random entry trading

[walterjennings]

Quote from Steveyd:

I did not imply that. Expectancy will be close to 0 no matter what the PT and SL are (as asap has been saying).

I agree this is the key issue needing discussion here. I am not 100% convinced that this is the case. Why are we limited to exp close to 0? We just need the exp to be greater than cost in any case.

Quote from Steveyd:


You can't say this without qualifying your exit method. You won't have a 50% chance of profit on the next trade if your PT is double your SL, for example.

incorrect. give me an exit strategy out of the set of all possibly exit strats, say S, we do not need to qualify it. given S, and the future F.

let pl equal the profit or loss given exit strat S running over the set of data F until close taking a BUY position at the start.

obviously S(F) = pl when buying. S(F) = -pl when selling.

now flipping a coin. we have 50% chance of buying and selling. ergo we have a 50% chance of +pl or -pl regardless of exit strategy when entering on a coin flip.

note this is true with the case of TP being double SL. since given TP = x and SL = 2*x. we know S(F) = pl , flipping a coin we have 50% chance of pl and 50% of -pl. given TP =x and SL = 2*x.

Quote from Steveyd:


Let me rephrase:

Thinking that you can find a random entry strategy with negative expectancy (of any significance) is the flaw in your logic.

there you go with the negative expectancy again. ignore the term negative and lets discuss it as 'non-zero' exp. thats all we care about. because we know if its positive, then we are done. if it is negative. we can flip it in the no cost world.

you are saying we are unable to get a non zero expectancy in a random entry strategy. why? this is the whole reason for the discussion on this thread, you simply stating that 'it is a flaw in logic' does not advance the conversation at all. why do you think it is impossible to get a non zero expectancy (of any significance).

lets define 'any significance' as |exp| > cost, because if we could find that, then the system would be profitable over time.

 

[illiquid]

Quote from walterjennings:

Iincorrect. give me an exit strategy out of the set of all possibly exit strats, say S, we do not need to qualify it. given S, and the future F.

let pl equal the profit or loss given exit strat S running over the set of data F until close taking a BUY position at the start.

obviously S(F) = pl when buying. S(F) = -pl when selling.

now flipping a coin. we have 50% chance of buying and selling. ergo we have a 50% chance of +pl or -pl regardless of exit strategy when entering on a coin flip.

Imagine you have a stop loss at 5 ticks and target of 100. If the market just ranges up and down 20 ticks, it doesn't matter if you flip heads or tails, you will have -pl either way.

 

[walterjennings]

Quote from illiquid:

Imagine you have a stop loss at 5 ticks and target of 100. If the market just ranges up and down 20 ticks, it doesn't matter if you flip heads or tails, you will have -pl either way.

note. this is very important when trying to understand any of this. everything must be the exact same when flipping between buy and sell. the entry and exit points in the BUY case must be the exact same in the sell case.

an easy way of doing this is by always running the buy case and said exit strategy on a paper trading account. then placing the 50/50 BUY or SELL on a live account, using the entry and exit signal from the paper trading account. therefore pl(S,F | buy)= -pl(S,F | sell). easy as pie.

sorry i did'nt make this clear before. it was an implied fact in my mind. no wonder people wernt seeing it the way i was.

 

[illiquid]

What is the point of that, having one account long X and another account short X? You are just negating expectancies, given 0 slippage and spread. Someone else can give a better argument than I can at the moment, I just woke up from a nap and my brain is foggy.

 

[walterjennings]

Quote from illiquid:

What is the point of that, having one account long X and another account short X? You are just negating expectancies, given 0 slippage and spread.

eh? you dont have to actually place the orders on the paper trading account. i was using it as an example for how to get independent entry / exit signals regardless of being in a long or short position. common illiquid. its not complex. ill put it to you another way

say we have a system running. it only does BUYs, it also simulates TP and SL. all the broker gets from the system is an open signal and a close signal. the system handles entry and exits. now lets intercept those original open, and flip a coin to leave it as an open buy, or change it to and open sell. the close will always close whatever position is open, so we dont have to mess with that.

so given the system, for any action. the pl can either be positive or negative. we dont care. we also dont care what ratio of positive or negative actions we have.

all we care about is that for that said pl. we have a 50/50 chance of it being either pl. or -pl. since we change the buy to a sell 50% of the time. the exits still execute as if they were all buys.


and correct! we are negating expectancies. thats the whole point! now that we know we agree we can negate expectancies giving no spread / slippage. we can get back to the original discussion about wether or not we can get a |exp| > cost(spread, commis) given random entry and some kind of exit strat. knowing we can always flip the exp to the positive side if need be.

 

[illiquid]

Quote from walterjennings:

note. this is very important when trying to understand any of this. everything must be the exact same when flipping between buy and sell. the entry and exit points in the BUY case must be the exact same in the sell case.

an easy way of doing this is by always running the buy case and said exit strategy on a paper trading account. then placing the 50/50 BUY or SELL on a live account, using the entry and exit signal from the paper trading account. therefore pl(S,F | buy)= -pl(S,F | sell). easy as pie.

sorry i did'nt make this clear before. it was an implied fact in my mind. no wonder people wernt seeing it the way i was.

I still don't understand. I have a method where my exit strategy is 5pt SL and a 100pt TP from my entry, regardless of direction -- how does the coin flip work here? I've flipped heads, what now?

 

[illiquid]

Quote from walterjennings:

so given the system, for any action. the pl can either be positive or negative. we dont care. we also dont care what ratio of positive or negative actions we have.

This is not the system, "for any action"; this is (-system), after the fact.

 

[walterjennings]

Quote from illiquid:

I still don't understand. I have a method where my exit strategy is 5pt SL and a 100pt TP from my entry, regardless of direction -- how does the coin flip work here? I've flipped heads, what now?

gah. ok. lets run it in the buy case. say it buys at 1.5$. so the SL is at 1.495 and the TP is at 1.6. so it runs and either stops at 1.495 or 1.6. yay. so in the case of 1.6 we made .1. in the case of we lost 0.005. double yay.

now watch this. this is where the magic happens. flip the coin. boom. we got heads. now instead of a buy. we did a sell. since we are running as if we are in a buy case. we still stop out at 1.6 and 1.495.

at 1.6. we lost 0.1. at 1.495. we gained 0.005!!!!!! notice regardless of which stop we hit. the pl gets negated when we flip the entry? hallelujah

edit: "This is not the system, "for any action"; this is (-system), after the fact." dont quite understand this.. the system stays the same. it is either action or (-action) (buy or sell) depending on the coin flip. note using TP and SL on the brokerage and changing the entry from a buy to a sell fundamentally changes the entry / exits. maybe thats where people are getting confused.

we are straying from the real nut meat of the idea though. which is wether or not we can get |exp| > cost(spread, commis) given a chance to run in a no cost world. basing entries on a random coin flip. all this fluff we are discussing is just a proof that given a negative exp, we can flip it to a positive in the no cost world. making it easier to find a |exp| > cost.

 

[illiquid]

Quote from walterjennings:

gah. ok. lets run it in the buy case. say it buys at 1.5$. so the SL is at 1.495 and the TP is at 1.6. so it runs and either stops at 1.495 or 1.6. yay. so in the case of 1.6 we made .1. in the case of we lost 0.005. double yay.

now watch this. this is where the magic happens. flip the coin. boom. we got heads. now instead of a buy. we did a sell. since we are running as if we are in a buy case. we still stop out at 1.6 and 1.495.

at 1.6. we lost 0.1. at 1.495. we gained 0.005!!!!!! notice regardless of which stop we hit. the pl gets negated when we flip the entry? hallelujah

What is the point of this? To make your example even easier, the strategy is just GO LONG X. So your opposite account is just SELL X. You have 2 accounts, one long and one short, and you flip coins between which account to look at, at any given time in future F.

All you have is 2 strategies that cancel each other out, ie, a perfect hedge, 0 total expectancy. What would be the point of flipping coins again?

 

[walterjennings]

Quote from illiquid:

What is the point of this? To make your example even easier, the strategy is just GO LONG X. So your opposite account is just SELL X. You have 2 accounts, one long and one short, and you flip coins between which account to look at, at any given time in future F.

All you have is 2 strategies that cancel each other out, ie, a perfect hedge, 0 total expectancy. What would be the point of flipping coins again?

you're kidding right? where are you getting this perfect hedging / multi account stuff from? i used the paper trading account as an example of how to get the same entry and exit signals, using a single TP and SL regardless of actually trading long or short.

the idea is this. single account. single trade (open and close). known stops x, y. two possible actions, buy or sell. note x and y remain constant regardless of buy or sell. pl for buy will be -pl for sell given constant x,y.

im actually confused now. i dont know how coin flipping got worked in. that was an argument that it is possible to get a 50% profitability given a random entry / any exit strat. lets get back to the idea of finding an |exp| > cost. is it possible with random entry? if not. why?

edit: ahhh i see. i stupidly used the term 'flipping' expectancy (negating action). which has nothing to do with flipping coins

edit 2:

Quote from Steveyd:


You can't say this without qualifying your exit method. You won't have a 50% chance of profit on the next trade if your PT is double your SL, for example.

i just really confused myself thinking i was wrong by calling this incorrect. say your exit strat gives a 60%-40% win-loss ratio. flipping a coin to negate action on each signal. say you trade 1,000,000 times. roughly you will have close to 500,000 no negated trades. giving you 0.6*500,000 wins and 0.4*500,000 losses. you will also have ~500,000 negated action trades. giving you 0.4*500,000 wins and 0.6*500,000 losses. add those together. 500,000 wins. 500,000 losses. or 50% chance of profitability regardless of the exit method using coin flipping to enter the position. phew. i thought i went insane for a second there. note this works for any exit strat. any win-loss ratio.