Reality based coin-tosser method that beats 95% of traders in the world.

The only rules that are "better" (in my book) are those that bring at least an extra tic per day on average over the original method, based on some sort of reality or at least empirical/anecdotal evidence.

Everything that you describe as "better" is simply a personal preference of yours. No offense.

You could say that I don't mind giving the beggar a daily coin, and my method is as random as it gets, and I don't need any backtesting crap for my method, because it is 50-50 minus commisions by definition. In other words reality does not need any backtesting "proof". It just is.

 

whisky


as we all know, a coin toss averages 50/50 outcomes in a large sample.

this is suboptimal because the market averages .52p up and .48p down in large sample. the result of using a 50/50 random trigger on a variable that exhibits a different distribution of outcomes is even more negative outcome, as the probability mismatch amplifies the negative expectancy.

in other words, using a 50/50 random process is by itself a way of introducing a bias in the decision process.

you should use a biased coin to reflect the actual market behaviour.

the reason the market (an index) has this behaviour and certainly will maintain it in the future is the following:

1. survivorship bias
2. inflation

so, why not using an adequate random process?


BTW excellent thread

 

Finally!. Someone wants to improve the coin. Any suggestions?.

I'll start: If you can elliminate 2 numbers in an European Roulette wheel and you play in a casino with no table limits, you can own the world in how many spins?.

Bipolar disorder can be turned into an asset for a trader. Welcome back to the thread.

 

Your coin toss will be subject to questions of fraud if the results were too positive. Also, if you wanted to cheat on the results, it would be very easy to do so. With the LottoBalls rules the entry times are defined before the market opened and the channel prices by exchange Time and Sales.... So not just 'personal preference' but for practical reasons.

Never offended on ET!!!

 

Point taken.

The main issue I have with this "potential improvement" is that it rules out any potential improvements in the coin-tossing algorithm.

 

... drill two holes in your coin, then sell it as a button for more than 100% of its denomination

 

You will have to prove to yourself and others that your suggestions will yield an extra tic per day average profit over the original method rules.

 

Coin toss is 50-50 - would be interested to see what would happen if you had a variable-bias coin - then set the bias to a measure of market's current direction.

(eg, ratio of up/down days over past 2 weeks, or 50 + %age net movement over past 1d/1w/1m etc, million and one ways to do this, across pretty much any timeframe).

That said, in my experience it is randomizing exit strategies that leads to the more illuminating results.

And no need to limit to one coin - computers are wonderful tools - use LOTS of coins controlling LOTS of parameters and monte carlo the hell out of it.

 

... can we see the data supporting your small loss claim for your original method rules?

 

Would you care to ellaborate or suggest some simple objective rules or process?.

Too much randomizing and Monte-Carlo lead to datamining troubles.