider my options in terms of ease of placing the bets, shopping around for the best price, and a million other factors which I probably have not considered yet.

[Rushian]

[Rushian]

Re: Statistical Football Prediction Model 2008 Following Smokin Joe's comments on bet selection and available odds, I decided to do some further analysis on the systems results. To set the framework, consider the following scenario: The system estimates the chances of a team to win at 40% and the best available odds are 2.55. The estimated edge (or perceived value) of the system is calculated as (40% * 2.55) - 1 = 2% Value betting instructs that you wouldn't bet on this team to win if the odds are less than 2.5 (i.e. 1/40%). What I have been trying to explain, is that the system places an upper limit too on the estimated edge before making the selections (in fact, I've set this to 20%). The following is a table with the system's number of bets, strike rate and yield for bets within different bands of estimated edge.

Estimated Edge Bands		Bets	Strike Rate	Yield
0%	5%	193	38.9%	-1.5%
5%	10%	164	39.6%	3.1%
10%	15%	117	38.5%	10.6%
15%	20%	127	34.6%	8.6%
20%	30%	142	32.4%	3.9%
30%	40%	78	28.2%	4.2%
40%	50%	55	20.0%	-21.7%
50%	Max	70	20.0%	-19.2%

The 601 bets of the system so far are the first four lines of this table. It seems that the system is registering marginal losses in those bets where the perceived edge is very small (less than 5%), it then starts to record profit in matches where the edge is higher (between 5% and 15%), before those profits start to diminish (edge between 15% and 40%). It then returns to red for anything bigger than 40%. So maybe a further improvement for the system would be to increase slightly the lower level. I would appreciate anyone's thoughts/experiences on this.

[Smokin Joe]

Re: Statistical Football Prediction Model 2008

Following Smokin Joe's comments on bet selection and available odds, I decided to do some further analysis on the systems results. To set the framework, consider the following scenario: The system estimates the chances of a team to win at 40% and the best available odds are 2.55. The estimated edge (or perceived value) of the system is calculated as (40% * 2.55) - 1 = 2% Value betting instructs that you wouldn't bet on this team to win if the odds are less than 2.5 (i.e. 1/40%). What I have been trying to explain, is that the system places an upper limit too on the estimated edge before making the selections (in fact, I've set this to 20%). The following is a table with the system's number of bets, strike rate and yield for bets within different bands of estimated edge.

Estimated Edge Bands		Bets	Strike Rate	Yield
0%	5%	193	38.9%	-1.5%
5%	10%	164	39.6%	3.1%
10%	15%	117	38.5%	10.6%
15%	20%	127	34.6%	8.6%
20%	30%	142	32.4%	3.9%
30%	40%	78	28.2%	4.2%
40%	50%	55	20.0%	-21.7%
50%	Max	70	20.0%	-19.2%

The 601 bets of the system so far are the first four lines of this table. It seems that the system is registering marginal losses in those bets where the perceived edge is very small (less than 5%), it then starts to record profit in matches where the edge is higher (between 5% and 15%), before those profits start to diminish (edge between 15% and 40%). It then returns to red for anything bigger than 40%. So maybe a further improvement for the system would be to increase slightly the lower level. I would appreciate anyone's thoughts/experiences on this.

Interesting stats. As you say, the easy conclusion to make is to increase your floor from 0% edge - this looks it will help your yield no end. How did the low band perform in your validation period?

[Rushian]

Re: Statistical Football Prediction Model 2008 I hadn't checked this angle so first of all, I think a "Thank you" is in order for pretty helpful and valuable feedback :ok I have just looked at my spreadsheets, and this is the corresponding table for the validation data:

VALIDATION DATASET
Estimated Edge Bands		Bets	Strike Rate	Yield
0%	5%	331	38.4%	3.0%
5%	10%	286	38.1%	14.1%
10%	15%	275	41.8%	13.7%
15%	20%	192	31.8%	-12.1%
20%	30%	295	31.5%	-2.6%
30%	40%	168	26.8%	3.0%
40%	50%	88	26.1%	-3.8%
50%	Max	175	18.9%	-5.7%

Not sure what to make of it yet...

[Rushian]

Re: Statistical Football Prediction Model 2008 Everything has been called off for tonight so next set of selections will be for the weekend.

[Rushian]

Re: Statistical Football Prediction Model 2008 Twenty selections for this weekend with quite a few longshots around... Maybe this is what is needed to push the mean bet/strike odds up, closer to the validation figures posted before...

No	Div	Date	Home Team	Away Team	Bet	Stake	Bet Odds	Bookmaker	W/L/P	Profit/Loss
602	E1	9/1/2009	Reading	Watford	A	10	7.00	Bet365	P	0.00
603	E0	10/1/2009	Arsenal	Bolton	A	10	11.00	Bet365	P	0.00
604	E0	10/1/2009	Newcastle	West Ham	H	10	2.20	Bet365	P	0.00
605	E1	10/1/2009	Blackpool	Birmingham	H	10	3.80	Bet365	P	0.00
606	E1	10/1/2009	Doncaster	Bristol City	H	10	2.50	VC Bet	P	0.00
607	E1	10/1/2009	Ipswich	Sheffield Weds	A	10	4.80	VC Bet	P	0.00
608	E1	10/1/2009	Plymouth	Crystal Palace	D	10	3.30	Bet365	P	0.00
609	E2	10/1/2009	Cheltenham	Millwall	H	10	3.20	VC Bet	P	0.00
610	E2	10/1/2009	Leeds	Carlisle	A	10	6.00	Bet365	P	0.00
611	E2	10/1/2009	Oldham	Hartlepool	A	10	5.00	Bet365	P	0.00
612	E2	10/1/2009	Southend	Crewe	H	10	1.80	Ladbrokes	P	0.00
613	E2	10/1/2009	Walsall	Brighton	A	10	3.75	Ladbrokes	P	0.00
614	E3	10/1/2009	Accrington	Darlington	A	10	2.40	Ladbrokes	P	0.00
615	E3	10/1/2009	Bury	Barnet	A	10	5.00	Bet365	P	0.00
616	E3	10/1/2009	Chesterfield	Rochdale	H	10	2.60	Bet365	P	0.00
617	E3	10/1/2009	Dag and Red	Wycombe	D	10	3.40	Bet365	P	0.00
618	E3	10/1/2009	Grimsby	Morecambe	H	10	2.40	Gamebookers	P	0.00
619	E3	10/1/2009	Shrewsbury	Chester	A	10	6.00	Bet365	P	0.00
620	E0	11/1/2009	Man United	Chelsea	A	10	3.50	Bet365	P	0.00
621	E0	11/1/2009	Wigan	Tottenham	A	10	3.00	Blue Square	P	0.00

Fingers firmly crossed as usual :hope

[Rushian]

Re: Statistical Football Prediction Model 2008 No surprises in the Friday match...

No	Div	Date	Home Team	Away Team	Bet	Stake	Bet Odds	Bookmaker	W/L/P	Profit/Loss
602	E1	9/1/2009	Reading	Watford	A	10	7.00	Bet365	L	-10.00
603	E0	10/1/2009	Arsenal	Bolton	A	10	11.00	Bet365	P	0.00
604	E0	10/1/2009	Newcastle	West Ham	H	10	2.20	Bet365	P	0.00
605	E1	10/1/2009	Blackpool	Birmingham	H	10	3.80	Bet365	P	0.00
606	E1	10/1/2009	Doncaster	Bristol City	H	10	2.50	VC Bet	P	0.00
607	E1	10/1/2009	Ipswich	Sheffield Weds	A	10	4.80	VC Bet	P	0.00
608	E1	10/1/2009	Plymouth	Crystal Palace	D	10	3.30	Bet365	P	0.00
609	E2	10/1/2009	Cheltenham	Millwall	H	10	3.20	VC Bet	P	0.00
610	E2	10/1/2009	Leeds	Carlisle	A	10	6.00	Bet365	P	0.00
611	E2	10/1/2009	Oldham	Hartlepool	A	10	5.00	Bet365	P	0.00
612	E2	10/1/2009	Southend	Crewe	H	10	1.80	Ladbrokes	P	0.00
613	E2	10/1/2009	Walsall	Brighton	A	10	3.75	Ladbrokes	P	0.00
614	E3	10/1/2009	Accrington	Darlington	A	10	2.40	Ladbrokes	P	0.00
615	E3	10/1/2009	Bury	Barnet	A	10	5.00	Bet365	P	0.00
616	E3	10/1/2009	Chesterfield	Rochdale	H	10	2.60	Bet365	P	0.00
617	E3	10/1/2009	Dag and Red	Wycombe	D	10	3.40	Bet365	P	0.00
618	E3	10/1/2009	Grimsby	Morecambe	H	10	2.40	Gamebookers	P	0.00
619	E3	10/1/2009	Shrewsbury	Chester	A	10	6.00	Bet365	P	0.00
620	E0	11/1/2009	Man United	Chelsea	A	10	3.50	Bet365	P	0.00
621	E0	11/1/2009	Wigan	Tottenham	A	10	3.00	Blue Square	P	0.00

Still 19 more picks to go! :hope

[Rushian]

[Rushian]

[Rushian]

Re: Statistical Football Prediction Model 2008 One pick for tonight (based on the odds of the originally scheduled match...)

No	Div	Date	Home Team	Away Team	Bet	Stake	Bet Odds	Bookmaker	W/L/P	Profit/Loss
610	E2	12/1/2009	Oldham	Hartlepool	A	10	5.00	Bet365	P	0.00

It's quite a longshot but ... fingers crossed :hope

[Rushian]

[Rushian]

Re: Statistical Football Prediction Model 2008 After a series of (unsuccessful) away picks, the system churns out a Home selection:

No	Div	Date	Home Team	Away Team	Bet	Stake	Bet Odds	Bookmaker	W/L/P	Profit/Loss
611	E2	13/1/2009	Yeovil	Scunthorpe	H	10	2.90	Blue Square	P	0.00

Let's hope for a return to winning ways! :hope

[Rushian]

[Rushian]

Re: Statistical Football Prediction Model 2008 Lots of picks to choose from for the weekend...

No	Div	Date	Home Team	Away Team	Bet	Stake	Bet Odds	Bookmaker	W/L/P	Profit/Loss
612	E2	16/1/2009	Hartlepool	Northampton	A	10	3.75	Blue Square	P	0.00
613	E0	17/1/2009	Blackburn	Newcastle	H	10	2.15	Blue Square	P	0.00
614	E0	17/1/2009	Chelsea	Stoke	H	10	1.25	Blue Square	P	0.00
615	E0	17/1/2009	Man City	Wigan	A	10	4.00	William Hill	P	0.00
616	E0	17/1/2009	Sunderland	Aston Villa	H	10	3.30	Ladbrokes	P	0.00
617	E0	17/1/2009	West Brom	Middlesbrough	H	10	2.40	Interwetten	P	0.00
618	E1	17/1/2009	Birmingham	Cardiff	H	10	2.00	Bet365	P	0.00
619	E1	17/1/2009	Bristol City	Wolves	H	10	3.20	Ladbrokes	P	0.00
620	E1	17/1/2009	Derby	QPR	H	10	2.38	William Hill	P	0.00
621	E1	17/1/2009	Norwich	Barnsley	H	10	2.20	Bwin	P	0.00
622	E1	17/1/2009	Nott'm Forest	Plymouth	D	10	3.40	VC Bet	P	0.00
623	E1	17/1/2009	Preston	Burnley	H	10	2.10	Bwin	P	0.00
624	E1	17/1/2009	Southampton	Doncaster	H	10	2.30	Bet365	P	0.00
625	E1	17/1/2009	Swansea	Reading	A	10	2.50	Ladbrokes	P	0.00
626	E1	17/1/2009	Watford	Sheffield United	H	10	3.00	Ladbrokes	P	0.00
627	E2	17/1/2009	Brighton	Leeds	D	10	3.40	Ladbrokes	P	0.00
628	E2	17/1/2009	Colchester	Cheltenham	H	10	1.73	Gamebookers	P	0.00
629	E2	17/1/2009	Leyton Orient	Bristol Rvs	H	10	2.40	Bet365	P	0.00
630	E2	17/1/2009	Millwall	Tranmere	A	10	4.00	Bet365	P	0.00
631	E3	17/1/2009	Bradford	Accrington	A	10	6.00	Ladbrokes	P	0.00
632	E3	17/1/2009	Brentford	Notts County	A	10	6.00	Ladbrokes	P	0.00
633	E3	17/1/2009	Chester	Chesterfield	A	10	2.75	Ladbrokes	P	0.00
634	E3	17/1/2009	Exeter	Bury	A	10	3.00	Ladbrokes	P	0.00
635	E3	17/1/2009	Morecambe	Gillingham	H	10	2.62	Blue Square	P	0.00
636	E3	17/1/2009	Rochdale	Lincoln	A	10	4.00	Bet365	P	0.00
637	E0	18/1/2009	Tottenham	Portsmouth	A	10	5.50	VC Bet	P	0.00
638	E0	18/1/2009	West Ham	Fulham	A	10	3.60	Bet365	P	0.00
639	E0	19/1/2009	Liverpool	Everton	A	10	6.50	William Hill	P	0.00

Fingers crossed :hope

[Rushian]

[Rushian]

[Rushian]

Re: Statistical Football Prediction Model 2008 The Merseyside derby ended in a draw, which spells further loss to the system... Whene will this bad spell end? Hopefully tonight!

No	Div	Date	Home Team	Away Team	Bet	Stake	Bet Odds	Bookmaker	W/L/P	Profit/Loss
639	E0	19/1/2009	Liverpool	Everton	A	10	6.50	William Hill	L	-10.00
640	E2	20/1/2009	Peterboro	Milton Keynes Dons	A	10	3.20	Blue Square	P	0.00
641	E3	20/1/2009	Accrington	Darlington	H	10	3.30	William Hill	P	0.00
642	E3	20/1/2009	Dag and Red	Exeter	H	10	2.00	VC Bet	P	0.00
643	E3	20/1/2009	Port Vale	Gillingham	H	10	2.50	Bet365	P	0.00

Fingers crossed as always :hope:hope:hope

[Rushian]