Pricing - time decay

Football, Soccer - whatever you call it. It is the beautiful game.
Post Reply
ColdWinner
Posts: 3
Joined: Sun Mar 31, 2019 7:14 am

Say you back the draw. In theory after 30 minutes with no goal scored, your bet should have risen in value by 33% 30/90. However anyone who knows football is well aware that it could take more than minutes for say Man City to go 1-0 up against Cardiff. The likelihood of the draw doesn't necessary increase just because the first half hour the score is still 0-0.

So rather than pricing a draw on a minute by minute basis. Is anyone aware of a way to price a draw giving more weight to the back end of a game. For instance if its still 0-0 after 60 minutes, then yes the value of your bet should get a bump of roughly 60% - perhaps even a bit lower.
User avatar
firlandsfarm
Posts: 2688
Joined: Sat May 03, 2014 8:20 am

I've often wondered about this time decay thing (and the reverse, whatever that is called for each team's price inflation until they score a goal). I remember asking a keen football fan about in-play and prices early on asking "how would you think of a game while watching it as to what the outcome might be?" and his reply was that unless something exceptional was happening he would probably wait and see what the score was at half time. Is there any evidence to support that a team being a goal down after say 20 mins. is more likely to draw level by the final whistle than a team that went a goal down after say 50 mins.? Yes there is more time for them to equalise but there is also more time for the other team to go 2 up! I can understand the theory but is there the evidence. I like to play with the stats on such issues but where can I get a full database of goal scoring times?
User avatar
ShaunWhite
Posts: 9731
Joined: Sat Sep 03, 2016 3:42 am

ColdWinner wrote:
Sun Mar 31, 2019 7:33 am
Say you back the draw. In theory after 30 minutes with no goal scored, your bet should have risen in value by 33% 30/90.
Price over time isn't a straight line for any given match. Lots of cleverer people than me have looked at this, and many factors come into play.
There's some interesting acedemic papers on the subject if you enjoy a bit of maths, but also 1000s of half baked theories on gambling sites so choose your sources carefully. imo, mispricings only exist briefly and without a mispricing, finding a value trade is tricky (impossible?).
dm1900
Posts: 71
Joined: Sun Jan 15, 2017 10:02 pm

It is very easy to estimate the decay using the exponential decay equation.
spreadbetting
Posts: 3140
Joined: Sun Jan 31, 2010 8:06 pm

Maybe have a look at The Power of Goals site if you're interested in automating your expected odds decay

http://thepowerofgoals.blogspot.com/201 ... t-ten.html
The rate at which a team's goal expectancy declines isn't constant. More goals are scored on average in the second half than the first as teams become more urgent in their efforts to score and fatigue leads to more space. The rates are around 44% for the former and 56% for the latter and the decay can be adequately described by an exponential equation of the following form.

Remaining Goal Expectancy = Initial Goal Expectancy x (Proportion of Time Remaining) ^0.84
User avatar
firlandsfarm
Posts: 2688
Joined: Sat May 03, 2014 8:20 am

spreadbetting wrote:
Thu Jun 20, 2019 5:35 pm
Remaining Goal Expectancy = Initial Goal Expectancy x (Proportion of Time Remaining) ^0.84
Understand and agree but why 0.84?
spreadbetting
Posts: 3140
Joined: Sun Jan 31, 2010 8:06 pm

Because it's an exponential decay rather than constant ,I'd guess the ^0.84 has been derived from actual statistical data.

If you look at actual Betfair data compared to the modelled odds they're very close so it does seem the majority of automated systems are using their own adjusted versions of Poisson distributions to model their odds. But the only saying of crap in crap out will always trip up most people trying to predict odds.

Unless you're very clued up it's a lot easier to pounce when odds veer out of line with expected values and wait for them to return to expected values.
User avatar
firlandsfarm
Posts: 2688
Joined: Sat May 03, 2014 8:20 am

spreadbetting wrote:
Fri Jun 21, 2019 9:49 pm
Because it's an exponential decay rather than constant ,I'd guess the ^0.84 has been derived from actual statistical data.

If you look at actual Betfair data compared to the modelled odds they're very close so it does seem the majority of automated systems are using their own adjusted versions of Poisson distributions to model their odds. But the only saying of crap in crap out will always trip up most people trying to predict odds.

Unless you're very clued up it's a lot easier to pounce when odds veer out of line with expected values and wait for them to return to expected values.
it was the source of the 0.84 I was querying, I got the exponential bit :) So a stats best fit (guess) but with no real/published foundation, like the querty keyboard! But what's driving the Betfair data … the actual probability or the automated bots? If enough people (bots) follow a set odds pattern the actual odds pattern will move to them. That gives two opportunities, one as you say when the odds veer from the expectation and the second when the expectation is over/under real probability because of trading distortion. A trading version of tell a lie often enough and people will believe it. I'm not doubting any of it, I just like to see the evidence.
spreadbetting
Posts: 3140
Joined: Sun Jan 31, 2010 8:06 pm

I imagine that ties in with the fact that that around 46% of goals are scored in the first half and 54% in the second half (stats) so we adjust 46/54 = 0.85 exponentially. If you look at the various tweaked poisson predictions they tie in with the actual Betfair data very closely. Bet Angel have their own soccer odds prediction dunno if that's stats driven rather than poisson and no idea how closely that follows the actual data either, but there must be a few users who bet from the data.

Image
Image
Image

http://thepowerofgoals.blogspot.com/201 ... t-ten.html
StellaBot
Posts: 818
Joined: Thu Jan 26, 2017 11:52 am

Interesting reading.
I have been building and testing(using real money) a football system
It uses a massive spreadsheet I constructed
Using non UK leagues at moment for the past 6-8 weeks

Main problems I have encountered are
(1)deciding what stake sizes to use for each price
Im guessing lets say 75% strike rate
All prices are odds on
If over evens it would be easy obviously
(2) When and if to "cash out"
(3) ?Spreading risk on other market/s on same game
(4) Cant backtest(at moment) for example previous seasons data(for many logical reasons) as the spreadsheet is continually updating the present teams in league

If anyone could help ,please answer here or pm me

EDIT
I have factored in many scenarios into spreadsheet
eg home, draw,away,correct score,over/unders
User avatar
firlandsfarm
Posts: 2688
Joined: Sat May 03, 2014 8:20 am

spreadbetting wrote:
Sat Jun 22, 2019 2:07 pm
If you look at the various tweaked poisson predictions they tie in with the actual Betfair data very closely.
Thanks for the sample graphs, they do fit well showing a correlation but my thoughts are more to the price v's the outcome? Need lots of goal times data but would be interested to see how closely the inplay odds match a perfect market expectation.
Post Reply

Return to “Football trading”