Price Pain

My approach to 'Conditional' bets prompted an email from one subscriber, who wrote:

Hope you are well. I've got another point to raise on your conditional selections, just an argument you might want to address. I must stress, I have no issues with the way you already do things and am happy with the service.

You say subscribers should be profitable on the conditional selections and if 3.15 can be obtained, then that is value. That may be true, but quite possibly, 3.15 would not be high enough to yield a profit on conditional picks long term. I'm sure you've got some evidence to back up that 3.15 is sufficient, though. It's quite possible for a subscriber of yours to look at your conditional picks, go out and secure prices of 3.15+, and end up with a lot worse P+L than your official results show. If I managed to back all your conditional picks at 3.15 this season, or even 3.2, along with your official picks, how different would my P+L be to your official results? These conditional picks could make a big loss, possibly due to variance. If a potential subscriber looked at your official results for the season, which are so much better than one of your subscribers who managed to secure 3.15 on conditional picks that made a loss, is that possibly misleading?

I am not sure if the conditional picks should be included in the email at all, that's just my opinion, not that it matters either way to me. I check Pinny on Friday afternoon and if the conditional picks are not at least 3.15, I don't back them at any price.

The subscriber is quite correct in saying that 3.15 may not be value. No price can ever guarantee a profit since as in a season such as the last (2013-14), draws were simply short in supply, so it is true that 3.15 is a best estimate of likely profitability based on historical results.

As I have mentioned before, I doubt that at the end of a season, any two subscribers will have exactly the same results. Having warned that 3.15 is borderline, I imagine that many people either omit selections anywhere near this price, or reduce their stakes. The idea of including them as Conditional picks is to let subscribers know that these matches historically are value draw selections, but historically draw prices have not been in the 3.0x range. Any subscriber backing at exactly, or close to, 3.15 does so in the knowledge that they are in a sense, on their own on this one, if Pinnacle's price doesn't meet that threshold.

As for "I am not sure if the conditional picks should be included in the email at all" my answer is really as simple as 'what harm does it do'? For Premium Charge payers, those 3.15 or close bets might be useful; for other more risk averse subscribers they may not be. at least by including them, the subscriber has a choice, which is generally a good thing!

Dmitri had some observations regarding the recording of prices, noting that some "use fair prices collected just BEFORE the email is sent". As he says, using old prices is not a fair way of recording prices, especially if they are in relatively illiquid markets. Dmitri writes:

To see the effect of using your method of recording (Pin prices on Friday) compare for example Combo’s official result of 9.84% with the 6.04% for Pin prices. Or Graeme's best system 31 with 8.40% down to 4.59%.

You are betting on the top divisions and are not influenced by odds movement, down from 11.95% to 11.46%.

I can see the problem from both sides, but as a punter I prefer your recording method.

Just a clarification here, that Dmitri is comparing my prices at the time of the email, versus my official price which is the independently taken Football Data Pinnacle price. My email price is given as a courtesy, along with my Cassini price and the current edge, but the recorded price is the minimum price that subscribers should be expected to be able to obtain, and is usually beatable. For example, this weekend the Newsletter was sent out on Tuesday morning, and gave the following for the Value Selections:

 

The official prices are yet to be published as I write, but subscribers have what I considered to be the 'true' price for the selections, and have several days to shop around for a price at which they are comfortable. The Pinny price was from Tuesday morning, and prices do move, even in the deep waters of the top leagues. It is mildly annoying when a (winning) selection is recorded at a significantly lower price than at the email time, but it's a small price to pay for the benefit of using an independently recorded price.

On the subject of the timing of releasing bets, my thoughts are that for a number of reasons, it is surely better to release them sooner rather than later. The more time subscribers have to place, or not place, bets the better. For anyone who is not full-time, it's also not always convenient to send out an email at a pre-determined time. Unless selections are based on individual player metrics and need to wait for injury / team selection news there is simply no need to wait. Send the bets out, with the price that is considered 'true' and give subscribers as much time as possible to find a price that they are comfortable with. Releasing selections at a set time late in the week, especially in soft markets, simply leads to crashes in price and little time for them to bounce back.

On the subject of 'true' prices, Dmitri observes:

If you as many punters do always use level staking this number is irrelevant. Then you are only interested in beating the bookie.

But if you are more advanced and use some sort of variable staking like fractional Kelly, then value is very important. On average these services are overestimating the value on their selections and this in turn will influence a bettor using variable staking negatively.

I do record my selections using level staking, but as a Kelly man, my real-world bets are not level. Dmitri continues, with a rather depressing conclusion:

Regarding minimum odds I would never consider signing up for a service that did not provide these prices. They represent the models estimate of fair prices.

Of course no model can get it right, but on a very large sample the profit of proofing to min odds should be about a round zero.

This is actually not the case. The profit ranges from -3.43% to -7.15%.

I can’t remember any service with a positive profit!

 We're all useless!  The true price hurts!