For about 3-4 months, I have been experimenting (paper trading) with a very specific trading strategy. To be more precise, it's not even "trading" in the strict sense of the word. In a nutshell, my strategy consists of entering a market when a certain event occurs, REGARDLESS OF ODDS AT THE TIME, and keeping my position open till the final whistle. So it's a "bet" in that sense.
Over this time period, I have tested it on over 1800 football matches in total, across various leagues and competitions in the world - from Premier League to some obscure leagues in Africa. My calculations show that I'd be in a healthy profit over 1800 matches. Further, for the last 350 matches (out of those 1800), I started trading with actual money - placing very small stakes, and this has generated a positive return so far.
What is strange to me, is that this strategy seems to apply to vast, vast majority of the games where these certain conditions occur. Essentially, these 1800 matches were chosen almost at random by me (just as long as above mentioned event happens), more or less regardless of inplay stats and deemed relative strength of the teams, etc.
It seems that I am systematically beating the market odds - but we all know that betfair is very efficient for the most part.
So my question is: how do I know if I have genuinely stumbled upon an edge that gives +EV? Or is my success just a result of statistical variance and if I kept up this experiment for longer, say 10,000 matches, it would all average out and my net profit / loss would come close to 0?
Thank you.
Do I have a genuine edge or is this just statistical variance?
-
- Posts: 12
- Joined: Sat Jan 12, 2019 2:26 pm
Your post begs the question, what is the specific event? Multiple people testing your theory would speed up the process of determining whether you have indeed found a genuine edge or it is just "statistical variance".
-
- Posts: 3140
- Joined: Sun Jan 31, 2010 8:06 pm
There are many edges on Betfair that have been running for years unexploited, only way you'll know for sure is when you put your money into the market as that's the one variable you can't account for simply looking at stats. You're entering the market when a certain situation occurs and it's likely others are entering on that same 'trigger' and giving away value for whatever reason the thing to look at now is just how much you can exploit that edge. You may find the edge disappears when you up your stakes to reasonable amounts.
It’s all about the odds.
https://www.pinnacle.com/en/betting-art ... 6ux3b552bg
https://www.pinnacle.com/en/betting-art ... f7s22scgmz
If you are laying high or backing low then you need a lot more matches to confirm it is profitable, what are your average odds?
https://www.pinnacle.com/en/betting-art ... 6ux3b552bg
https://www.pinnacle.com/en/betting-art ... f7s22scgmz
If you are laying high or backing low then you need a lot more matches to confirm it is profitable, what are your average odds?
*laying low or backing high.
Otherwise I agree, variance depends a lot on the average odds.
This is also a good article:
https://www.pinnacle.com/en/betting-art ... SEAMWSY89K
Otherwise I agree, variance depends a lot on the average odds.
This is also a good article:
https://www.pinnacle.com/en/betting-art ... SEAMWSY89K
I am very hungover today but i think all four count.pythonic wrote: ↑Fri May 03, 2019 10:28 am*laying low or backing high.
Otherwise I agree, variance depends a lot on the average odds.
This is also a good article:
https://www.pinnacle.com/en/betting-art ... SEAMWSY89K
But it works the opposite way as well, the higher the chance to win a single bet, the more bets you need to even out randomness.
I can lay all horses at 1000 and make money for months before I start to see a negative pnl.
I admit that it is tricky but I am pretty sure that for a constant positive edge you will minimize variance by adjusting your stake after the odds like in the Kelly formula (stake more when backing shorter odds, keeping the amount to win constant).
Of course, If your edge turns out to be negative, you would also lose more.
Of course, If your edge turns out to be negative, you would also lose more.
Like Joseph Buchdahl writes in the article:
"Unsurprisingly, the longer the odds the greater the variance or spread in outcomes. Of course, for these expected break-even scenarios the variance (or square of the standard deviation) is directly proportional to the odds minus 1.
Betting at longer odds will mean you have a greater chance of doing a lot better than expected just because of good luck (the distribution tails are fatter at higher yields). Of course, the reverse, is unfortunately also true since the distributions are symmetrical."
"Unsurprisingly, the longer the odds the greater the variance or spread in outcomes. Of course, for these expected break-even scenarios the variance (or square of the standard deviation) is directly proportional to the odds minus 1.
Betting at longer odds will mean you have a greater chance of doing a lot better than expected just because of good luck (the distribution tails are fatter at higher yields). Of course, the reverse, is unfortunately also true since the distributions are symmetrical."
- ShaunWhite
- Posts: 9731
- Joined: Sat Sep 03, 2016 3:42 am
Who needs this Joseph Buchdahl geezer, he seems surplus to requirements while we have pythonic, LinusP around (to name but a few)pythonic wrote: ↑Fri May 03, 2019 12:29 pmLike Joseph Buchdahl writes in the article:
"Unsurprisingly, the longer the odds the greater the variance or spread in outcomes. Of course, for these expected break-even scenarios the variance (or square of the standard deviation) is directly proportional to the odds minus 1.
Betting at longer odds will mean you have a greater chance of doing a lot better than expected just because of good luck (the distribution tails are fatter at higher yields). Of course, the reverse, is unfortunately also true since the distributions are symmetrical."
It's funny how once you have a 'name' you can get plaudits (and paid!) for writing articles stating the bleedin' obvious. That's not a dig at you for reproducing it, far from it, i'm sure there's lots who hadn't twigged what that abstract explains or gone the extra step and figured out the maths, or just needed a big name to confirm what's already been said on this thread.
But he really is just saying that for a given return, your income stream will be more predictable if you sell beans than it would be if you sell boats. At least he shows the reader some respect by starting sentances with 1 x "Unsuprisingly" and 2 x "Of course" which makes a nice change.
Have you read his book? Its hard to recommend due to the underlying feeling that he believes making money from gambling is impossible and anyone who says they are is a lier but then thats why he is probably writing rather than betting..ShaunWhite wrote: ↑Fri May 03, 2019 2:35 pmWho needs this Joseph Buchdahl geezer, he seems surplus to requirements while we have pythonic, LinusP around (to name but a few)pythonic wrote: ↑Fri May 03, 2019 12:29 pmLike Joseph Buchdahl writes in the article:
"Unsurprisingly, the longer the odds the greater the variance or spread in outcomes. Of course, for these expected break-even scenarios the variance (or square of the standard deviation) is directly proportional to the odds minus 1.
Betting at longer odds will mean you have a greater chance of doing a lot better than expected just because of good luck (the distribution tails are fatter at higher yields). Of course, the reverse, is unfortunately also true since the distributions are symmetrical."
It's funny how once you have a 'name' you can get plaudits (and paid!) for writing articles stating the bleedin' obvious. That's not a dig at you for reproducing it, far from it, i'm sure there's lots who hadn't twigged what that abstract explains or gone the extra step and figured out the maths, or just needed a big name to confirm what's already been said on this thread.
But he really is just saying that for a given return, your income stream will be more predictable if you sell beans than it would be if you sell boats. At least he shows the reader some respect by starting sentances with 1 x "Unsuprisingly" and 2 x "Of course" which makes a nice change.
- ShaunWhite
- Posts: 9731
- Joined: Sat Sep 03, 2016 3:42 am
I haven't, but those unused audible credits keep stacking up so i might.
But it's hard to believe that any educated person can't understand the concept of value when they make a bet, so maybe I won't
-
- Posts: 61
- Joined: Wed Feb 20, 2019 10:01 pm
Thanks for all your replies.
spreadbetting - it seems to me that my edge, if it does exist, comes from that in certain matches (usually occuring in small less known leagues), the given odds are consistently not indicative of the true chances of the event happening. As I wrote in my first post, I have bet across all sorts of leagues, and it seems to me (I don't have a large enough sample) that some value comes from me placing these bets in these obscure leagues, where the total matched volume is quite small. This naturally brings up the point you made - can this strategy be scaled up to yield very large profits.
LinusP: the average implied probability from the odds that betfair gives me at the time I enter the market is around 12%. The spread is very large, though. Over my 1800 bets, the lowest probability I took was around 1%, and the highest was around 75%.
Given the average taken odds imply that the chances are around 12%, how many matches would I need to confirm my strategy is profitable?
spreadbetting - it seems to me that my edge, if it does exist, comes from that in certain matches (usually occuring in small less known leagues), the given odds are consistently not indicative of the true chances of the event happening. As I wrote in my first post, I have bet across all sorts of leagues, and it seems to me (I don't have a large enough sample) that some value comes from me placing these bets in these obscure leagues, where the total matched volume is quite small. This naturally brings up the point you made - can this strategy be scaled up to yield very large profits.
LinusP wrote: ↑Fri May 03, 2019 7:55 amIt’s all about the odds.
https://www.pinnacle.com/en/betting-art ... 6ux3b552bg
https://www.pinnacle.com/en/betting-art ... f7s22scgmz
If you are laying high or backing low then you need a lot more matches to confirm it is profitable, what are your average odds?
LinusP: the average implied probability from the odds that betfair gives me at the time I enter the market is around 12%. The spread is very large, though. Over my 1800 bets, the lowest probability I took was around 1%, and the highest was around 75%.
Given the average taken odds imply that the chances are around 12%, how many matches would I need to confirm my strategy is profitable?
- ShaunWhite
- Posts: 9731
- Joined: Sat Sep 03, 2016 3:42 am
Sample sizes may be chosen in several ways:Speculator_3 wrote: ↑Sat May 04, 2019 12:41 amGiven the average taken odds imply that the chances are around 12%, how many matches would I need to confirm my strategy is profitable?
1. Using experience – small samples, though sometimes unavoidable, can result in wide confidence intervals and risk of errors in statistical hypothesis testing.
2. Using a target variance for an estimate to be derived from the sample eventually obtained, i.e. if a high precision is required (narrow confidence interval) this translates to a low target variance of the estimator.
3. Using a target for the power of a statistical test to be applied once the sample is collected.
4. Using a confidence level, i.e. the larger the required confidence level, the larger the sample size (given a constant precision requirement).
You won't be suprised to know that I didn't know that by heart, I grabbed that from the Wiki page on sample size determination but I've never been able to understand it fully. We seem suprisingly short of mathematicians on here who can put it into a formula with just the parameters necessary for what we do. 100s of years of experience kicking around on here and nobody knows any semi-serious maths, it's strange. Either that or they're not the sort who help much.
Being constructive though I think a good place to start is 100 of the least likely outcome...ish?
200 @ 50%
400 @ 25%
800 @ 12% etc
But there's clearly a lot more to it than that. "As soon as the charts look reasonable" is my other goto on sample size and I suspect that goes under #1 above.