HT/FT Odds

[MattR]

This may be one for Mat.,

I'm not entirely sure but I believe these are based on match odds and I guess half time match odds.

I got to thinking that the 'vagueness' of that must leave it an area that could potentially be exploited.

The question is how would you go about working out theoretical odds for the various outcomes. I've used goals scored for home and away teams to a degree of success on the half time match odds markets before so that would be a foundation for the HDA frst half part. From there comes the hard part calculation wise I think.

IE if Stoke are 1-0 up v Liverpool at HT, you would need a calculation that gave the chance of Liverpool winning a half against Stoke by 2 or more goals. Similarly for a 2 goal lead, you'd need Liverpool's probability of winning by 3 goals in a half, and so on and so on.

So in theory (and this is where your expertise comes in Mat :D ) If you were to use the probability of scores for a half of football (using half the ave goals from the original formula from goals scored home v away, for and against) then I guess you could add the necessary scores to approximate all outcomes. Ie get the half time 'result' from the ht score probabilities and then add all the half scores that would make that outcome become HA HD HH etc etc.

[mathare]

Ooh, this is the sort of area I have thought about getting into in the past and never found the time as there are a lot of variables to consider.

In your Stoke 1-0 Liverpool at HT example how do you work out the probability of Liverpool scoring at least two more than Stoke in the second half. You need to work out the probability of Stoke scoring 0, 1, 2, 3 goals etc and thus Liverpool scoring 2, 3, 4 etc. But what data do you use as the basis for this?

You could use:

• Liverpool's goalscoring frequency away from home
• Frequency of Stoke conceding at home
• Combination of above (straight average)
• Weighted average of above
• Above data based on 90 minutes or just second half
• etc

But suppose you somehow get as far as knowing that Liverpool are expected to score 1.3 goals in the second half, what use is that? You can then use Poisson probabilities to work out the likelihood of them scoring 2 or more, 3 or more etc.

As with anything though it's a model and thus contains assumptions and inherent inaccuracies. But it's a good starting point

[MattR]

What I have used to some success although I'm still trying to accumulate data over time to see if there are any trends is this:


HT (Home team)
AT (Away team)
GF (goals for)
GA (Goals against)

HT GF + AT GA / ( (HT home games played + AT away games played)/ 2)

to give a Home team goal total/potential call it what you will

then opposite for away team ie AT GF + HT GA etc

then these two totals are used for the poisson distribution of possible scores to give odds of any particular score happening.

I've had a bit of success dutching the four most likely scores together but never really tested it at any length. I've also used it to generate match odds by adding each potential score up ie home/draw/away

Then I tried halving the original team goal totals and using the poisson on scores again to get likely half time scores and also via that a half time match odds prediction. This again seems fairly good although again with limited testing.

But what I am thinking is if I have the lprobability of scores per half then perhaps by adding the permutations you can come up with odds for HH, HD etc. Ideally I imagine you should really break the original calculations up into the data for each half that the teams score their goals but that could become very long winded getting the data and would it be any more useful?

The hardest part would be getting the sheet set up I imagine as there are numerous possibilities to be added. IE Home half time 1-0 to become HH would have to add 1-0 one half to a 0-0, a 1-0,2-0,3-0,3-1 and so on. Then a 2-0 to 0-0,0-1,1-2,1-0 etc etc

[Street cry]

you could get soccer mystic that might help you

[mathare]

What I have used to some success although I'm still trying to accumulate data over time to see if there are any trends is this:

HT GF + AT GA / ( (HT home games played + AT away games played)/ 2)

to give a Home team goal total/potential call it what you will

then opposite for away team ie AT GF + HT GA etc

That's certainly a sensible way of doing it as it measures home attack against away defence and vice versa.

then these two totals are used for the poisson distribution of possible scores to give odds of any particular score happening.

You know about Poisson distributions, good. Were you aware that Excel can do all the hard work for you in calculating the probabilities?

Then I tried halving the original team goal totals and using the poisson on scores again to get likely half time scores and also via that a half time match odds prediction. This again seems fairly good although again with limited testing.

If you have the HT and FT scores you can take the FT score from the HT score to give two separate scorelines - first half and second half. You could then calculate your figures above based only on first half or second half data if you liked and see how this compares to simply halving the expected goal totals.

But what I am thinking is if I have the lprobability of scores per half then perhaps by adding the permutations you can come up with odds for HH, HD etc. Ideally I imagine you should really break the original calculations up into the data for each half that the teams score their goals but that could become very long winded getting the data and would it be any more useful?

Two ways of doing this really:
1) do it based on scores so calculate the probaility that the home team will be winning 1-0, 2-0, 2-1 etc at half-time and then calculate the second half score probability based on what you want the full-time result to be. That's the long (thorough?) way
2) check the HT results to see how often a team is winning at HT and FT, how often they go in ahead at HT but draw at FT etc.

The HT score data is all easily available online in Excel/CSV format

The hardest part would be getting the sheet set up I imagine as there are numerous possibilities to be added. IE Home half time 1-0 to become HH would have to add 1-0 one half to a 0-0, a 1-0,2-0,3-0,3-1 and so on. Then a 2-0 to 0-0,0-1,1-2,1-0 etc etc

This is where it all gets messy, yes. What's to stop a 1-0 HT result from becoming a 4-4 draw, for example? Nothing, so for each HT score possibility you need to calculate a raft of second half score possibilities too. You need to check all the combinations and there are going to be lots!

[MattR]

Yes I've got a sheet that I found some years back with poisson correct scores that I've adapted in various formats.


Just been playing around with what should be the simplest outcome to do, the DD. This would seem to be the least number of possible second half scores so the simplest to do first. The sheet goes up to 7-0/7-7/0-7 which is ample I think to cover all but the rarest of scores. Maybe even 6 would be enough. Anyway it's 7 for now as it's already there.

So as most games today are cup games and using it on euro with teams in different leagues is not really ideal I'll pick a game from yesterday to get the goal stats.

I'll use Grimsby Mat!

So we have Macclesfield v Grimsby (and I have remembered to remove last nights games from the league table) This could be quite a skewed one looking at it as Macclesfield have managed to lose 1-4 and 0-6 at home I see! anyway let's see how it pans out

this gives us a home and away team initial goals potential of:

1.56 to 1.39 I have seen stats and used the ones at football data that shows on average around 43% of goals are scored in the first half so could split it into two sets of poisson scores using 43% for 1st half and 57% for second half and that may be a good idea but for the moment I'll just split the match total in half.

So using 0.78 v 0.695 for a half's goals we come up with ht match odds of:

Home 2.94
Draw 2.70
Away 3.45

draw scores of
00 22.88%
11 12.40%
22 1.68%
33 0.10%
44 0.00%
55 0.00%
66 0.00%
77 0.00%

naturally the second half also has to be a draw so those 8 scores are the only possible second half scorelines that will produce a DD result.

This is where I'm getting confused on what calculation we'd need to do now. For the half I've used 1/ the total of the draws (37.07%) to get the 2.70 odds but what calculation is required to get the DD odds??

[mathare]

Nothing is ever as simple as it seems is it?

What you need to do is calculate the chance of the first half being a draw and multiply that by the chance of the second half being a draw. If the chance of the HT result being a draw is 37.07% then as you've halved the goals totals the second half must be the same chance as the first so another 37.07%. So if 37% of games are draws at HT and only 37% of second halves are a draw then the chance of a game being a draw at HT and FT is 37% x 37% = 13.74% or odds of around 7.28 for the double result.

I think.

[MattR]

I think.

That makes sense. I wonder what the actual odds were? I'm going to do it on a polish cup game that's on today as although not ideal being a cup tie at least the two teams are in the same division.


OK we have Cracovia Krakow v Wisla Krakow which apart from being a fairly impressive scrabble score I don't know an awful lot more about. As I say it is a cup match so who knows what strength teams they are putting out but if we base it on the league Crac are 15th and Wisla 4th

Potential half goals gives us

0.28 v 0.50

a drawing half % of 1.91 match odds 52.49% so DD predicted odds of 3.63

Betfair currently at 6.00

However Cracovia home record of 3 scored 6 against in 7 games and Wisla's 10 scored in 7 games suggests not too many goals in the game assuming proper teams are put out.

So, (in theory!) Sounds a good value bet at 6.00!


So to get the other ht/ft odds it's going to be a case of multiplying each first half home win score % (in case of HH, HD, HA) by all other possible scores that create the final result, then totalling them up and divide into 1?

[mathare]

I'm wondering if this can be simplified at all. Take the HD example - we know that at HT the home team is winning so during the second half we need the away team to 'win' by the same amount to force the draw. Can we easily work out the chances of the away team scoring 1 goal, 2 goals etc more than the home team during the second half? Do we need to?

The chances of a draw at full time is built up from the probabilities of both sides scoring the same number of goals throughout the game. So the chance of a 0-0 is the probability of the home side scoring 0 and the away side doing the same, based on Poisson distributions using the full match goal total figures. Similarly 1-1, 2-2 etc. Using those we can work out the chances of a draw at the final whistle.

What about a home win at HT? Then we use the first half goal totals and do similar to the above calculation but summing the probabilities of 1-0, 2-0, 2-1 etc. Or, and this has just occurred to me we work out the Poisson probability of a team scoring 1 goal or less, 2 goals or less etc - which Excel can do for us - and then use those. For example the combined probabilities of 2-0 and 2-1 are the same as multiplying the probability of the home team scoring 2 goals and the away team scoring 1 goal or less.

Am I confusing anyone else or just me?

[MattR]

Actually that makes sense Mat, narrowing it down into groups. So for instance the home team is 2 goals up we'd then just need the chance of the away team scoring 3 or more in a half and the home team not scoring for the HT result to become HH etc etc . At least I think that's what you mean :D



EDIT: I wonder if you could use a vlookup table for this

[mathare]

Actually that makes sense Mat, narrowing it down into groups. So for instance the home team is 2 goals up we'd then just need the chance of the away team scoring 3 or more in a half and the home team not scoring for the HT result to become HH etc etc . At least I think that's what you mean :D

That's exactly what I meant, yes. Thank you.

[MattR]

Actually i think the easiest way to break this down will be into

ht scores 1 more than at
2 more than etc and vice versa

then those can be applied to the ht score

rather than every score

[MattR]

I think I've cracked it (possibly!)


I've got a table of ht by 1, by 2 and so on, on to draw and then away team leading by 1 etc, with same across the top. Then those events that don't give the desired result ie 1-0 ht and 2 goals by away team in second half are not included in that particular line.


Just did it for HH, DD, AA and DD came out the same at 3.62 so it all looks correct

so far we have on betfair in black (106% market) , mine in blue

HH 10.5 8.8
HD 19.5
HA 28
DH 12.5
DD 5.5 3.62
DA 5.0
AH 55
AD 17.5
AA 2.96 3.61


Will add the rest as I get done with the table.

[MattR]

Ok the final figures were


HH 10.5 8.8
HD 19.5 28.6
HA 28 381
DH 12.5 12.04
DD 5.5 3.62
DA 5.0 6.01
AH 55 65.14
AD 17.5 28.62
AA 2.96 3.61


So probability wise, DD and HH were two perceived 'value' bets whilst DD and AA looked the most likely outcomes with DA a decent possibility too. This all of course depends on what sort of sides two Polish teams put out in the league cup!


Well the game finished about 20 mins ago and it was 0-0 so DD was infact the result!


I just ran the other Polish cup game that kicked off at the same time out of interest and that gave

HH 4.43
HD 20.85
HA 75.27
DH 7.82
DD 5.08
DA 8.40
AH 89.24
AD 20.85
AA 4.87


The actual result of this one was AA 0-1/1-2



I might run the champs league for the hell of it, not sure what to do goals wise though. Not really enough info in 4 group games so may add the group game record to their league record and do it that way. Not ideal by any means but I'll see what it throws up anyway!

[MattR]

These are the mostly like outcomes according to the data I've used. The data isn't really that reliable i don't think because I've had to use their own form in their own leagues which doesn't really match up well when you have the likes of Inter v Panathinaikos for example where Pana's home league stats make them appear much better than they are in comparison to Inter. But for the sake of curiosity I want to see how they fare. It will be a much better test to do on a set of league fixtures so will do some for the weekend and see how they go. For now though here are the champs lge for tonight.

Those in bold are under the actual current betfair price and technically value bets if the stats are reliable and also if the teams in question have anything to play for and play full teams of course!


Anorthosis v Werder HH 2.95
Atl Madrid v PSV HH 2.52
Brodeaux v Chelsea AA 3.64
Cluj v Roma HH 3.11
Inter v Panthinaikos DD 3.99
Liverpool v Marseille HH 3.71
Shakhter v Basel HH 3.94
Sp Lisbon v Barcelona AA 2.77



Below are the three most likely outcomes

Anorthosis v Werder HH DH AA
Atl Madrid v PSV HH DH DD
Brodeaux v Chelsea AA DD HH
Cluj v Roma HH DD DH
Inter v Panthinaikos DD HH AA
Liverpool v Marseille HH DD AA
Shakhter v Basel HH AA DD
Sp Lisbon v Barcelona AA DA DD

[MattR]

4/8 correct which is encouraging! 6/8 were correct for the half time result as well. Not going to read too much into it though at this stage until I've seen how it goes on league matches.


Anorthosis v Werder HH 2.95 result DD
Atl Madrid v PSV HH 2.52
Bordeaux v Chelsea AA 3.64 result AD
Cluj v Roma HH 3.11 result AA
Inter v Panthinaikos DD 3.99 result DA
Liverpool v Marseille HH 3.71
Shakhter v Basel HH 3.94
Sp Lisbon v Barcelona AA 2.77

[MattR]

Got chance to play around a bit more with this over the weekend and though it's a little bit more work getting the first half goals for/aga I've decided that's a better way to split it. The question now is what to do with the predicted odds against the real odds on offer. Do you take the lowest predicted odds and just bet, or the once with the most 'perceived' value price or or or or....

So I've decided to run two tests. The lowest price and the three best 'value' odds dutched together. Will be intermittent as it depends on having the time available. I've run tonights championship games so will see what comes from that.

Will post them shortly and see how they go

[MattR]

Test 1 - Lowest odds from predicted odds

Team (pred/actual)

Burnley HH 3.52/3.80
Coventry AA 3.53/5.90
Sheff Utd AA 3.72/4.70
Birmingham AA 4.32/4.00
Preston HH 3.84/3.00
Sheff Weds HH 2.28/4.40
Swansea HH 2.72/3.05
Wolves HH 2.11/2.54
Reading HH 2.62/2.32

Test 2 - Dutching three predicted odds that are under real odds - if there are more than three then I will take the three lowest, if there are less than 3 then will just take what there is as the bet


Burnley HH,DD,AA BF Dutched odds: 1.70
Charlton DA,AA 2 only BF Dutched odds: 3.50
Nottm F DD,DA,AA BF Dutched odds: 1.90
Plymouth HH 1 only BF odds: 6.40
Preston DH,DD,AA BF Dutched odds: 2.20
Sheff W HH,DH 2 only BF Dutched odds: 2.60
Swansea HH,AD,AA BF Dutched odds: 2.00
Wolves HH 1 only BF odds: 2.54
Reading DH,DA,AA BF Dutched odds: 3.30


Stand out odds (according to predicted odds v available odds)
Sheff Weds HH 2.28 actual odds 4.40
Sheff Utd AA 3.71 actual odds 4.70
Plymouth HH 4.47 actual odds 6.40
Blackpool AA 12.20 actual odds 17.50


Will assume a £10 stake on these tests, although only a test and not betting them myself. Although I may have a fun couple of pounds on the stand out ones.

[counterfeit]

I've always thought that the football markets were not quite right so it is with some relief and gratitude that some geniuses (might have known it would be this forum) have worked it out and proved the odds are wrong.

I guess the long term profit is in the biggest value bets (presumably in terms of % rather than price). There could be a magnificently expensive piece of software in this!

[mathare]

There could be a magnificently expensive piece of software in this!

Tell me about it!

[MattR]

As an added interest to this I'm listing below the HT match odds as they are produced on my sheet when I'm putting in the first half goals data.


Predicted odds

Home Team Home Odds/Draw Odds/Away Odds

Burnely 2.86---2.12---5.60
Charlton 3.72---2.68---2.79
Nottm F 4.45---2.22----3.08
Plymouth 3.34---2.50---3.34
Preston 3.20---1.93---5.93
Sheff W 2.24---2.327---8.76
Swansea 2.49---2.66---4.50
Wolves 1.96---2.67---8.74
Reading 2.29---2.43---6.62

From that we have

'Value Priced' Half Time Match Odds

Team Pred/Actual

Coventry (Away) 2.79---3.90
Plymouth (Home) 3.34---4.30
Preston (Draw) 1.93---2.24
Sheff W (Home) 2.24---3.25
Wolves (Home) 1.96---2.14


and for a bit of fun on these value ones as betfair have half time corr score markets

Ht Score pred/actual

Coventry
0-1 5.33---6.00
0-2 13.32---24.00
1-2 20.49---32.00
Any 13.44---24.00


Plymouth
1-0 5.80---6.20
2-0 17.00---25.00
2-1 24.00---34.00
Any 14.90---28.00

Preston
0-0 2.23---2.86

Sheff W
1-0 3.60---5.20
2-0 9.70---18.00
Any 23.48---28.00

Wolves
1-0 3.70---3.95
2-0 7.70---9.40
Any 13.39---14.50

[MattR]

Good start on these....:)

'Value Priced' Half Time Match Odds

Team Pred/Actual

Coventry (Away) 2.79---3.90 Won +29.00
Plymouth (Home) 3.34---4.30 Lost -10
Preston (Draw) 1.93---2.24 Won +12.40
Sheff W (Home) 2.24---3.25 Lost -10
Wolves (Home) 1.96---2.14 Won +11.40


Ht Score pred/actual

Coventry
0-1 5.33---6.00Won +50
0-2 13.32---24.00Lost -10
1-2 20.49---32.00Lost -10
Any 13.44---24.00Lost -10


Plymouth
1-0 5.80---6.20Lost -10
2-0 17.00---25.00Lost -10
2-1 24.00---34.00Lost -10
Any 14.90---28.00Lost -10

Preston
0-0 2.23---2.86Won 18.60

Sheff W
1-0 3.60---5.20Lost -10
2-0 9.70---18.00Lost -10
Any 23.48---28.00Lost -10

Wolves
1-0 3.70---3.95Lost -10
2-0 7.70---9.40Won 84.00
Any 13.39---14.50Lost -10


Half time Match odds total: +32.80 :)
Half time scores total: +32.60 :)

[MattR]

Lowest odds from predicted odds

Team (pred/actual)

Burnley HH 3.52/3.80 Lost -10
Coventry AA 3.53/5.90 Won +49.00
Sheff Utd AA 3.72/4.70 Won +37.00
Birmingham AA 4.32/4.00 Lost -10
Preston HH 3.84/3.00 Lost -10
Sheff Weds HH 2.28/4.40 Lost -10
Swansea HH 2.72/3.05 Lost -10
Wolves HH 2.11/2.54 Won +15.40
Reading HH 2.62/2.32 Won +13.32

Dutching three predicted odds that are under real odds - if there are more than three then I will take the three lowest, if there are less than 3 then will just take what there is as the bet


Burnley HH,DD,AA BF Dutched odds: 1.70 Won +7.00
Charlton DA,AA 2 only BF Dutched odds: 3.50 Won +25.00
Nottm F DD,DA,AA BF Dutched odds: 1.90 Won +9.00
Plymouth HH 1 only BF odds: 6.40 Lost -10
Preston DH,DD,AA BF Dutched odds: 2.20 Won +12.00
Sheff W HH,DH 2 only BF Dutched odds: 2.60 Won +16.00
Swansea HH,AD,AA BF Dutched odds: 2.00 Lost -10
Wolves HH 1 only BF odds: 2.54 Won +15.40
Reading DH,DA,AA BF Dutched odds: 3.30 Lost -10



Stand out odds (according to predicted odds v available odds)
Sheff Weds HH 2.28 actual odds 4.40 Lost -10
Sheff Utd AA 3.71 actual odds 4.70 Won +37.7
Plymouth HH 4.47 actual odds 6.40 Lost -10
Blackpool AA 12.20 actual odds 17.50 Lost -10





Nights results:
Lowest Predicted Odds HTFT's: +64.72 :) 4/9

Dutched HTFT's: +54.40 :) 6/9

Stand Out Value HTFT's: +7.70 :) 1/4

Half time Match odds: +32.80 :) 3/5

Half time scores: +32.60 :) 3/4


Wow

[counterfeit]

well, that's a good experiment. If that keeps up for the next 6 weeks or so you'll be rich. Rich, rich, rich. £190 per match day.

Actually, being serious, it does seem very good and as I said above, really does prove the poor value being offered.

Well done and let's hope it keeps going that way.

[MattR]

Thanks CF. Certainly a promising start! In the end I actually put small bets on the value ones and the lowest predicted htft odds just for a bit of fun and made £26 odd so that was nice!



Lowest Priced Predicted Odds:

Team, pred/actual
Nice HH 3.70/3.05 Lost -10.00
Bellinzona HH 2.62/3.35 Postponed
Ipswich HH 3.31/3.20 Lost -10.00
Watford HH 3.04/3.60
Godoy Cruz HH 3.09/5.80

Value Dutched:

Game, Picks, Dutched price
Nice HD,DA,AA, 4.80 Lost -10.00
Bellinzona HH,AH,AD,AA (done all 4 here as price ok), 1.90 Postponed
Ipswich DD,DA 2 only, 3.90 Lost -10.00
Watford HH 1 only, 3.70
Godoy Cruz HH, DH 2 only, 3.40

Stand Out Value HTFT - some interesting high one's thrown up today

Nice HA 52.37/70.00 Lost -10.00
Nice HD 16.68/20.00 Lost -10.00
Nice DA 6.92/14.50 Lost -10.00
Ipswich DA 8.12/12.50 Lost -10.00
Bellinzona HH 2.62/3.05 Postponed
Bellinzona AH 24.44/32.00 Postponed
Bellinzona AD 12.48/16.50 Postponed
Bellinzona AA 4.93/7.60 Postponed
Watford HH 3.04/3.70
Godoy Cruz HH 3.09/5.90
Godoy Cruz DH 5.84/8.00

If I were betting these stand out value one's then I think I would dutch the one's with more than one selection in.

Half time match odds bits to follow later.