the end expected value for coin flipping is actually ruin not breakeven anyway probability of ruin for 2 person coin flip game is bankroll divided by total of bankrolls or something like that. the point is if you have a fair game you still shouldn't play cause the house has a much larger bankroll and you will lose all your money. and the 1-100 thing is misleading everyone. if you're choosing two, it's either larger or smaller. whether its larger by 1 or 25 makes no difference. in the case that they're random, this is still such a trivial problem.. and I fail to see any application to trading, but maybe that's why I can't trade for beans. to make it even clearer look at it this way. say you somehow get two numbers, either by choosing or randomly or whatever. one is higher than the other. you either show the higher one or the lower one, and those are your two choices, it doesn't make any difference what the numbers are as long as they're not 1 or 100
Successful traders do not approach the market with a random expectation and then decide to somehow manage the outcome with a great level of skill to the point of positive expectancy. Cute thread but of no practical value in the markets. Buying bids and selling offers consistently in a very low vol market is an edge. Modeling some sort of positive expectancy and being able to execute that trade in the market is an edge. Flipping a coin and attempting to translate any logical interpretation of that result into the market is not an edge.
I agree that coin flipping does not apply directly to trading but a lot of traders have applied trading to coin flipping
I must respectfully disagree. I provided two examples in this thread where 'coin-flipping' can be used to in a trading environment. Now, I am not saying the markets are random or a coin flip (ala heads market goes up, tails the market goes down). But, comparing one's results to a coin-flip distribution is insightful and useful. I showed a year of live trading resulted in nearly perfect 50-50, profit factor 1, results. And, as such, the profit curve and the daily profit probability mirrored a coin flip. Ever wonder why most systems created generally have ~50% win percentage and a profit factor of 1, it's because they are modeling/trading market noise. Systems that demonstrate notable positive or negative equity curves (ie what all automated traders desire) are hard to create as they are trading real edges not noise. I am more than comfortable saying my system was trading noise...Therefore, it was in my best interest to improve either the win% or the PF such that my results were no longer like a coin flip. That is one example where knowing the coin flip distribution and logic is useful. Second, I showed 3 plots of possible equity curves using a coin flip, even money simulator (when discussing the arcsine law). The point here is that a 50-50, PF 1 system can also have equity curves that appear to be either very good (ie steady slope to the upper right) or poor (steady slope to the lower right) in addition to the expected sawtooth about the 0 line curve. Again, knowing that these distributions are possible given a even-money system, one would be fool-hearty to trade a system without proving to oneself that the back tested/forward tested results are actually better than a coin-flip. So, in my humble opinion, the topic is valid for system development and critique, not necessarily so for market price action. Just a few cents tossed your way, Masterjaz
Couldn't of said it better myself. Thanks for taking the time to post your stuff. Many here seem to believe that randomness is something to trivialize and disregard. The monty hall paradox and the secretary problem are two classic examples that most shrugg off either due to the difficult math involved, or, by thinking these problems add little value to their trading. The value is in the nature of the problem. It causes one to think analytically and use non-intuitive reasoning to formulate a conclusion. It still baffles me that many here do not see that any market dynamic is a distribution with certain unique properties ... a coin flipped over time has a certain distribution as does volatility over time. Which one of these distributions has easily identifiable/exploitable characteristics and how does one determine those characteristics are genuine? The answer certainly isn't found by "looking at the chart".
If your trading strategy/system is pretty mechanical and can be evaluated in the context of winners and losers, then statistical probabilities for something as easy as coin flipping provide insight. It's already been said, but if you determine your trading has a 50% win rate, then you know that 5,6,7+ losers in a row is in your future. To continue trading through a string of losses is exactly what "disciplilne" means. IMO, insight from coin-flipping distributions (as the previous posters said) is very helpful. I built my own statistical model and played around with trading "parameters". It was based on the coin-flipping example. It helped me. That's the only proof I need. J.Scott
I find this an outstanding post! Masterjaz: I have some comments/suggestions that I can make in PMs. If you would be interested let me know. Regards!
This is another outstanding post! Thank you Mike! You are right that the problems may not be easy to grasp in their own right. The second road block is the applicability of the problems to trading. One still has to formulate a trading problem in terms of a Monty Hall/Secretary problem. I believe there can be more than one application. So there is potential for new developments. Mike: Could one ask you to share some of what you know in terms of applications of those two problems to trading ? I would not be offended if you do not share, as I believe it is your right to keep things private. You already greatly shared by making comments on the existence of the potential. If discussing of this is better suited to PMs, I would be delighted to read some of your insights. Best regards!
Coin flipping is random, unfortunately. Cards counting ain't on another hand. Certain trading strategies can be based on coin flipping or cards. The problem of coin flipping is that it carries minimum probability of next event taking place, even though it may seem more likely to get heads after 5th consecutive tails toss. You can toss tails another 5 times in a row or 10. In card counting it is way different as you have 36 cards & if you counted 4 kings there is no chance of another one being in that deck. Therefore, past events in conjunction with limited amount of cards provides scope for much better probability when compared to coin flipping which offers 0 edge. When I "flip trades" going for one breakout to the opposite one I try to gauge using volatility analysis, range analysis & momentum analysis. If I was to implement coin flipping going for random entries, then I would wipe out account eventually.