by Adrian Worton
Having developed a model which converts bookies' odds into probabilities, and having looked at real-life results to see how the model needs tweaking, we are now in a position where we can analyse our model in order to see which bets are of a value which makes us likely to return a profit.
Having developed a model which converts bookies' odds into probabilities, and having looked at real-life results to see how the model needs tweaking, we are now in a position where we can analyse our model in order to see which bets are of a value which makes us likely to return a profit.
How do bookies make their money?
As we covered in our pre-Grand National article, bookies weight odds such that the odds are slightly shorter than they ought to be, to make sure that they return a profit.
Whilst it is, of course, possible to have the occasional victory with a bet, in the long run you will be expected to lose money, and with millions of bets placed every day the bookmakers almost always see a healthy return.
Therefore in order to turn the tables, we need to see if there is a group of bets which are under-valued, such that in the long run we would be expected to win money from them.
Individual constituencies
The transformation of our model we carried out last week essentially saw the probabilities adjusted such that the favourites became stronger and the outsiders weaker, in keeping with results from the past. Therefore, by comparing our adjusted probabilities to the odds, there will be some bets which look to return money, on average. The best of these return between 10-30p per £1 placed. If we pick enough of these, the law of averages which benefits bookmakers can be turned against them to make the chances of a small profit highly likely.
Without going into the very specifics of how the best-value bets were found, in general the best-value ones are the favourites in reasonably marginal seats, often ones who have a 70% chance of winning, but not the nailed-on favourites, as their odds of 1/50 or lower are so small as not to be worth the risk.
For the full list, scroll to the end of the article.
Other bets
We ran our usual 1000 simulations on our model to get an idea of the total number of seats the parties are looking at finishing on. By comparing these to some specialist bets, we can make a handful of further bets:
Majority. None of our simulations, in any version of the model, has yielded a single majority. Therefore it is a no-brainer to bet on such a likely occurrence, given that the odds are not yet pointlessly low.
Conservative seats. We found that the Conservative seats ranged between 260 and 291, with the majority b eing between 270 and 280. Therefore, in the 'banded' markets, betting on both 251-275 and 276-300 seats are a solid bet. Similarly, in the 'over or under' markets, there are good odds available on the party scoring below 286.5 seats.
Labour seats. In our simulations Labour's range was slightly wider, with the number of seats always being between 254 and 290. So once again the 251-275 and 276-300 bands represent a very good-value bet, as does the above 268.5 in the over or under markets.
Liberal Democrat seats. Whilst we are anticipating an onslaught of lost Lib Dem seats, we are currently expecting their total to be in the high-20s. Therefore with evens available on them scoring above 25.5, this represents good value.
UKIP seats. The number of UKIP seats in our model have always exceeded those predicted by polls, as they are second-favourite in a huge number of constituencies. Therefore if our model is to be trusted, the evens available on them scoring above 3.5 seats is another good-value bet.
Conclusion
It is only possible to find so many worthwhile bets due to the vast number of markets available on the General Election, with 650 seats up for grabs, plus markets such as party seat totals. It means we can make such a number of bets where the odds are in our favour that the element of risk that is inevitable with a small number of bets is greatly diminished.
The list of bets
So, these are the TGIAF-recommended bets for the 2015 General Election. The odds were checked over the weekend of the 12th-13th of April, and if they are now shorter than listed, they are no longer recommended, as the whole nature of the bets rests on the balance between the length of the odd given and the probability of it coming through.
The bets are listed below, in the format: Market, Bet (Odds). So for example, if a recommendation was to bet on Plaid winning the seat of Buckingham, at odds of 2/5, then it would be listed: Buckingham, PLA (2/5). All initials represent compass points (i.e. North, East, South and West). An even amount should be placed on each bet, except those marked with *, which will be explained at the bottom.
Firstly, constituencies:
As we covered in our pre-Grand National article, bookies weight odds such that the odds are slightly shorter than they ought to be, to make sure that they return a profit.
Whilst it is, of course, possible to have the occasional victory with a bet, in the long run you will be expected to lose money, and with millions of bets placed every day the bookmakers almost always see a healthy return.
Therefore in order to turn the tables, we need to see if there is a group of bets which are under-valued, such that in the long run we would be expected to win money from them.
Individual constituencies
The transformation of our model we carried out last week essentially saw the probabilities adjusted such that the favourites became stronger and the outsiders weaker, in keeping with results from the past. Therefore, by comparing our adjusted probabilities to the odds, there will be some bets which look to return money, on average. The best of these return between 10-30p per £1 placed. If we pick enough of these, the law of averages which benefits bookmakers can be turned against them to make the chances of a small profit highly likely.
Without going into the very specifics of how the best-value bets were found, in general the best-value ones are the favourites in reasonably marginal seats, often ones who have a 70% chance of winning, but not the nailed-on favourites, as their odds of 1/50 or lower are so small as not to be worth the risk.
For the full list, scroll to the end of the article.
Other bets
We ran our usual 1000 simulations on our model to get an idea of the total number of seats the parties are looking at finishing on. By comparing these to some specialist bets, we can make a handful of further bets:
Majority. None of our simulations, in any version of the model, has yielded a single majority. Therefore it is a no-brainer to bet on such a likely occurrence, given that the odds are not yet pointlessly low.
Conservative seats. We found that the Conservative seats ranged between 260 and 291, with the majority b eing between 270 and 280. Therefore, in the 'banded' markets, betting on both 251-275 and 276-300 seats are a solid bet. Similarly, in the 'over or under' markets, there are good odds available on the party scoring below 286.5 seats.
Labour seats. In our simulations Labour's range was slightly wider, with the number of seats always being between 254 and 290. So once again the 251-275 and 276-300 bands represent a very good-value bet, as does the above 268.5 in the over or under markets.
Liberal Democrat seats. Whilst we are anticipating an onslaught of lost Lib Dem seats, we are currently expecting their total to be in the high-20s. Therefore with evens available on them scoring above 25.5, this represents good value.
UKIP seats. The number of UKIP seats in our model have always exceeded those predicted by polls, as they are second-favourite in a huge number of constituencies. Therefore if our model is to be trusted, the evens available on them scoring above 3.5 seats is another good-value bet.
Conclusion
It is only possible to find so many worthwhile bets due to the vast number of markets available on the General Election, with 650 seats up for grabs, plus markets such as party seat totals. It means we can make such a number of bets where the odds are in our favour that the element of risk that is inevitable with a small number of bets is greatly diminished.
The list of bets
So, these are the TGIAF-recommended bets for the 2015 General Election. The odds were checked over the weekend of the 12th-13th of April, and if they are now shorter than listed, they are no longer recommended, as the whole nature of the bets rests on the balance between the length of the odd given and the probability of it coming through.
The bets are listed below, in the format: Market, Bet (Odds). So for example, if a recommendation was to bet on Plaid winning the seat of Buckingham, at odds of 2/5, then it would be listed: Buckingham, PLA (2/5). All initials represent compass points (i.e. North, East, South and West). An even amount should be placed on each bet, except those marked with *, which will be explained at the bottom.
Firstly, constituencies:
Amber Valley, LAB (1/2) Belfast S., SDLP (4/6) Brigg & Goole, CON, (3/10) Brighton Pavilion, GRE (2/7) Bristol W., LIB (evens) Cardiff Central, LAB (4/9) Carmarthen E. & Dinefwr, PLA (2/7) Carmarthan W. & S. Pembrokeshire, CON (4/7) Chippenham, CON (8/15) Cleethorpes, CON (4/6) Crawley, CON (3/10) Dover, CON (4/9) E. Dunbartonshire, SNP (4/6) E. Renfrewshire, LAB (4/9) Eastleigh, LIB (4/7) Edinburgh W., SNP (4/5) Erewash, LAB (4/7) Great Yarmouth, CON (5/6) Hampstead & Kilburn, LAB (1/3) Harlow, CON (1/3) Hastings & Rye, LAB (4/9) | Ilford N., CON (2/5) Newcastle-under-Lyme, LAB (2/7) N. Thanet, CON, (4/9) Peterborough, CON (1/3) Plymouth Moor View, LAB (1/4) Plymouth, Sutton & Devonport, LAB (4/11) Portsmouth N., CON (2/5) Portsmouth S., CON (evens) Somerton & Frome, CON (2/7) S. Basildon & E. Thurrock, CON (8/15) S. Thanet, UKIP (4/6) Southport, LIB (1/2) St. Austell & Newquay, CON (1/9) Taunton Deane, CON (4/7) Thurrock, UKIP (4/6) Upper Bann, DUP (2/7) Walsall N., LAB (1/3) Waveney, LAB (2/5) W. Aberdeenshire & Kincardine, SNP (1/2) Wyre Forest, CON (8/15) |
All other bets:
Conservative seats, below 286.5 (10/11)
Conservative seats banded, 251-275 (7/2)*
Conservative seats banded, 276-300 (13/8)*
Labour seats, above 268.5 (8/11)
Labour seats banded, 251-275 (2/1)
Labour seats banded, 276-300 (2/1)
Liberal Democrat seats, above 25.5, (evens)
Majority, none (1/6)
UKIP seats, above 3.5 (evens).
* - these bets are directly competing against each other, so they both cannot win. Therefore the stake placed needs to be adjusted so that the return is the same no matter which one wins. For the odds listed, this is 36.8% on the 251-275 bracket, and 63.2% on the 276-300 bracket. The banded Labour seat bets are also in competition with each other, but as their price is the same, so is the stake.
Oddschecker is the best place to look to get an overview of all odds available on a market.
As a demonstration of faith in my predictions, I have personally placed £100 (£2 per bet) on the recommendations, and will be reviewing the profit/loss of these bets after the vote.
Conservative seats, below 286.5 (10/11)
Conservative seats banded, 251-275 (7/2)*
Conservative seats banded, 276-300 (13/8)*
Labour seats, above 268.5 (8/11)
Labour seats banded, 251-275 (2/1)
Labour seats banded, 276-300 (2/1)
Liberal Democrat seats, above 25.5, (evens)
Majority, none (1/6)
UKIP seats, above 3.5 (evens).
* - these bets are directly competing against each other, so they both cannot win. Therefore the stake placed needs to be adjusted so that the return is the same no matter which one wins. For the odds listed, this is 36.8% on the 251-275 bracket, and 63.2% on the 276-300 bracket. The banded Labour seat bets are also in competition with each other, but as their price is the same, so is the stake.
Oddschecker is the best place to look to get an overview of all odds available on a market.
As a demonstration of faith in my predictions, I have personally placed £100 (£2 per bet) on the recommendations, and will be reviewing the profit/loss of these bets after the vote.
General Election articles
Previous: General Election Simulator - Easter Update
Next: Dissecting Scotland
Previous: General Election Simulator - Easter Update
Next: Dissecting Scotland