Wednesday, 16 March 2011

Playing The Percentages

Just as I was struggling for something to say, Mouldhouse helps me out with a comment...

It's a bit pernickity (if that's even how you spell it) but I was always taught that the level of accuracy should fit the context, as far as mathematics is concerned.

For example, if you had 10,000 matches in a sample, and 3,667 of them had ended in a draw, the statistic "I have a 36.67% strike rate" would be more than acceptable.

If you've got a sample size of 30, let us say, and 11 of them have ended in draws (or indeed now a sample of 31, and 12 have ended in draws, congrats, btw) then that level of accuracy really isn't justified; simply because the statistical significance - indeed, if that is the case, the probability that the results have been obtained by pure fluke and the system overall is a losing one is still 13.71% (assuming that all draws have hit at a price of 3.45, a big assumption I grant you!). Or indeed, more suited to this context, about one in 7.

Point being that strike rates to 2 d.p. over samples only just large enough to apply large-number estimates to is a bit of a misnomer - if my stabs above were correct, last night's result would make a difference of some 3%ish on the "advertised" strike rate, which is very volatile for one iteration.

The main purpose is not to be a total a$%ehole by pointing this out but hopefully to point out to some who may well risk their hard-earned following any of the posted picks that there might be a lot more luck involved than they would hope. I suppose its a downside of having a blog that posts picks without a concrete record, although I note gold-all-over has been updated after you nearly set a world record for getting bored of a blog :)

Incidentally, the strong draw picks posted on Gold-A-O have a nearly 1 in 3 chance of having been generated by pure fluke so far - but it says more about small sample sizes than anything else. I look forward to your witty riposte. I should also echo those who've praised it as the best betting blog on the internet - how you keep finding new subject matter that is actually interesting is beyond my comprehension, and isn't something I could do!
Not pernickety at all. I actually was never taught that, or certainly don't remember being taught that, the 'level of accuracy should fit the context', but it makes perfect sense to me and I shall endeavour, moving forward, to keep my decimal places appropriate.

The percentages quoted are actually those from the 2010-11 season, not just those since the start of the calendar year, which is some 1,364 matches.

Of those, 128 qualified as strong draws, of which 43 finished drawn - 33.59375% - (the other 85 lost to 90' goals!) with the average draw price being 3.432352941... (sorry, I just couldn't resist).

Yes, certainly not a big sample, and I hope no one is stupid enough to follow my picks without doing their own research as well, and being aware that these numbers are from this season only. The problem with a system finding winners at 3.4 is that long runs are inevitable - currently 8 of the last 9 have lost - but then along comes runs of 4 winners in 5, or 3 in 4 which more than compensate.

Gold All Over has a couple of problems. One is that blogger doesn't allow (at least to my knowledge) the display of tables, so the data looks really untidy. It's also time consuming, and I have enough trouble finding time to keep one betting blog going, never mind two, so while the idea isn't necessarily dead, don't expect too much unless I can work out a tidy format for it.

Anyway, I appreciate the advice on the number of decimal places to use.

Jamie commented on my thought that playing poker for matchsticks is useless.
Perhaps a little bit different, but I've found kind of the same thing with playing the poker application on my phone.

I play quite a fair bit online and in casino tournaments, and have had a moderate amount of success. But I've found when I play the (free) poker app on my phone, I make a lot of decisions that I would never make when playing for decent stakes.

I also found that when playing "free" poker online to sharpen my skills - I played a lot differently when the stakes were low or non-existent, and it was starting to mess up the skills I had honed by playing for proper stakes. I now don't play at all unless it is for decent stakes, because that's where I believe my best performances are.

I also trade pre-post horses, and have found the same thing... trading for minimum stakes is OK to learn the system or the software, but can be detrimental to your overall success once you know what you are doing.

Keep up the great work on the blog
With poker, such a big part of the game is 'reading' your opponent, and being sure that the hands are clean (i.e. no collusion between players) that I have never seen the appeal of playing online. I shouldn't say never, because I have tried it years ago for entertainment, but I soon tired of it, and when Betfair came along, that was a lot more entertaining.


Jamie said...

No arguments there mate... though

The analogy had less to do with online poker than with playing poker with competitors under what I would call my optimum playing field...

If I'm playing with 10 mates and a $10 buy in, I play a lot differently that I do when I am playing at a casino tournament with a $2,00 buy in and 100 people to get through.

I personally found the low stakes games were instilling bad habits in me, so I ditched them. I've found the same with trading. When I started trading the 2pm at Haydock with $5 backing and laying, I was quite happy to lose the lot.

Now that I am putting 4 figures on the same sort of trades, I find my entry and exit points and rationale for trading completely different...

Anyway, thanks for the mention. I agree completely that paper / matchstick / low stakes trading is a blunt tool in the long term trader's arsenal (Go the Gunners ;-)


Jamie said...

And re: the table issue - just convert the table to a jpeg and upload it like you do with any other graphic.