False edge? Am i fooled by variance?

User avatar
jimibt
Posts: 3641
Joined: Mon Nov 30, 2015 6:42 pm
Location: Narnia

Jukebox wrote:
Mon Jun 29, 2020 12:22 pm
offlimit88 wrote:
Mon Jun 29, 2020 10:57 am

If you toss a coin 120 times, what is the chance that you have 70% heads and 30% tails?
A way to find out:
First set up a spreadsheet with the 1,329,227,995,784,915,872,903,807,060,280,344,576 different possible outcomes of 120 consecutive single coin tosses 'P' - then count up the number of of those outcomes that have exactly 84 heads 'n'. Then simply calculate n/P
OR... you could use statistical calculations (if you don't want to wait for excel 2050) :D

https://www.universalclass.com/articles ... tcomes.htm
offlimit88
Posts: 68
Joined: Mon Feb 25, 2019 2:29 pm

Wow this is going to be interesting :D
Is there anyone smarter than me who can figure this calculation out? :D
User avatar
jimibt
Posts: 3641
Joined: Mon Nov 30, 2015 6:42 pm
Location: Narnia

offlimit88 wrote:
Mon Jun 29, 2020 2:35 pm
Wow this is going to be interesting :D
Is there anyone smarter than me who can figure this calculation out? :D
lol - if you read the article i linked, then the 2nd or 3rd worked example is almost exactly what you're after

https://www.universalclass.com/articles ... tcomes.htm
User avatar
ruthlessimon
Posts: 2094
Joined: Wed Mar 23, 2016 3:54 pm

ShaunWhite wrote:
Sun Jun 28, 2020 4:37 pm
If you see an anomaly in the data then without a real world reason it's a prime candidate for reversion.
Hence why modelled strategies are powerful. If it works long-term (good sample) & the drawdowns are manageable, the real world reasons come for free.

https://financial-hacker.com/build-bett ... d-systems/
User avatar
ShaunWhite
Posts: 9731
Joined: Sat Sep 03, 2016 3:42 am

ruthlessimon wrote:
Mon Jun 29, 2020 2:56 pm
ShaunWhite wrote:
Sun Jun 28, 2020 4:37 pm
If you see an anomaly in the data then without a real world reason it's a prime candidate for reversion.
Hence why modelled strategies are powerful. If it works long-term (good sample) & the drawdowns are manageable, the real world reasons come for free.

https://financial-hacker.com/build-bett ... d-systems/
Nice link.
Perhaps if people spent as long learning about this stuff as they do reading TiTZ, they wouldn't need to read TiTZ. It's more related to another thread but top traders don't struggle with psychology, not because they're mastered their mind but because they've mastered their craft.
User avatar
ruthlessimon
Posts: 2094
Joined: Wed Mar 23, 2016 3:54 pm

ShaunWhite wrote:
Mon Jun 29, 2020 4:53 pm
Perhaps if people spent as long learning about this stuff as they do reading TiTZ, they wouldn't need to read TiTZ
I thought that was a porn mag for a sec :mrgreen:
offlimit88
Posts: 68
Joined: Mon Feb 25, 2019 2:29 pm

I give up. Waiting for reversion
lol
offlimit88
Posts: 68
Joined: Mon Feb 25, 2019 2:29 pm

ShaunWhite wrote:
Mon Jun 29, 2020 4:53 pm
ruthlessimon wrote:
Mon Jun 29, 2020 2:56 pm
ShaunWhite wrote:
Sun Jun 28, 2020 4:37 pm
If you see an anomaly in the data then without a real world reason it's a prime candidate for reversion.
Hence why modelled strategies are powerful. If it works long-term (good sample) & the drawdowns are manageable, the real world reasons come for free.

https://financial-hacker.com/build-bett ... d-systems/
Nice link.
Perhaps if people spent as long learning about this stuff as they do reading TiTZ, they wouldn't need to read TiTZ. It's more related to another thread but top traders don't struggle with psychology, not because they're mastered their mind but because they've mastered their craft.
Titz are two, easy math, i love titz
User avatar
jamesedwards
Posts: 2234
Joined: Wed Nov 21, 2018 6:16 pm

jameegray1 wrote:
Sun Jun 28, 2020 8:41 pm
ShaunWhite wrote:
Sun Jun 28, 2020 5:32 am

Problem is James that charts are like fractals, you don't know what zoom level you're looking at. That dip might be almost invisble after a year, and you might not know what happened before it started. How does it look if you put a trendline on that? origin zero. I'm guessing it's still +ve.
Crazyskier wrote:
Sun Jun 28, 2020 7:28 pm

Reversion to mean can indeed be a bitch!

LOL. I've set it running again on lowest possible stake. Will keep you informed ;)
Calling ShaunWhite. Can I stop it now? ;)
latest v3.PNG
You do not have the required permissions to view the files attached to this post.
User avatar
gazuty
Posts: 2547
Joined: Sun Jun 26, 2011 11:03 am
Location: Green land :)

Jukebox wrote:
Mon Jun 29, 2020 12:22 pm
offlimit88 wrote:
Mon Jun 29, 2020 10:57 am

If you toss a coin 120 times, what is the chance that you have 70% heads and 30% tails?
A way to find out:
First set up a spreadsheet with the 1,329,227,995,784,915,872,903,807,060,280,344,576 different possible outcomes of 120 consecutive single coin tosses 'P' - then count up the number of of those outcomes that have exactly 84 heads 'n'. Then simply calculate n/P
https://stattrek.com/online-calculator/binomial.aspx
Capture.PNG
You do not have the required permissions to view the files attached to this post.
offlimit88
Posts: 68
Joined: Mon Feb 25, 2019 2:29 pm

Thank you very much,very helpful. So , if i understand correctly, it's quite impossible to have such a strike rate with just plain luck. it is like 0,0004 % of chanches.
Edit : for my specific case it s a bit more 0.02%, still relevant

There has to be more in here
Emmson
Posts: 3363
Joined: Mon Feb 29, 2016 6:47 pm

"There is such a thing as 'the law of big numbers'.

It means that the more times the same event is repeated, the more actual results stochastically converge on expected results.

Imagine a fair coin toss. The law of large numbers predicts that after 10,000 iterations (plays of the same event) actual results are more likely to have heads coming up 50% of the time than after 100 iterations.

The problem is that in sports you don't get the same idealized event replayed countless times. In fact each event is individual--player injuries, good or bad form etc., circumstances of the match, confidence or mood etc. " askari1

https://en.wikipedia.org/wiki/Law_of_large_numbers

"There are 2 kinds of people on BF. Those who bet based on what they anticipate the result to be, and those who bet against what other people anticipate the result to be. " pxb

I thought stochastically may have been a misspelling but no it is an actual word. :D
User avatar
ShaunWhite
Posts: 9731
Joined: Sat Sep 03, 2016 3:42 am

Emmson wrote:
Wed Nov 08, 2023 12:40 am
"There is such a thing as 'the law of big numbers'.

It means that the more times the same event is repeated, the more actual results stochastically converge on expected results.
When you don't have a big enough sample to use Emmson's advice you can also look at Central Limit Theorem to see if your sample typical of what's expected. It's essentially the distribution of your results, the simplest being a histogram of your PLs. It's a way to try to see if you're currently winning or losing due to outliers or due to short term randomness.

It's tricky creating an example but your results distribution (the number of times you lost 0-5 or won 0-5, 6-10, 11-16 etc) should start to form a bell curve especially if you squint your eyes a bit, any notable gaps or peak 'could' mean you're due some in that region. But this is where LLN and CLT start to sound like they support the Gambler's Falacy but it doesn't. Prior results have no influence, the universe doesn't 'know' you're due a £50 win, but with a big enough sample you'll win as often as normalised distribution says you will.

Here's the rough 2min example, any band especially far from the distribution or way out at either end could mean your current sample isn't typical yet. Stats aren't really about yes or no they still need to be interpretted though.
Screenshot_4.jpg
You do not have the required permissions to view the files attached to this post.
User avatar
ShaunWhite
Posts: 9731
Joined: Sat Sep 03, 2016 3:42 am

I don't like posting anything that's wrong so i checked...#1 is quite funny.
Screenshot_5.jpg
...and it re-wrote it for us. I might get it to do all my posts, it's faster than I am ;)


When your sample size isn't large, you might still consider the Central Limit Theorem (CLT) to understand if your sample is representative of what's expected. The CLT deals with the distribution of sample means — for example, you might look at a histogram of your data points. This can help you determine if current variations are due to outliers or just short-term randomness.

Creating a sample distribution (the frequency of certain ranges of wins and losses) should reveal a pattern that resembles a bell curve, albeit not perfectly. Notable deviations from this pattern could indicate unusual results, but remember, this is not an endorsement of the Gambler's Fallacy. Prior results don't affect future outcomes; the universe doesn't compensate for past losses or wins. However, over a large number of trials, the Law of Large Numbers (LLN) assures that your results will align with the expected probability distribution.

Please note that a large enough sample size is crucial for the CLT to be applicable, and in practice, 'large enough' can vary depending on the actual distribution of your data.
You do not have the required permissions to view the files attached to this post.
Emmson
Posts: 3363
Joined: Mon Feb 29, 2016 6:47 pm

Little of what I posted is mine (Only the sto·chas·tic sentence) look closely and see the quotation marks, I even named the sources :D


This is where I pulled it from

https://community.betfair.com/general_b ... f-averages

it just caught my eye and that is why I brung it here
Post Reply

Return to “Tennis Trading”