In my first lesson in betting using a model, I'm going to explain some of the mechanics of how a bookmaker makes money. I talk about the "overround" quite a lot in my posts so I will explain this in a bit of detail.
What is Overround?
Overround is essentially how a bookmaker makes money. A bigger overround means more money for a bookmaker, and conversely less money for the punters. Overround is represented by a % number.
So how do you calculate your overround?
Liverpool @ 1.36
Draw @ 5.5
West Ham @ 9
First take your odds above, and convert these into percentages. for example, to convert home win, (1/1.36) = 73.5%
Liverpool 73.5%
Draw 18.2%
West Ham 11.1%
Then add these up. These three percentages add to 102.8%. Why is this important? It means your bookmaker is taking a 2.8% margin on whatever bets are placed (in theory). With a perfectly balanced book, he will make a 2.8% profit.
For example, lets assume those odds are a true reflection of the publics perception of the actual odds of the game. The bookmaker receives £7,352.94 in bets on Liverpool, £1,818.18 in bets on the draw, and £1111.11 in bets on West Ham. The bookmaker has then received £10,282.23 in bets, and will pay out £10,000 no matter who wins. Therefore, the bookmaker is completely neutral to the result and will make a profit of £282.23 irrespective of the winner.
Of course, this is in a perfect world and it doesn't always work out like that. People lump a lot of money on the favourite, teams that are well backed, for example Liverpool, will get bets placed on them no matter what the odds are. It is this irrational behaviour which makes it rational for a bookmaker to price up not based on his beliefs, but on his beliefs of the public's beliefs. This weekend it allows the bookmaker to price up Liverpool to win at 73% chance, when a basic model will tell you their chance is closer to 57%.
In any case, overround is quite important. Here is the overround for a couple of bets from the first game of the season, Manchester United vs Tottenham:
Home-Draw-Away: 102.3%
Over-Under: 105.6%
So what does this mean? It means a home/draw/away bet is the best bet to place for someone betting using a football model. It gives the bookmaker the smallest advantage possible (a perfectly fair game would have an overround of exactly 100%) and means that when we are working with small margins, even if the results we are picking are random, our expected profit is only a 2.3% loss.
So what about other markets?
Double Chance: 106.1%
Both Teams to Score: 108.5%
Goals odd/even: 103.9%
Draw No Bet: 106.7%
Result/Both Teams to Score: 114.6%
Score bet (eg 1-1): 122.6%
What this demonstrates is that generally, the more bets you can place, the higher the typical overround. In a way, overround is similar to the bid/ask spread of a typical financial market. You want this number to be as small as possible. If you plan on betting properly, you need to test the markets you are going to bet on to make sure that your overround is as small as possible - bigger overrounds will erode your profits and make gambling unprofitable in the long run.
Saturday, 29 August 2015
Thursday, 27 August 2015
Gameweek 4 - August 29th
Its been a funny old season. Here we are, 30 games in and results seem to have defied convention. What do i mean by this? Well, across the 30 games so far we have:
Home Win: 6
Draw: 11
Away Win: 13
A draw rate of 37% at season end is unusually high. We normally see a draw rate between 25% and 34% (ill publish this study another day) with an increasing number of draws as the spread of talent in the league becomes more equal. However in reality this is only approximately 3-4 draws more than we would usually see. However 13 away wins and just 6 home wins is very unusual. This weekend we are actually predicting some home favourites are going to continue this pattern and will see a few top teams struggling for a result.
AST vs SUN - Double Chance Sun/Draw @ 2.01
What im about to show you will save you money. It may not make sense and I will do a future blog post on this to explain further, but for now work with me. The bet we want is Aston Villa not to win (it is -18% on the value meter). The bet we want is therefore double chance.
Do not bet on double chance.
Why not? because it is poor value and is for people who don't understand numbers. Double Chance sunderland/draw is priced currently at 1.9. So lets say we bet £100 on this, to return £190 as long as Villa don't win.
Great. Except it isn't. You've just paid the bookmaker because you are lazy. The draw is 3.5 and the away win is 4.75. If you had bet £54.29 on the draw, and £40 on the away win, you would have the following payout table:
Aston Villa win: Return £0
Draw: Return £190
Sunderland win: Return £190
You have therefore reconstructed your desired bet all for an outlay of £94.29. You have saved yourself £5.71 by not being lazy. Ill explain in another post why this happens in further detail. Of course this is completely irrelevant because Aston Villa will win and we will lose money either way.
CHE vs CPL - Double Chance Draw/CPL @ 3.44
A strange bet has caught my eye. Mourinho's home record is famous (9 years unbeaten at home in a row) but with Chelsea out of form and Palace with a great team it is not difficult to imagine Chelsea losing points. At 3.44 and a value of around 22% there are worse bets to take this weekend.
SWA vs MNU - Double Chance SWA/Draw @ 1.75
Interesting that the model thinks that Chelsea, Man United and Liverpool are massively overvalued this weekend. In any case, this bet doesnt pay anywhere near as well as going against Chelsea and Liverpool (both priced up at 1.36 this weekend) but still indicated as a value bet of around 7%.
LIV vs WHU - Double Chance Draw/WHU @ 3.41
Liverpool are at Home to West Ham and its telling me to do the same bet at 16% (model 57% vs bookmaker 73%). On the face of it this looks a terrible bet. West Ham have been awful, their victory over Arsenal aside, and so i am not going to do this bet. Then I am reminded that I am but a mere monkey standing in the way of this supreme machine's power until it learns how to blog by itself. So we have our bets for this weekend.
Thats it for this week guys. Other value bets include an Everton win against Spurs (7.5%), Over 2.5 goals for the same game (5%), Newcastle Arsenal over 2.5 (8%), Liverpool West Ham under 2.5 for (6.7%), under 2.5 for the Chelsea game (8%), and Under 2.5 for the Villa game (5%).
List in full:
NEW 1-2 ARS
AST 1-1 SUN
BOU 0-1 LEI*
CHE 1-0 CPL
LIV 1-0 WHU (see!!)
MCI 3-1 WAT*
STO 1-0 WBA
TOT 1-1 EVE
SOU 1-2 NOR*
SWA 1-1 MNU
*still not including any promoted teams in anything i publish, or when I tally my perfect scores at various points in the season.
Home Win: 6
Draw: 11
Away Win: 13
A draw rate of 37% at season end is unusually high. We normally see a draw rate between 25% and 34% (ill publish this study another day) with an increasing number of draws as the spread of talent in the league becomes more equal. However in reality this is only approximately 3-4 draws more than we would usually see. However 13 away wins and just 6 home wins is very unusual. This weekend we are actually predicting some home favourites are going to continue this pattern and will see a few top teams struggling for a result.
AST vs SUN - Double Chance Sun/Draw @ 2.01
What im about to show you will save you money. It may not make sense and I will do a future blog post on this to explain further, but for now work with me. The bet we want is Aston Villa not to win (it is -18% on the value meter). The bet we want is therefore double chance.
Do not bet on double chance.
Why not? because it is poor value and is for people who don't understand numbers. Double Chance sunderland/draw is priced currently at 1.9. So lets say we bet £100 on this, to return £190 as long as Villa don't win.
Great. Except it isn't. You've just paid the bookmaker because you are lazy. The draw is 3.5 and the away win is 4.75. If you had bet £54.29 on the draw, and £40 on the away win, you would have the following payout table:
Aston Villa win: Return £0
Draw: Return £190
Sunderland win: Return £190
You have therefore reconstructed your desired bet all for an outlay of £94.29. You have saved yourself £5.71 by not being lazy. Ill explain in another post why this happens in further detail. Of course this is completely irrelevant because Aston Villa will win and we will lose money either way.
CHE vs CPL - Double Chance Draw/CPL @ 3.44
A strange bet has caught my eye. Mourinho's home record is famous (9 years unbeaten at home in a row) but with Chelsea out of form and Palace with a great team it is not difficult to imagine Chelsea losing points. At 3.44 and a value of around 22% there are worse bets to take this weekend.
SWA vs MNU - Double Chance SWA/Draw @ 1.75
Interesting that the model thinks that Chelsea, Man United and Liverpool are massively overvalued this weekend. In any case, this bet doesnt pay anywhere near as well as going against Chelsea and Liverpool (both priced up at 1.36 this weekend) but still indicated as a value bet of around 7%.
LIV vs WHU - Double Chance Draw/WHU @ 3.41
Liverpool are at Home to West Ham and its telling me to do the same bet at 16% (model 57% vs bookmaker 73%). On the face of it this looks a terrible bet. West Ham have been awful, their victory over Arsenal aside, and so i am not going to do this bet. Then I am reminded that I am but a mere monkey standing in the way of this supreme machine's power until it learns how to blog by itself. So we have our bets for this weekend.
Thats it for this week guys. Other value bets include an Everton win against Spurs (7.5%), Over 2.5 goals for the same game (5%), Newcastle Arsenal over 2.5 (8%), Liverpool West Ham under 2.5 for (6.7%), under 2.5 for the Chelsea game (8%), and Under 2.5 for the Villa game (5%).
List in full:
NEW 1-2 ARS
AST 1-1 SUN
BOU 0-1 LEI*
CHE 1-0 CPL
LIV 1-0 WHU (see!!)
MCI 3-1 WAT*
STO 1-0 WBA
TOT 1-1 EVE
SOU 1-2 NOR*
SWA 1-1 MNU
*still not including any promoted teams in anything i publish, or when I tally my perfect scores at various points in the season.
Gameweek 3 - Another Failure
Pretty disappointed yet again and i think we can safely conclude that the model isnt working with 0 profitable weeks out of 3.
MNU vs NEW - Home - result 0-0
While Newcastle were never going to win this these are exactly the sort of games I try to avoid - massive odds on home favourite. In truth United should have won as a goal was wrongly disallowed for offside but this was never worth betting on given Uniteds fixtures this season in the league.
Lei vs TOT - Home - result 1-1
As mentioned, Aquitis was actually spot on, predicting a perfect score of 1-1. The model essentially said Spurs wouldn't win, and i made the same mistake that I made in GW1 with Liverpool/Stoke - only this time it was costly.
WBA vs CHE - Over 2.5 result 2-3
The only winner of the weekend.
ARS vs LIV - Home - result 0-0
How this game ended 0-0 I will never know. Arsenal scored a legitimate goal that was incorrectly given offside (I'm seeing a pattern...)
A quick scan of the predictions in full and I see most games I was a goal off. I'll keep this going but its clear I need more data. The only reason I think these bets are value bets is because bookmakers have more sophisticated models than me.
Happy Hunting
Aquitis
MNU vs NEW - Home - result 0-0
While Newcastle were never going to win this these are exactly the sort of games I try to avoid - massive odds on home favourite. In truth United should have won as a goal was wrongly disallowed for offside but this was never worth betting on given Uniteds fixtures this season in the league.
Lei vs TOT - Home - result 1-1
As mentioned, Aquitis was actually spot on, predicting a perfect score of 1-1. The model essentially said Spurs wouldn't win, and i made the same mistake that I made in GW1 with Liverpool/Stoke - only this time it was costly.
WBA vs CHE - Over 2.5 result 2-3
The only winner of the weekend.
ARS vs LIV - Home - result 0-0
How this game ended 0-0 I will never know. Arsenal scored a legitimate goal that was incorrectly given offside (I'm seeing a pattern...)
A quick scan of the predictions in full and I see most games I was a goal off. I'll keep this going but its clear I need more data. The only reason I think these bets are value bets is because bookmakers have more sophisticated models than me.
Happy Hunting
Aquitis
Thursday, 20 August 2015
Gameweek 3- August 22nd
After last week's 3 correct scores, I have decided to post score predictions after my four key value bets. I've been doing a lot of thinking about value bets - basically a discrepancy between my model and the bookmakers price - and while value bets may be identified in equilibrium, its quite likely that during parts of the season when there is a lot of ongoing chance (August, January) this may simply be bookmakers having a better model with more information.
Still, my bets for Gameweek 3:
MNU vs NEW - Home @ 1.36
I really do not like this bet. Apparently Man Utd have an 85% chance of success in this game, meaning this bet has a value of roughly 10%. I prefer backing the outsider but there really isnt that much on this week. Interesting this was picked - ill have to keep an eye on how bookmakers look at teams who have had a midweek game in the european cups going forward.
LEI vs TOT - Home @ 2.6
Now HERE is a good bet! Leicester at a massive 2.6 when Aquitis predicts a 50% chance of winning gives a juicy value of 12%. Oddly the most likely score is 1-1 @ 10.2%, though its clear to see the model really doesn't fancy Spurs' chances away. I would normally take a Draw No Bet on this but the value for the draw is pretty poor on this game so we will stick to the home win.
WBA vs CHE - Over 2.5 @ 2.07
I really dont like this bet. I wont elaborate - just think about the respective managers.
I really really dont like this one.
ARS vs LIV - Home @ 1.64
This week has thrown up some very strange suggestions, none of which I particularly like. But as I said, its slim pickings this week- which is why we have this at a pretty mediocre value of 5.4%. Arsenal should have enough about them to see off Liverpool, but then I'd have said the same about West Ham...
Guys, predictions in full
MNU vs NEW 2-0
CPL vs AVL 1-1
LEI vs TOT 1-1 (But if you rebase it from Ranieri's arrival then a 4-2 victory..haha..)
NOR vs STO n/a (with hardly any data the model suggests 0-2)
SUN vs SWA 0-1
WHU vs BOU n/a (apparently a 40% chance of a 0-0...lets give the promoted teams time)
WBA vs CHE 1-2
EVE vs MCI 1-1 (!!!)
WAT vs SOT n/a (0-0 is 100% chance....)
ARS vs LIV 1-0 though 2-0 is preferred in the non time adjusted model.
Happy Hunting!
Aquitis
Still, my bets for Gameweek 3:
MNU vs NEW - Home @ 1.36
I really do not like this bet. Apparently Man Utd have an 85% chance of success in this game, meaning this bet has a value of roughly 10%. I prefer backing the outsider but there really isnt that much on this week. Interesting this was picked - ill have to keep an eye on how bookmakers look at teams who have had a midweek game in the european cups going forward.
LEI vs TOT - Home @ 2.6
Now HERE is a good bet! Leicester at a massive 2.6 when Aquitis predicts a 50% chance of winning gives a juicy value of 12%. Oddly the most likely score is 1-1 @ 10.2%, though its clear to see the model really doesn't fancy Spurs' chances away. I would normally take a Draw No Bet on this but the value for the draw is pretty poor on this game so we will stick to the home win.
WBA vs CHE - Over 2.5 @ 2.07
I really dont like this bet. I wont elaborate - just think about the respective managers.
I really really dont like this one.
ARS vs LIV - Home @ 1.64
This week has thrown up some very strange suggestions, none of which I particularly like. But as I said, its slim pickings this week- which is why we have this at a pretty mediocre value of 5.4%. Arsenal should have enough about them to see off Liverpool, but then I'd have said the same about West Ham...
Guys, predictions in full
MNU vs NEW 2-0
CPL vs AVL 1-1
LEI vs TOT 1-1 (But if you rebase it from Ranieri's arrival then a 4-2 victory..haha..)
NOR vs STO n/a (with hardly any data the model suggests 0-2)
SUN vs SWA 0-1
WHU vs BOU n/a (apparently a 40% chance of a 0-0...lets give the promoted teams time)
WBA vs CHE 1-2
EVE vs MCI 1-1 (!!!)
WAT vs SOT n/a (0-0 is 100% chance....)
ARS vs LIV 1-0 though 2-0 is preferred in the non time adjusted model.
Happy Hunting!
Aquitis
Wednesday, 19 August 2015
Gameweek 2- Steady
So...2 correct and 2 incorrect. Its early season so Im still giving the model the benefit of the doubt but my theory on Value bets doesn't seem to be performing particularly well. Perhaps value bets only seem to be valuable because my model just isnt as sophisticated as the one that bookmakers use? Perhaps my model said southampton had a 70% chance of winning but in reality their chances were closer to the 50% quoted by bookmakers, due to transfers, Europa commitments etc. Still, we wont be able to see the true performance or derive any conclusions this early.
SOT vs EVE - Home - result 0-3
Not a great result. Everton seem to be the most inconsistent team in the league, and this was a splendid performance. The corresponding fixture last season was a 3-0 victory to Southampton (though Martinez did win the fixture at Goodison 1-0 later in the season). Southampton were missing Claise and it showed.
SWA vs New - Home - result 2-0
The model predicted a 2-0 home win (both the basic and time adjusted) with an average % for that exact scoreline 13.3%.
WHU vs LEI - Home - result 1-2
Ranieri's first two games have yielded 6 points to leave them top of the table. It will be interesting to see what happens from here on in.
MCI vs CHE - Home - result 3-0
Wow. It just goes to show how difficult defending the Premier League title is. Mourinhos men were dreadful and didnt look up to scratch, with their captain subbed at half time. My least confident forecast was the best result (Again!).
This means on LSP, I was up this week (10 pence..) but unfortunately the adaptation of Kellys means i placed more money on the southampton game and thus I am down on this wallet.
LSP: £95.60
Kellys: £87.87
I will be publishing an article soon explaining LSP/Kellys and theories on how to portion your bets, just as a bit of side reading.
One more note: An interesting thing happened this weekend. Of the 7 games the model forecast (remember promoted teams still excluded due to the messy translation of results from championship to premier league), the model predicted the following results:
AVL vs MNU 0-1
SOT vs EVE 2-0
SWA vs NEW 2-0
TOT vs STO 1-0
WHU vs LEI 1-0
CPL vs ARS 0-2
MCI vs CHE 2-1
Firstly, I know thats a lot of clean sheets - typically these are the most common results. What is interesting is that of these results, the model predicted 3 scores out of 7. Even at terrible odds of 6/1, you only need to get one result to breakeven - Ill keep track of the perfect score bet (even with terrible overround) to see if the model is any better at predicting this.
Happy Hunting,
Aquitis.
SOT vs EVE - Home - result 0-3
Not a great result. Everton seem to be the most inconsistent team in the league, and this was a splendid performance. The corresponding fixture last season was a 3-0 victory to Southampton (though Martinez did win the fixture at Goodison 1-0 later in the season). Southampton were missing Claise and it showed.
SWA vs New - Home - result 2-0
The model predicted a 2-0 home win (both the basic and time adjusted) with an average % for that exact scoreline 13.3%.
WHU vs LEI - Home - result 1-2
Ranieri's first two games have yielded 6 points to leave them top of the table. It will be interesting to see what happens from here on in.
MCI vs CHE - Home - result 3-0
Wow. It just goes to show how difficult defending the Premier League title is. Mourinhos men were dreadful and didnt look up to scratch, with their captain subbed at half time. My least confident forecast was the best result (Again!).
This means on LSP, I was up this week (10 pence..) but unfortunately the adaptation of Kellys means i placed more money on the southampton game and thus I am down on this wallet.
LSP: £95.60
Kellys: £87.87
I will be publishing an article soon explaining LSP/Kellys and theories on how to portion your bets, just as a bit of side reading.
One more note: An interesting thing happened this weekend. Of the 7 games the model forecast (remember promoted teams still excluded due to the messy translation of results from championship to premier league), the model predicted the following results:
AVL vs MNU 0-1
SOT vs EVE 2-0
SWA vs NEW 2-0
TOT vs STO 1-0
WHU vs LEI 1-0
CPL vs ARS 0-2
MCI vs CHE 2-1
Firstly, I know thats a lot of clean sheets - typically these are the most common results. What is interesting is that of these results, the model predicted 3 scores out of 7. Even at terrible odds of 6/1, you only need to get one result to breakeven - Ill keep track of the perfect score bet (even with terrible overround) to see if the model is any better at predicting this.
Happy Hunting,
Aquitis.
Friday, 14 August 2015
Gameweek 2 - August 15th
Southampton Vs Everton - Home @ 1.95
Great Value. Martinez has a poor away record - 5 wins and 4 draws last season - while southamptons home record was the best outside the top four. Aquitis predicts approximately a 70% chance of southampton victory, so at those odds you have value of around 19% - making this the bet of the weekend.
Swansea Vs Newcastle - Home @ 1.95
Sticking with Home wins, there is approximately 13% value in the Home win in Swansea this weekend. Ok, so the managerial change bolt on wont give me data, but the other two are flagging this as a good bet.
My only worry with this is that Mclaren has made a lot of squad changes - this is a very different side to the one that played under Frank Carver. Perhaps the bookmakers have put more emphasis on how good newcastle looked last game? I suppose we shall see this weekend.
West Ham Vs Leicester - Home @ 2.25
I know, I know. Given that both managers are new hires over the summer, I have less data to play around with and thus less expect more volatile returns. Still, I cant do as bad as last week huh?! Another Home win, the model reckons a 56% chance in this one so at least 12% value to take out.
Sticking to my strict rule of last week, my final bet SHOULD be Man City to beat chelsea. But i just dont like the big games. I especially dont like going against Mourinho in big games. So now i am torn. Do i put on a bet I dont trust? Or do i find something with less value?
.......
Man City Vs Chelsea - Home @ 2.1
Here goes nothing. Pulling results off of my depreciation curve I get a 59.7% chance of a home win. So thats my fourth bet of the weekend. In reality, Mourinho will park the bus and this will lose. But if the model says so, the model says so.
Right! Thats it for this week guys. Speak on Monday. There was a tiny bit of value in the Arsenal win too and ill go through those on Monday in a Bullseye-torturous like fashion (lets have a look at what you could have won!) but until then, best of luck.
Happy Hunting,
Aquitis
Monday, 10 August 2015
Gameweek 1 - A lesson in overconfidence
Well that was a complete and total disaster wasn't it?
So much for the theory that sat behind my model. A resounding 1/4 results leaves me £9.20 down on my adaptation of Kellys, and £4.50 down on a LSP £2 portfolio. It wasn't a good weekend. So..what went wrong with the bets?
MNU vs TOT - Over 2.5 goals - result 1-0
My model seemed to suggest there was some value in the over 2.5 bet. What went wrong? For those who watched the game, you will have noted hardly any chances made. is this perhaps two sides cancelling each other out? Last seasons fixtures were a 3-0 at Old Trafford and a 0-0 at WHL. The season before this the teams battled it out in a 2-2 draw at WHL and a 1-2 victory for Spurs at Old Trafford. What happened here was both teams lacked the incision to play that killer ball - united often spreading play wide to young, who would run at walker and either put in a poor cross or play it back to Carrick/ Schniderlin. While Kane and Eriksen showed great promise Kane's passing was sloppy and Eriksen wasted a great opportunity.
So...was this game lost due to a faulty model? Or was this just noise? After all, the model predicted a 63% chance of over 2.5, not 100%. I note that the model also said there was 5% value in backing United to win, but instead i went for the juicier overs bet (albeit this bet has more over round - Ill do more analysis on this later).
CHE vs SWA - Under 2.5 goals - result 2-2
Oh dear. Chelsea conceded just 9 goals at home last season - 0.47 per game. Swansea scored 1 per game. The average was 1.1 per game - on paper this looked a good bet, especially at Evens. yet what we saw was a Swansea team that clicked together. In the 2014 season both games ended 1-0 to Chelsea, although last season we saw 4-2 and 5-0 victories. So was this necessarily a good prediction? If the model believes there was value in under 2.5, should it not have seen what happened last season? Chelsea scoring just two goals could be as a result of not having a fully fit striker - perhaps Monk vs Mourinho styles mean a very open game?
I am toying with the idea of having a new bolt on (this will be the fourth) which places a greater emphasis on last seasons results. obviously this allows for some error in player replacement but currently i find myself looking through previous results for further clarity. This shouldn't happen. Over the coming weeks I will test as to whether past results indicate future results (after adjusting for quality of teams and manager changes of course).
In any case, this bet fell flat on its face. I entered Sunday praying for a Mame Biram Diouf Hattrick
STO vs LIV - Home win - result 0-1
Before the game I realised I had made a mistake. I had gambled slightly. The model was indifferent between a Stoke win and a Stoke win/draw (double chance). I had gone for a Stoke win because of the enhanced odds. Genuine question to my readers (who am I kidding....), what is better in your opinion, a bet at evens with a 75% chance of winning, or a bet of 3/1 with a 50% chance of winning? Only one chance to roll the dice so I don't want expected value analysis, just your opinion.
Onto the game. The model was fairly sure Stoke was the value bet here. I commented above that my new bolt on would emphasise previous results from last season. In this scenario my model would have SCREAMED to bet the house on stoke after the 6-1 demolition. But it just wasn't to be. Was it the loss of N'zonzi? Shawcross? Bojan? Arnautovic? Begovic (though i doubt that is what caused the lack of attacking swerve.) Stoke were allegedly very unlucky not to take something from the game so i am willing to write this off as noise, but things were really not looking good....
WBA vs MCI - Over 2.5 - result 0-3
Just! In truth City dominated this game. Probably the only team to look sharp out of the big boys. There was 13.5% value in backing this and coincidentally in my preview I mentioned that this was the one I felt least confident about!
So what did I not bet on? As you guys know, I dont want to do the promoted teams just yet. Translating data from Championship to Premier league is messy (probably due to the high number of transfers promoted clubs do and how varied the successes and failures are. My model also said there was great value in City away win (9.1%) though I don't like betting on the same game twice.
It also said:
Newcastle Vs Southampton - Bookies priced it correctly. My first bolt on suggested a tiny bit of value in Southampton win (3%) but my second bolt on said this was the worst bet you could do on this game. No possible prediction.
Arsenal Vs West Ham - No value in any bets at all - you bet on this game and you're just paying the over round. However there is a bit of value in Under 2.5 Goals (7%). Why did I ignore this? because honestly I didn't believe it was correct. This is something I will have to watch in future. There should be no emotion in this model, and that's a possible lost value bet there at generous odds. (2.33x)
Chelsea Vs Swansea - Home win is a very poor bet (-6.9%). Draw has the best value but still nothing to get excited about (3.5%). Under 2.5 goals is definitely the best bet (15.6%...)
Leicester Vs Sunderland - Bookmakers have priced this absolutely perfectly. H/D/A all loses value. Some value in Under 2.5 goals (3.3% - 6%) but again not too much to be excited about.
Man Utd Vs Tottenham - A bit of value in the Home win (4%) but Over 2.5 presents about 7.7% value - that's the bet to place.
I wonder..should i run separate books on over/under and FT result? They have very different overrounds and the model may be better at one over the other. Its something to consider going forward. I will archive this model, update the data with this weekends results and publish something later this week for gameweek 2 - hopefully better predictions!
I also plan to do a couple articles every now and then just with some thoughts, betting ideas and analysis for you guys to read. They may not necessarily be good betting ideas but I hope they will be entertaining nevertheless.
Happy Hunting Guys.
Aquitis.
So much for the theory that sat behind my model. A resounding 1/4 results leaves me £9.20 down on my adaptation of Kellys, and £4.50 down on a LSP £2 portfolio. It wasn't a good weekend. So..what went wrong with the bets?
MNU vs TOT - Over 2.5 goals - result 1-0
My model seemed to suggest there was some value in the over 2.5 bet. What went wrong? For those who watched the game, you will have noted hardly any chances made. is this perhaps two sides cancelling each other out? Last seasons fixtures were a 3-0 at Old Trafford and a 0-0 at WHL. The season before this the teams battled it out in a 2-2 draw at WHL and a 1-2 victory for Spurs at Old Trafford. What happened here was both teams lacked the incision to play that killer ball - united often spreading play wide to young, who would run at walker and either put in a poor cross or play it back to Carrick/ Schniderlin. While Kane and Eriksen showed great promise Kane's passing was sloppy and Eriksen wasted a great opportunity.
So...was this game lost due to a faulty model? Or was this just noise? After all, the model predicted a 63% chance of over 2.5, not 100%. I note that the model also said there was 5% value in backing United to win, but instead i went for the juicier overs bet (albeit this bet has more over round - Ill do more analysis on this later).
CHE vs SWA - Under 2.5 goals - result 2-2
Oh dear. Chelsea conceded just 9 goals at home last season - 0.47 per game. Swansea scored 1 per game. The average was 1.1 per game - on paper this looked a good bet, especially at Evens. yet what we saw was a Swansea team that clicked together. In the 2014 season both games ended 1-0 to Chelsea, although last season we saw 4-2 and 5-0 victories. So was this necessarily a good prediction? If the model believes there was value in under 2.5, should it not have seen what happened last season? Chelsea scoring just two goals could be as a result of not having a fully fit striker - perhaps Monk vs Mourinho styles mean a very open game?
I am toying with the idea of having a new bolt on (this will be the fourth) which places a greater emphasis on last seasons results. obviously this allows for some error in player replacement but currently i find myself looking through previous results for further clarity. This shouldn't happen. Over the coming weeks I will test as to whether past results indicate future results (after adjusting for quality of teams and manager changes of course).
In any case, this bet fell flat on its face. I entered Sunday praying for a Mame Biram Diouf Hattrick
STO vs LIV - Home win - result 0-1
Before the game I realised I had made a mistake. I had gambled slightly. The model was indifferent between a Stoke win and a Stoke win/draw (double chance). I had gone for a Stoke win because of the enhanced odds. Genuine question to my readers (who am I kidding....), what is better in your opinion, a bet at evens with a 75% chance of winning, or a bet of 3/1 with a 50% chance of winning? Only one chance to roll the dice so I don't want expected value analysis, just your opinion.
Onto the game. The model was fairly sure Stoke was the value bet here. I commented above that my new bolt on would emphasise previous results from last season. In this scenario my model would have SCREAMED to bet the house on stoke after the 6-1 demolition. But it just wasn't to be. Was it the loss of N'zonzi? Shawcross? Bojan? Arnautovic? Begovic (though i doubt that is what caused the lack of attacking swerve.) Stoke were allegedly very unlucky not to take something from the game so i am willing to write this off as noise, but things were really not looking good....
WBA vs MCI - Over 2.5 - result 0-3
Just! In truth City dominated this game. Probably the only team to look sharp out of the big boys. There was 13.5% value in backing this and coincidentally in my preview I mentioned that this was the one I felt least confident about!
So what did I not bet on? As you guys know, I dont want to do the promoted teams just yet. Translating data from Championship to Premier league is messy (probably due to the high number of transfers promoted clubs do and how varied the successes and failures are. My model also said there was great value in City away win (9.1%) though I don't like betting on the same game twice.
It also said:
Newcastle Vs Southampton - Bookies priced it correctly. My first bolt on suggested a tiny bit of value in Southampton win (3%) but my second bolt on said this was the worst bet you could do on this game. No possible prediction.
Arsenal Vs West Ham - No value in any bets at all - you bet on this game and you're just paying the over round. However there is a bit of value in Under 2.5 Goals (7%). Why did I ignore this? because honestly I didn't believe it was correct. This is something I will have to watch in future. There should be no emotion in this model, and that's a possible lost value bet there at generous odds. (2.33x)
Chelsea Vs Swansea - Home win is a very poor bet (-6.9%). Draw has the best value but still nothing to get excited about (3.5%). Under 2.5 goals is definitely the best bet (15.6%...)
Leicester Vs Sunderland - Bookmakers have priced this absolutely perfectly. H/D/A all loses value. Some value in Under 2.5 goals (3.3% - 6%) but again not too much to be excited about.
Man Utd Vs Tottenham - A bit of value in the Home win (4%) but Over 2.5 presents about 7.7% value - that's the bet to place.
I wonder..should i run separate books on over/under and FT result? They have very different overrounds and the model may be better at one over the other. Its something to consider going forward. I will archive this model, update the data with this weekends results and publish something later this week for gameweek 2 - hopefully better predictions!
I also plan to do a couple articles every now and then just with some thoughts, betting ideas and analysis for you guys to read. They may not necessarily be good betting ideas but I hope they will be entertaining nevertheless.
Happy Hunting Guys.
Aquitis.
Thursday, 6 August 2015
Gameweek 1 - August 8th
Already we have a slight problem. Trying to translate championship results into premier league performance has proved to be very erratic - I could of course take an average extrapolation but I would rather not at this stage, especially if i am looking at value bets around 7-12%. Therefore, I will not be making bets on games involving promoted teams until Gameweek 7, until i have more data. The promoted teams seem to have a balanced set of opening games:
Bournemouth: AVL, liv, whm, LEI, nor, SUN
Norwich: CPL, sun, STO, sot, BOU, liv
Watford: eve,WBA, SOT, mci, SWA, new
Perhaps a slightly easy run for bournemouth but it should give a good idea of how attacking these teams are.
Without further delay, here are my bets:
Mnu v Tot - Over 2.5 @ 1.8 (Aquitis - 63.2%, bookies 55.6%). Bet £3.46
Che v Swa - Under 2.5 @ Evs (Aquitis 62.3%, bookies 50%). Bet £4.91
Sto v Liv - Home @ 3.4 (Aquitis 49.9%, bookies 29.4%). Bet £5.80
WBA vs Mnc - Over 2.5 @ 1.75 (Aquitis 66.7%, bookies 57.1%). Bet £4.46
Oddly, the result Aquitis is most sure about, I am least sure about. West Brom vs Manchester City on the 10th, over 2.5 goals with Aguero a doubt and Dzeko/Jovetic in Spain? A Pulis team, over 2.5 goals? Are you sure?
What do I know. Im just a human.
As mentioned I will be running two balances, one for Kellys and one with £2 LSP with a £100 bankroll on each just to see where they end up. Bet365 is my bookmaker of choice - honest company, great reviews, the best platform, and an overround between 2%-3% on most PL games. What more could you ask for.
LSP Bankroll: £100
Kelly Bankroll: £100
Next week, I'll review the first week and give my next set of predictions. I am very conscious that early season results fluctuate a lot, so am not expecting immediate success, but lets see how the weekend goes.
Happy hunting.
Aquitis.
A new season
The season kicks of on Saturday, 8th August 2015. That day will also mark the first published predictions of my model, Aquitis. Aquitis is a statistical based model that is designed to use probability theory to work out value bets. Why value bets? A bet is only worth doing if you think it is more likely than the bookmaker. The best model in the world will not have 100% accuracy - there will be bad weeks. The most important thing is to see where i end up in May. As the season goes on there will be minor tweaks to Aquitis methodology but I intend to keep it the same.
Throughout the season I intend to publish my value bets on the Thursday / Friday, and record my progress in two ways. 1) using a LSP returns record and 2) using the Kelly criteria. Please note that I will be betting on these bets before publishing - obviously if you think the purpose of this blog is to bet on something, publish positive findings, move the markets on an exchange and then hedge then please do not follow my bets. I do not intend to use an exchange or to hedge these bets but be clear - I have placed money on these bets.
Please let me know comments / suggestions. I do intend to publish eventually with a formal PDF but for now I just need somewhere to store results. Who knows - i could be bankrupt by christmas and have no need for PDFs!
One more thing - as mentioned before, no model can predict with 100% accuracy. I am an amateur gambler. If you do follow these tips, bet responsibly and in small increments. Its just meant to be a bit of fun.
Have a good season.
Aquitis.
Throughout the season I intend to publish my value bets on the Thursday / Friday, and record my progress in two ways. 1) using a LSP returns record and 2) using the Kelly criteria. Please note that I will be betting on these bets before publishing - obviously if you think the purpose of this blog is to bet on something, publish positive findings, move the markets on an exchange and then hedge then please do not follow my bets. I do not intend to use an exchange or to hedge these bets but be clear - I have placed money on these bets.
Please let me know comments / suggestions. I do intend to publish eventually with a formal PDF but for now I just need somewhere to store results. Who knows - i could be bankrupt by christmas and have no need for PDFs!
One more thing - as mentioned before, no model can predict with 100% accuracy. I am an amateur gambler. If you do follow these tips, bet responsibly and in small increments. Its just meant to be a bit of fun.
Have a good season.
Aquitis.
Subscribe to:
Posts (Atom)