What is a good sample of bets

4 figures+  and 12 months

Cheers Delta. Was starting to think GB was just tumbleweed

I can do the 12 months no problem, but 4 figures may take quite a while. 16 bets in 5 days so far. Though I suppose that isn't too far off the mark.

Seen "clear" profitable patterns last many hundreds of bets and then turn on a dime and head south for no clear reason.  Seen "clear" edges turn bad as soon as my bet size became a little bit over the minimum - so smart punters started picking any bet that was poor value and avoiding the value ones just enough to skew the results from a few per cent profit after commission to a few per cent loss.

Low volume makes you less visible to the smart punters but takes a long time to see if the edge is real.  How you get matched matters a lot.  If you control this it is better.  If you wait to get matched passively you will be picked off selectively by the knowledgeable.  Trust form reading edges more than statistical ones.

Try randomising 1000 hypothetical bets  with wins and losses at say 2 to 1 odds using random numbers to create wins and losses of say a 5% edge.  Graph the results the recalculate and watch the graph dance.  Edges will appear and disappear before you eyes due to randomness.  A massive edge will be clear with hundreds of bets.  A thin apparent edge cannot be trusted until you have thousands.

Commission is the killer though - I've found quite a few edges with 2 or 3 %  in my favour (no mean feat in a tough market) but the steady drain of commission turned them into a losing approach.  If you analyse past betfair data you will find pots of gold everywhere in statistical patterns.  Most are illusions - just wrinkles in the randomness. But even the real ones once you put them into practice the market so often it matches you in ways that weight the result to the poorer value bets.

48 bets in so far. Only 952 to go...

still winning though ?

Oh yeah, but it's a ridiculous return right now. Just waiting for the inevitable balancing to happen!

if its ridiculous then a change isnt far off. take it steady,but hold your nerve when it does happen and keep plodding on.The Universe will try to put you off ! good luck

It depends on what you are looking at and if there is
logic in the stats.
For instance if, say Chelsea were playing say, Stoke at
home, and over the last 100 games at home to teams out
side the top six, they had won xxxxx matches and were
available at a better price than the stats indicate, then
the sample IMO would be big enough. Are your bets similar
to that?

Have to disagree to a certain extent with Delta and zooot. It is not always advisable to pick an arbitrary figure out of the air and say you must have that number as a sample size. One size does not fit all and betting and systems don't always work like that.

From personal experience; I had a very profitable system from 2002-2007 when certain market forces started to make it less profitable and I began to lose the edge I had. Then other changes meant by 2009 it began and currently does make a small loss (I don't use it anymore). It averaged at one bet a week and had I waited to sample it over 1000 bets then I would have missed out on £1000's. In the grand scheme of things over the life history of that particular sport (or even a few thousand events) the years 2000-2007 (backtested for 2 years) would make it seem like it was just an illusion or random statistical event but this wasn't the case, there were very real reasons why it worked then and why it doesn't now.

What I am saying is that you have to be careful not to over sample a system and miss an opportunity. I have had and still use several different systems and they often need to be tweaked or revised or gotten rid of completely as things change within various sports and within the betting markets themselves. Past statistical analysis is all very well and good but it only works guaranteed one way and that's backwards. The example I used above was probably one of my best systems and I still hold some hope that it may well be applicable again at some point in the future. Although it is certainly not the only one that eventually failed. In my opinion there are very few systems that would last forever without being tweaked in some way, if any, sport and the world just don't work like that.

That said if you can get a large sample size then it is a good idea to do so, if it can be done relatively quickly. Something like horse racing I usually spend a minimum of 3-6 months of paper trading, with a further 3 months at minimum stakes. It also allows you to adapt or tweak the system if necessary and if it doesn't work then you have only lost a small amount (as well as your time).

ken - I realize you're mainly a football guy, and what I am doing wouldn't really fit with that example.

Tiger - yeah I have already reached a big losing streak, but as my stake is just 2 quid a time there is no chance of me losing much. Lost 10 of the last 14.

Trousers - thanks for your input. I can see what you mean about different sample sizes fitting better for some systems rather than others. I'm not really one for paper trading,so have done this with 2 quid a bet.

Not that it means much yet, but these are the results so far:

Bets = 54. W = 33 L = 21 P/L = 34.99 after comm

Here is an article on testing systems

http://web.archive.org/web/20041009202829
/http://www.0dd5.com/matharchie.shtml

You don't need a sample of thousands

ok.

Hows Hank Jr doing mate?

Did I see posted somewhere, that you said it was a crucial year coming up?

All the best to you both.

Dave - thanks for that I'll take a good look when I have a moment.

kenbo - thanks mate, new update on chit chat.


100 bets on the board now.

Bets = 100 W = 55 L = 45 P/L = 17.58 after comm

the next 100 could very easily result in a loss of 17.58 or even more.

well done so far and keep plodding on.

TTT is correct of course, as playing, say roulette,
any one of the evens chances could come up 55+ from
a 100 spins and would mean nothing, nevertheless GL
with what you are doing.

TTT and ken. Absolutely correct that the reverse could be true for the next batch of bets, but I believe I have something better than a toss of a coin chance. We'll see. I could well be wrong but I will keep with this as I think I am getting good value. Appreciate all input though.

Your figures suggest that your average odds are around 2.05 (depends on your comm rate).

I don't know much about stats so this may be wrong but I think you can assess this quite easily to get a rough idea of your situation using a crude methodology:

Assuming only one bet in each market and a comm rate of 4% the standard deviation on the total return on 100 back bets with zero expected value at 2.05 for £2 a throw is about £19. Without getting into anything more sophisticated a useful rule of thumb is that you want to be two standard deviations away from the mean before you ascribe any significance to a result. In this case the mean would be about -£4 because of comm.

Therefore after 100 bets you should expect to be up by more than £34 before you can have significant confidence that you're probably doing better than chance.

At 200 bets 2*standard deviation + mean would be £51 and at 1000 it would be £83.

So those are your benchmarks: £51 after 200 bets, £83 after 1000.

You might notice that as the number of bets goes up the target gets less onerous, in simple terms that's because after a lot of bets there has been more time for good and bad bits of luck to even out.

Like I said though, I'm not really into stats so I stand to be corrected.

There is an excel spreadsheet here http://www.sportspreadsheet.com/Betting_Calculators.html that uses the chi-squared test.  You just need to put in some simple info

https://docs.google.com/file/d/
0Byg8Wbbo5p2uaDlzUWg5cmRQUUE/edit?usp=sharing