polpolpol

Cycling time: £2 e/w on Sagan to win today's stage (15-2) £1 e/w on Kenny van Hummel (50-1)

J. Degenkolb in the Criterium of Dauphine prologue @ 66-1. 4th by 6 seconds last year, the course will suit him a bit more. The hairpin turns mean his acceleration will help. For the whole thing, Wiggins looks short at odds on; it would be a strange if he tried to hold his condition from Paris-Nice through Romandy and the Dauphine to the Tour. On the other hand, there is a long time trial and not too many mountaintops. With the little ramps and descending finishes, I'm going for Nibali (18) and Sanchez (80), and a few long shots e/w: Weening (200) who started the season late and was 10th in California, and the Lotto pair Vandendert (300) and Van Den Broeck (25) who can climb.

In tonight's semi I have Greece, Denmark to win / top3 Hungary top 3 and Montenegro, Austria to qualify Not too sure about the last two though.

Ah well, the streak is over. Philipstown in the next, back on the horse.

Next Random horse = Cheddar Gorge edit With Sunny Future for 2nd

Shit the bed - http://img195.imageshack.us/img195/6867/betss2.jpg

Its like that 12,000-1 Italian football bet (many pages ago) that I had 1p on. Always wiser to bet more when you win.

Red Cadeaux - Be Fabulous Exacta coming up.

Berry, Market, its the same Anyway, embarrassingly small stakes and it too me an age to upload. No trebles, just two quids. http://img692.imageshack.us/img692/7996/betswc.th.jpg

Just had Sir Patrick Moore (40) and Easy Terms (14) come in, going for Momentary in the next at Newmarket.

Hushovd and Felliu e/w in today's stage of the Giro (@ c. 100-1) - depends on how Cavendish is recovering. They've been around the middle of the top 10 so far.

Good call.

I got someone with the programming skills to code it for me, I just told him how to weight form, home / away, context and some other stuff. It basically takes every game from the season so far and comes up with correct score possibilities, but the more recent games add more to the count, so it has form weighting. How did you decide on the weighting to apply? And do you feel as if it works? Often it comes up with the same odds as the bookies - when it doesn't, sometimes it can be right - spotting form or under rated teams - but equally, sometimes when it's off it is because there are other factors beyond the pure maths and it works to bet against it. The main problems are that (1) it can't simulate when a team is huge favourite, even with weighting for teams being 'top4' or whatever. (2) It can't understand how underdogs are often happy with a draw. (3) It underrates the frequency of 0-0 and overrates 3-0, 0-0 is basically a hard scoreline to break, so nils should cluster with other nils, but the distribution can't do it. I use it more for correct score accys than win - loss. It's a tool, but judgement is much more useful. The dream was to scrape the data directly from betfair and have it as a bot, but it proved too much work. So have you analysed the occasions on which your system disagrees with the bookies to see who's right? And if I'm understanding it right, and your data pool builds over the course of the season, it should get more accurate over time. Do you find that? I did a bit of analysis, but I had to do it manually and it was a real grind to get all of those different bits of data into a table. If the bookies price and the price I have is significantly different - more than 50%, or even 100%, it's worth looking at usually - but I don't always go for the maths. If there's injuries, players rested, or - at this time of year - teams with nothing to play for, sometimes instinct is better. It certainly gets better over time. Up to Christmas, it isn't so good.

I got someone with the programming skills to code it for me, I just told him how to weight form, home / away, context and some other stuff. It basically takes every game from the season so far and comes up with correct score possibilities, but the more recent games add more to the count, so it has form weighting. How did you decide on the weighting to apply? And do you feel as if it works? Often it comes up with the same odds as the bookies - when it doesn't, sometimes it can be right - spotting form or under rated teams - but equally, sometimes when it's off it is because there are other factors beyond the pure maths and it works to bet against it. The main problems are that (1) it can't simulate when a team is huge favourite, even with weighting for teams being 'top4' or whatever. (2) It can't understand how underdogs are often happy with a draw. (3) It underrates the frequency of 0-0 and overrates 3-0, 0-0 is basically a hard scoreline to break, so nils should cluster with other nils, but the distribution can't do it. I use it more for correct score accys than win - loss. It's a tool, but judgement is much more useful. The dream was to scrape the data directly from betfair and have it as a bot, but it proved too much work.

I got someone with the programming skills to code it for me, I just told him how to weight form, home / away, context and some other stuff. It basically takes every game from the season so far and comes up with correct score possibilities, but the more recent games add more to the count, so it has form weighting.

I'd agree that the context of the match matters, but since both teams have a lot to play I'd assume that will cancel itself out. I'll definitely be going in large on County tomorrow, along with Arsenal and Sheff U My own version of the 'sheet' (right window) suggests the bookies (well, betfair, left) are about right. http://img716.imageshack.us/img716/1543/wycnottsc.jpg

Romain Bardet for the Fleche Wallon tomorrow @ 150-1. He looked really strong in Amstel, if he can keep that form he can go close. Van Den Broeck @ 80 as well, if he can stay on his damn bike.

Valverde to win the Amstel Gold Race, Van Den Broeck to place. Both covered if Gilbert gets it back.

The assumption that Ba can keep up his outstanding form, and play every game of the season like a super-elite striker seems quite a big one (especially considering his knee and his release clause). Even with that assumption though, what you prove is that we could play awful football, lose heavily against the 'top 6' and still finish upper-mid table. The Stoke model. The replacement of Hughton with 'Percentage' Pardew shows that the risk averse board likes this compromise, but 1) it's horrible to watch at times; 2) the border of the European places is the ceiling of its effectiveness; and 3) there's little scope for evolving the strategy into something more progressive.

Peter Sagan to win Mailan San Remo. He's the fastest of the riders who can definitely get over the hills. Pozzato to place. Hausler and Breschel to place now they're recovering. A bit on Hoogerland and Boom so I can hopefully cheer on a break, and Degenkolb if he gets out to 100-ish. I'm also trying to back Cavendish at about 5 for the win, and laying him at about 2.5 for the top 3. I can't see him being 2 or 3; it's either the win, or not being able to stay with the front group.

On form, Gladbach to beat Kaiserslauten @2.5 has to be the bet of the day. And Gladbach love to score 3; I'm going for 0-3 and 1-3: "Lautern are winless in 11 BL games (seven draws, four defeats) - currently the longest run of its kind in the top flight. Their last win came over four months ago when they beat Freiburg 1-0. Borussia have only lost one of their last 11 BL games (eight wins, two draws) and are unbeaten in their last five (four wins)."

The Strongest - Santos game in the Copa Libertadores (watching it on the betfair stream) is fantastic. Totally end to end, 1-1 after 30 mins. I assume it'll get more crazy as the altitude breaks Santos.

£5 on Vitesse @ 30-1 at h/t In their game against Feyenoord. They were looking good in the first half, unlucky to be 2-1 down.

I did all of the Serie A games today as 5, 4, and 3 folds at 1p, 1p and 3p, so the total bet was about £3.50. With hindsight, I wish it was a bit more.

Had a theory that the cold in Italy might lead to some low scores: Landed this at 12,00-1 . . . . with a penny stake! http://img194.imageshack.us/img194/1340/onepo.jpg Should have splashed out with 5p