I've been working on a couple different models. Having done some backtesting the results weren't great (link). However that was against closingish lines (recorded Friday) and the model doesn't take injuries into account. I'm still working on it but I'm hoping that taking opening lines and considering injuries myself I can do better. In other words I'm not betting blind off the model but using as a jumping off point.

Right now I only have money on Bookmaker (GTFO NOOB IMO) a situation I'll be fixing shortly now that I've read through the FAQ and several posts here and have a clue. I'll be posting results and perhaps more importantly Friday nights I'll look at Pinnacle and see what percentage of the time I got a better line than what they have.

Here are my picks so far this week. All odds are decimal since most soccer bettors are probably Euros. My pick is bold:

Result	Odds	Model says
o3 Real Madrid - Getafe	1.69	1.36
Xerez - Sevilla	1.6	1.27
o2.5 Athletic - Atletico	1.75	1.46
Espanyol - Valladolid	3.85	3.15
Mallorca - Racing	1.8	1.26
draw Villarreal - Tenerife	3.4	2.84
Malaga - Valencia	1.95	1.58

I hope this a joke with Model says on OVER 3 in Real game 1.36 , Sevilla 1.27 ?? Mallorca 1.26 ???? Does Model says mean thats what true odds are supposed to be going by your system ?
Good luck with system ...
PS : never in life you are going to see odds of under 3.00 on a draw unless its a tournament match and outcome doesnt affect teams placement .

I'm not going straight off the model. Since we're only 8 matches in it can easily be thrown off due to a couple of huge results.

Xerez are extremely weak offensively - the worst in the liga. Sevilla are second best defensively. That's true looking at just the league and what the model says. I think the model definitely does underrate Xerez defensively due to a couple huge results against them (5-0 at Real Madrid and 0-3 versus Depor). Given Sevilla's results I don't see any that would tilt things in either direction. Significantly improving Xerez's defense puts Sevilla at 1.45 making that bettable.

Real Madrid is tougher because of Ronaldo still being out. Like I said the model doesn't take that into account. Getafe are about an average defensive side in the league - the model has them 11th. Looking at their results I think that's about right. They have a 3-0 defeat but nothing else that's huge that could sway the model. On the other hand Real Madrid have several big wins so I adjusted their attacking score downward. That put o3 at 1.5 making it bettable. To be honest while I think the model is off I don't think it's as far off as you imply especially if they had Ronaldo (or did for 3/4 of the time or whatever).

edit: just saw your edit. Yeah I agree actually. I didn't look at that one closely enough. I ran it through another model which I think more accurately predicts draws at it put it at 3.34 which is still value but is more thin than I'd like to bet on that kind of longshot with flat betting. It might well be -EG.

Interesting to me would be how many of those lines are bet heavily or faded heavily. For example Valladolid sticks out to me since I had them as a bet at 4.4, and the eventual price settled at about 3.7 or even lower. Not sure about the overs and unders as i don't bet on them. Mallorca was certainly weakish and available around 2.03. Not sure about Valencia as I recall they were pretty hot and bet in from a -0.25 line on the asian handicaps to a -0.5 line.

I'm working on tweaking the model to add stats other than just goals. So far the results are promising but I'll know more tomorrow when I run it through some backtesting. I'm also adding a couple books which is surely more important anyway.

I added a couple premiership bets and some over/unders. I'm not confident at all in the over/unders so I'm laying off those for a while.

Here were all the results:
Saturday:
Arsenal 1.57 - win
Aston Villa 2.65 - loss (this line was a trap I guess - looked ridiculously +EV to me)
Sevilla 1.6 - win
Real Madrid - Getafe o3 1.69 - loss
Osasuna - Barcelona u3 1.77 - win
Xerez - Sevilla u2.5 1.77 - win
Athletic - Atletico o2.5 1.74 - loss
not including: Oregon - USC moneyline. Ship the 2.3!

Sunday:
Valladolid 3.85 - loss
Mallorca 1.8 - win
Villarreal - Tenerife draw 3.4 - loss (only 5 goals off IMO)
Valencia 1.95 - win
Mallorca - Racing o2.5 1.77 - loss
Villarreal - Tenerife u2.5 - loss

Total winnings: -2.54 units

Yeah meant to check the lines before the games but a combination of forgetting and then going to the Oregon (American) football game got in the way. That's surely the best way to judge.

Something I've noticed is that the lines don't seem to move much at all in the midweek. Is that your experience?

I find they move initially, if there are any quick errors or early team news, then settle down until the matchday, where they pick up again, unless there's any really major news at work.

I had some of those picks as value (bilbao overs, certain others) but strongly disagreed with the villa pick, everton at 2.75 were one of the bets of last weekend IMO.

Jared. I am not being rude but you are talking gibberish. I hope this helps.

I was a tad hasty, apologies. I believe that using your model you will lose consistently over time by a small margin. You will know this because the sites that you bet on will allow you to bet with them. When they start to restrict you, you will know you are taking value bets. I don't believe there is any value whatsoever on Premiership matches (1X2 not AHCSs) and if bets are struck after the price moves then certainly no chance. You may run hot but you won't be making +ev bets especially with those margins. Even the 5% on betfair will hold you back on the Premiership for the most part.

I would like to discuss some in-running modelling with you if you have the time though...

Jared, if the lines your model produces is consistently largely off actual lines, then it is much much more likely your model is at fault.

If your model does bad in back testing, go back to the drawing board.

you may also want to track how your "big" plays do vs your borderline plays. if you find you are doing much better on your multiunit plays and not on borderline, then adjust accordingly.

Thanks for the feedback guys. I appreciate all whether it's positive or negative. Right now I'm working on adding some stuff. I've gotten good results in preliminary backtesting but the sample is small.

As far as getting cut off, right now I'm betting just 5 bucks a game so I don't think that'll be a problem Hopefully it becomes a problem due to me gaining confidence in the (new) model so I'm betting real money. For now since I'm just getting started I'm keeping it small.

You have a model so you can measure your edge. So you should be betting Kelly or at least fractional Kelly, no? You can still keep it small and bet Kelly.

The model I've been using up until this point was a small loser in testing. I thought that was due to not taking injuries into account, so I was just using it as a jumping off point. I'm less sure of that now, which is a big reason for just betting a flat $5.

The one I'm working on now seems a lot better and actually a decent winner, but we'll see if that was just the small sample size. I'll keep Kelly betting in mind for that, though I'm just using flat betting in testing to reduce variance and therefore be less likely to fool myself into thinking it's a winner.

What % of losses are attributable to vig? i.e, if you adjust the lines to be vigless (there are calculators online that will do this afaik), how would your results look?

Jared, I don't know anything about modelling myself, but a bit of advice......

Bet (or backtest if possible) the model against the early market. If you're onto something, you should beat the closer. Much easier to pick off weak spots in the opener. It might seem irrelevant since the model numbers are so different to the market, but if they're different in the right direction, betting early should get you the price you need to have an edge.

If you can convert your numbers to asian lines, you're betting into much tighter markets unless you got an exchange with commission discount.

A good site for a more efficient number than pinnacle is sportmarket.com. This is the line straight out of Asia, a feed from ****** (the pinnacle of soccer) that's not readily available apart from here. Though I would not bet here due to having to pay commision, the odds are displayed in a great format so it beats using the SBO website to monitor the asian line.

Some very useful posts on this thread. The points about being allowed to bet with pinn/sbo etc are not true though IMO, you are "allowed" to be a winner there. Just because you are taking bets at pinn doesn't automatically make them -EV.

The post above about comparing to opening lines and then again to closing lines is spot on. You would be looking to be moving in the right direction 60%+ of times of openers vs closers and that would give me, if I were you, a lot more confidence in any model.

SBO, IBC 1***** you can bet as much as you like. These are the best bookmakers in the world. They are the quickest to change their prices and therefore it is tough to take advantage of any mistakes. In fact these mistakes do not always get detected by 'models'. You should have a good enough knowledge to take advantage of early lines, early team news etc.

However, sportingbet, bet365, stan james who bet on strange games (where your edge can be higher) will restrict you over time. They take a larger vig (106-107%) on most matches to compensate for their inability to price up/adjust efficiently.

I would challenge you to beat ****** over 100 plays with the overound using simply a model.

Using the early market its a lot easier than the final market. Having said that 100 plays would not prove anything sufficiently, easy enough to make 100 +ev bets and not finish in front. Also at their margins also possible to make 100 -ev bets and finish in front!

The crucial point is, as above, if you are outside the closing lines on the correct side on more than 60% of your plays, you've got a very decent model working for you. From the initial posts in the thread I'd be alarmed that things are available at 1.66 that I make 1.27 or whatever - that's clearly not correct - but all good models start somewhere.

^^ Yep, with the asians going main stream its almost made pinny redundant for soccer. SBO etc very useful for US sport too. I have been limited at 1***** but I think they might just be struggling. The big players (SBO, IBC, a few others) take massive bets and very efficient, but I'd take your 100 bet challenge on the early markets, maybe variance would get me but I'd beat the closer often enough to overcome the small spread and have a small (2%ish) edge.

It would be interesting to know what is more heavily bet, the NFL or the EPL. Cetainly more money is bet on soccer than grid iron due the sheer number of games across leagues and longer seasons, and lots of nice volatile markets like the Europa league tonight.

I'd also guess that more money goes down on soccer bets from Asia than Europe / UK. Living in Asia I know that everyone is betting and soccer mad, and the asian line is available everywhere from the swankiest Hong Kong cocktail bar to the dingy street rice whisky bar in rural Thailand.

I think the EPL is bet more for sure, esp by Asia, syndicates, etc. The AHCs offered are so much more attractive than the nfl's spreads. I believe there is money to me made on early prices with or without a model and in-running reacting to certain events in a match, not just in football.

Anyways, gl with the model, hope it works out for you.

Superbowl or World Cup final?

How fast is that XML feed from sportsmarket.com if you use it keemao?

It refreshes every minute or you can refresh manually. It's pretty spot on, not right down to the second though. Not sure about in run, I only look at pre-game. Let me know if you know of anything better, pay or free pls.

Will do - also when I get some time I'll look at how it works in running for speed.