zupcon
Experienced member
- Messages
- 1,162
- Likes
- 322
It has been revealed to you time after time
how many more times do you need to be told ?
It has been revealed to you time after time
how many more times do you need to be told ?I confess to being confused about how R-squared is used to calculate a success rate (maybe confused past the point of being able to ask the right questions). What "model" is used for correlation purposes to calculate R-squared for a series of trades?
Claudia123
Established member
- Messages
- 724
- Likes
- 57
come on everyonerich explain this to us
everyonerich
Experienced member
- Messages
- 1,109
- Likes
- 43
It has been revealed to you time after time
its not everyone gonna buy, theres always the seller.
dont forget, if everyone had extra cash, its not like they gonna burn it. they will indirectly spend it and thus revive the current economy crisis. its good for all because money is just circulating (for eg: money reached personal bank acc, the consumer will spend the money and the bills always end up in bank or any banks again) and from the big picture nobody is really losing anything anyway.
everyonerich
Experienced member
- Messages
- 1,109
- Likes
- 43
i only use manual trading at the moment.
everyonerich
Experienced member
- Messages
- 1,109
- Likes
- 43
who cares? money is just a belief, life is short, you cant take all the money when youre in coffin anyway
everyonerich
Experienced member
- Messages
- 1,109
- Likes
- 43
because capitalism is double-edge sword..
oiltanker
Established member
- Messages
- 955
- Likes
- 89
..What "model"..
again in short
in this case its a trading strategy used which is 'a model' of price action. There should be an expected outcome for that strategy [model].
the maths guys don't use TA which they regard as 'believing in santa'. They say price has 'no memory' and that statistics show the market is only slightly better than 'noise'. So the strategy [usually on momentum] they use is batch trading ie expected outcome over a number of trades.
Because the market is always changing - who does carry trade now? where is the vix now compared to Nov? etc -so strategy performance will change. Some at 0.6 might go down to 0.3 and vica versa. Through knowing what the expected outcome is one can monitor changes in the market. If the expected outcome starts to deviate from that and something that worked once is now like flogging a dead horse it gives you warning to go find another horse and thus preserve capital.
a simple way [non maths complex] to use the idea is just work out how many trades a trader should have taken in the last week or last month [whatever time length the trader bases the accounting on] with the strategy and the conditions for the strategy e.g the strat is 58x on the 15m chart trading only London session in the gbp/eur. It has to be every trade.Say it was for simplicity 20 trades in a week of which 10 wins 10 losses. Which gives 0.5. So based on last weeks data we have an expected outcome of 0.5. If at the end of next week and assuming the trader was discipline enough to take every trade and could cover the potential stop loss for the strategy, the outcome is 0.4 and the week after its 0.3 we can see a downtrend in the strategy which tells us the market is changing . If it get to 0.2 then the strategy could go on the demo account until we start to see an uptrend in the number. etc.
if a trading house had the resources they could monitor lots of strats on different time frames, different sessions and different markets and only use those with an expected outcome of 0.5 or more. Which would suit automation.Most markets are now 60% automated [although that covers a lot of different things] and that percentage is expected to continue to grow.
again in short
in this case its a trading strategy used which is 'a model' of price action. There should be an expected outcome for that strategy [model].
the maths guys don't use TA which they regard as 'believing in santa'. They say price has 'no memory' and that statistics show the market is only slightly better than 'noise'. So the strategy [usually on momentum] they use is batch trading ie expected outcome over a number of trades.
Because the market is always changing - who does carry trade now? where is the vix now compared to Nov? etc -so strategy performance will change. Some at 0.6 might go down to 0.3 and vica versa. Through knowing what the expected outcome is one can monitor changes in the market. If the expected outcome starts to deviate from that and something that worked once is now like flogging a dead horse it gives you warning to go find another horse and thus preserve capital.
a simple way [non maths complex] to use the idea is just work out how many trades a trader should have taken in the last week or last month [whatever time length the trader bases the accounting on] with the strategy and the conditions for the strategy e.g the strat is 58x on the 15m chart trading only London session in the gbp/eur. It has to be every trade.Say it was for simplicity 20 trades in a week of which 10 wins 10 losses. Which gives 0.5. So based on last weeks data we have an expected outcome of 0.5. If at the end of next week and assuming the trader was discipline enough to take every trade and could cover the potential stop loss for the strategy, the outcome is 0.4 and the week after its 0.3 we can see a downtrend in the strategy which tells us the market is changing . If it get to 0.2 then the strategy could go on the demo account until we start to see an uptrend in the number. etc.
if a trading house had the resources they could monitor lots of strats on different time frames, different sessions and different markets and only use those with an expected outcome of 0.5 or more. Which would suit automation.Most markets are now 60% automated [although that covers a lot of different things] and that percentage is expected to continue to grow.
zupcon
Experienced member
- Messages
- 1,162
- Likes
- 322
R measures the degree of correlation. If you have a strategy that historically always returns 50% winners over a given time period, and you plot these results, week in week out and the strategy always returns exactly 50% then you'll get an R value equal to 1. The results are a perfect fit against the model
If however the strategy suddently starts returning 100% winners, you'll get an R value of much less than 1, similarly if the strategy starts returning 0% winners, the R value will drop to be less than 1. Therefore R is completely irrelevant, R does not tell you if the strategy is winning more, or losing more, it just tells you something is changing, and thats why its not used as a performance measure for systems
As for markets being random noise, well, thats why quants are quants and traders are traders
oiltanker
Established member
- Messages
- 955
- Likes
- 89
...the strategy suddently starts returning 100% winners, you'll get an R value of much less than 1...
not in isolation. it's a new number each week. there are lots of ways to use it. it is used as a measure of traders.
as long as people are winning that is the important point.
for anyone interested there are courses at oxford uni on it including how they take on the chartists.
not in isolation. it's a new number each week. there are lots of ways to use it. it is used as a measure of traders.
as long as people are winning that is the important point.
for anyone interested there are courses at oxford uni on it including how they take on the chartists.
Last edited:
R measures the degree of correlation. If you have a strategy that historically always returns 50% winners over a given time period, and you plot these results, week in week out and the strategy always returns exactly 50% then you'll get an R value equal to 1. The results are a perfect fit against the model
If however the strategy suddently starts returning 100% winners, you'll get an R value of much less than 1, similarly if the strategy starts returning 0% winners, the R value will drop to be less than 1. Therefore R is completely irrelevant, R does not tell you if the strategy is winning more, or losing more, it just tells you something is changing, and thats why its not used as a performance measure for systems
As for markets being random noise, well, thats why quants are quants and traders are traders
My thoughts as well.
zupcon
Experienced member
- Messages
- 1,162
- Likes
- 322
R is a very common and very simple statistical concept, probably still taught to 12 year olds in schools, its far from an advanced concept. Its a measure of correlation, and thats all it is.
R will NOT give an indication of winning or losing, its really is rather basic stuff, and to argue otherwise is deliberately misleading and serves no purpose.
zupcon
Experienced member
- Messages
- 1,162
- Likes
- 322
R is NOT a measure of statistical significance, its a measure of correlation.
So far youve asserted that R isa measure of system performance, when it clearly isnt, and now you are suggesting its a measure of statistical significance, when again, it clearly isnt. Its hardly rocket science.
new_trader
Legendary member
- Messages
- 6,770
- Likes
- 1,655
everyonerich
Experienced member
- Messages
- 1,109
- Likes
- 43
is that mean book = wrong information? do u mind explain more a bit..? thanks
There are entire funds that use nothing but automated systems. They usually employ mathematicians to develop their algorithms. So yes, there are some out there. The real question is whether or not the system you're putting YOUR money into works or not.
