I think your right on many points Captain Sensible, but also that you are missing some, i'm going to go straight to it, two points:
Commission MattersIn a commission free world you are right, there is only one true / fair price so if you take two opposing bets on a selection at different prices one has value by definition (or is a fair bet) and the other doesn't (or is a fair bet).
When you introduce commissions the picture changes.
A losing bet, that has an expected value of say -1% (and if you analyze your PL it shows you losing that 1%), will bring you value because it avoids commission. This is similar to some situations with taxes, where a losing proposition may save more tax than the loses on the deal. This is an example, not one from a book or pure theoretical, one that my bots encounter daily:
You find a value bet (i agree with you that you have to find value, just the definition of value that i have a different view) on a selection that is trading at 1.50 and you back it, You estimate it's fair value is around 1.45. If the selection is available to lay at 1.47, in a commission free world, you should not lay it because it has negative expected value. If it trades at 1.45, then it's a fair bet, and you should lay it because it releases funds and it ends the opportunity cost of that money.
But if you pay commission you should lay it at 1.47. Although that lay bet will appear to have a negative expected value (and it has) it will save you costs (commission).
Scenario 1Back 100£ at 1.50, with a 5% commission your net odd is ((1.50-1) * (1-5%)) + 1 = 1.475
The Expected Value of that bet is 1,72£
Scenario 2Back 100£ at 1.50, lay 102,04£ at 1.47, with a commission of 5% you make 2,04£ on that trade.
If you then analyze your PL you will find that the lay at 1.47, without accounting for commission, destroyed value because the true / fair price was at 1.45. But if you account for commission the conclusion is the other way around.
An extreme but realistic example: A selection has a true / fair value of 2.00. With a 5% commission everyone that backs under 2.053 and everyone that lays above 1.950 is losing money if they hold the bet to the end. Someone who trades the spread and makes multiple back trades at 2.02 and closes at 1.98, all losing bets on a hold to maturity view (negative expected value by their own because of commission), will make money.
PS: Premium charges may distort this conclusion.
You may not care where the value is (at least in the way you look at your PL)People make money discovering fundamentally miss priced securities or bets, but they also make money modeling people's behavior. In financial markets many of the largest hedge funds in the world do this, firms like Renaissance, Winton, Two Sigma, AQR, AHL, QIM, etc, hire mainly physics, engineers, and not economists, because they make their money mainly by uncovering statistical patterns.
A simple example may be a short term trend following CTA. The fund may have discovered that investors, on average, sell their holdings on a Friday one hour before the market closes because they don't want to carry the risk overnight thru the weekend. So the fund knows that if they sell two hours earlier and buy at the market close they make money (they are exploring this herding effect, not caring about the fundamental value of the securities they trade). They make this for the 500 stocks that are part of the S&P 500 every week and it proves to be profitable.
After some months they dig in the trades (the PL) to try to improve and get some insights. The first trade is always a sell so they check where the stock traded one month after in an effort to access the fundamental value of the stock at that time. They see that 75% of the stocks went up and 25% went down. So they conclude that 75% of the first trades they put in were bad bets, value destroying trades on a fundamental point o view, and the value creating leg of the trade was when they bought the stock at market close.
Although this is true, it doesn't help in any sense to improve the strategy. They don't want to know which of the legs of the trade was right or wrong on a fundamental value perspective, they just want to know if the herding behavior is persistent.
On a sports market exchange the true value of the securities (bets) is more explicit because they are, in the end, when the Schrodinger cats reveals if they are dead or alive, discovered. The price will always be found at the end and is immune to investors / bettors opinion.
But before that, the same rules apply. Imagine that you analyze the historical data patterns and discover that bettors bet on the outcomes they want to see and this creates a behavioral bias around the goals markets in soccer. They herd to drive the price of the Over 2.5 Goals lower, independently of where the price is at (they want to see a lot of goals so they bet on that outcome), Further tests show that the probability of the Over 2.5 Goals odd going down on the 2 hours before the off is very high. So you tell your obedient bot to back every Over 2.5 Goals selection two hours before the off and close the trade right before the off.
At the end of the year the strategy proves to be very profitable but you decide you want to improve it. You analyze the PL and break it by back and lay, opening and closing trade. You see that some times the back bet creates value meaning that the odd was higher than it should be but other times it was the other way around. In some games the back bet was lower than it should be and the opening trade was value destroying.
In aggregate terms the back and lay may prove to be tilted to one side or not but this information, in this particular case, wasn't useful. Although there was in fact always a value creating and a value destroying bet the strategy didn't depend on that, your were paid for anticipating the bettors behavior.
If you don't agree on something please lets discuss, it's very interesting.
As a side point, i'm not just talking from books, i'm fortunate to be successful in sports and financial markets applying this kind of rationals and i do make a living from both. i have had the good fortune of visiting and talking to some of those hedge funds i mention above and learn from them and they do make a lot of money exploiting this kind of patters.
I do agree with this in general terms:
If we take things back to basics we'll only win long term gambling if we obtain value, the same applies when trading, we need at least one side of the trade to be value in order to have a chance of profiting. Obviously our optimum aim would be to have both sides of the trade value bets as we'd then profit from our lay bets and our back bets, simply suming your bethistory will show you if your betting is unbalanced and to what degree you're leaking potential profits.
Just this one that i don't fully agree:
Obviously from a bank management point of view people may well want to close but that's a different matter. Most people with bots won't place more than 1 opening and closing bet per runner and it's likely there effort at finding value is in the opening bet with the closing bet having little thought other than to close the position nd ultimately leaking value. Again a simple summing of opening bets and closing bets will higlight the fact for most botters.
Sorry for the long long post, i'm passionate about this issues.