Stable MTS - page 23
You are missing trading opportunities:
- Free trading apps
- Over 8,000 signals for copying
- Economic news for exploring financial markets
Registration
Log in
You agree to website policy and terms of use
If you do not have an account, please register
It's very simple really, remember, as in the old anecdote: "... I'll take the goods at the wholesale market for a ruble apiece, and spread the stalls for a triple, 3-1 = 2, then these 2% and live ..." . But seriously, all trading comes down to trivial buying and selling operations. But it's the principle of determining the right moment for these operations that determines the logic of operation and profitability/loss of the trading algorithm. And the description of this principle in my mind goes beyond the concept of "in general". Therefore, the question arises: "Do you want to get by or do you want to get by?
I will not lecture you on how and in what way to attract and interest investors, but I will venture to make the following analogy: imagine a situation where an employee of a prestigious car dealership wants to sell a customer a flashy Mercedes or something similar. Will he be interested in consumer attributes such as comfortable leather interior, smooth running and instantaneous acceleration or will he tell her about the technology of dressing of bison skins and technology of engine assembly and production of its components?
When you understand it, they will be able to make you such an algorithm for 10$ in freelance service... What's my interest then?
Well, I don't mind.
So it works for me too)))
You have my spreadsheet for testing. Drop testing according to spreads from the spreadsheet on several pairs from 2001 to 2016 and from 2011 to 2016. You can do so in a private message. If it holds up, we can talk.
Yes, and bar times from five minutes and up ...
You've seen the results of mine anyway, so it's all real .... Several tools are always more interesting
Three months at a time counts. No need to recalculate, the market statistics we need change extremely slowly
Close, but not quite. Corr() and cov() are completely different in form. Just look at plot(corr()) and plot(cov()) and feel the difference.
You have my spreadsheet for testing. Drop testing according to spreads from the spreadsheet on several pairs from 2001 to 2016 and from 2011 to 2016. You can do so in a private message. If it holds up, we can talk.
Yes, and bar times from five minutes and up ...
You've seen the results of mine anyway, so it's all real .... Several instruments are always more interesting
It's very simple really, remember, as in the old anecdote: "... I'll take the goods at the wholesale market for a ruble apiece, and spread the stalls for a triple, 3-1 = 2, here's how these 2% and live ..." . But seriously, all trading comes down to trivial buying and selling operations. But it's the principle of determining the right moment for these operations that determines the logic of operation and profitability/loss of the trading algorithm. And the description of this principle in my mind goes beyond the concept of "in general". Therefore, the question arises: "Do you want to drive or drive in?
I will not lecture you on how and in what way to attract and interest investors, but I will venture to make the following analogy: imagine a situation where an employee of a prestigious car dealership wants to sell a customer a flashy Mercedes or something similar. Will he be interested in consumer attributes such as comfortable leather interior, smooth running and instantaneous acceleration or will he tell her about the technology of dressing of bison skins and technology of engine assembly and production of its components?
When you understand it, they will be able to make you such an algorithm for 10$ in freelance service... What's my interest then?
Well, I don't mind.
Sergei, do you mean the spreadsheet from FxPro? Was it in those documents, the link to which you posted here? I didn't see it there for some reason.
I did not sleep last night under the impression of this thread and it suggested to me that I should improve it, it has improved but 10%)
As the glory of the heroes was not resting and 70% of successful trades (and even 80% in one test) are striking, but the profit/loss ratio of 1/1, is a bit tense. I also was up all night thinking about improving my TS.
My last TS on FORTS gave up to 60% successful trades, but the average profit/loss ratio in a trade was ~(70-80)/(25-30)- numbers, those are points.
Couldn't think of anything until I remembered the uncertainty principle. Let me remind you, dx*dp>=h/2, i.e. the more precisely we determine the position of a particle, the less we know about its momentum (velocity). Of course the analogy is indirect.
We can improve credibility of entry, but it requires time for more plausible measurements, and hence although the number of unsuccessful trades will decrease, we will enter the trade later, and hence the profit will decrease. And in all successful trades. Thus increasing the ratio of successful and unsuccessful trades, we decrease the profit/loss ratio in trades.
If you look at transactions, a part of potentially successful ones is closed using a stop loss due to quotes fluctuations. If we move the stop, the ratio of successful/unsuccessful trades will increase, but the average loss per trade will increase. Yes, the ratio of profitable/lossmaking trades will increase but it will also decrease the ratio of average profit to average loss in a trade.
And we will obtain the same 70-80% of successful trades and 50/50 profit/loss ratio.
Now the question naturally arises - which system is better (more efficient)? The first thing that comes to mind for efficiency index is something like E = (probability of profitable trade)*(M trade profit/M trade loss). I haven't checked the maths or the physics yet.
As the glory of the heroes was not resting and 70% of successful trades (and even 80% in one test) are striking, but the profit/loss ratio of 1/1, is a bit tense. I also was up all night thinking about improvement of my TS.
My last TS on FORTS gave up to 60% successful trades, but the average profit/loss ratio in a trade was ~(70-80)/(25-30)- numbers, those are points.
Couldn't think of anything until I remembered the uncertainty principle. Let me remind you, dx*dp>=h/2, i.e. the more precisely we determine the position of a particle, the less we know about its momentum (velocity). Of course the analogy is indirect.
We can improve credibility of entry, but it requires time for more plausible measurements, and hence although the number of unsuccessful trades will decrease, we will enter the trade later, and hence the profit will decrease. And in all successful trades. Thus increasing the ratio of successful and unsuccessful trades, we decrease the profit/loss ratio in trades.
If you look at transactions, a part of potentially successful ones is closed using a stop loss due to quotes fluctuations. If we move the stop, the ratio of successful/unsuccessful trades will increase, but the average loss per trade will increase. Yes, the ratio of profitable/lossmaking trades will increase but it will also decrease the ratio of average profit to average loss in a trade.
And we will obtain the same 70-80% of successful trades and 50/50 profit/loss ratio.
Now the question naturally arises - which system is better (more efficient)? The first thing that comes to mind for efficiency index is something like E = (probability of profitable trade)*(M trade profit/M trade loss). I haven't checked the maths and physics yet.
Yuri, you don't have to invent anything. Everything has already been invented before us. The generalizing indicator is the mathematical expectation. The turtle traders used it, and they probably took it from someone else too.
The profit from trading can be represented by this simple equation:
Profit = N x Av(P) - (1-N) x SL, where
N - the share of profitable trades,
Av(P) - the average size of a profitable trade,
(1-N) - the share of losing trades,
SL - the average size of a losing trade (ideally it should be close to the stop loss).
Profit - the expected profit from the trade, i.e. a mathematical expectation.
Increase SL, and N will crawl upwards (in an ideal world with infinite capital and infinite SL (that is stop is absent at all), N = 1.
However, at the same time the second term of the difference also increases. So the maturity expectation perfectly reflects the optimality of the system.
(someone here asked why everyone attaches such importance to the mathematical expectation).
Yes, the discussion has stalled.
Colleagues, I repeat the question, I am very interested to see the really working systems, demonstrating a yield of 10-15% per annum, with a maximum drawdown not exceeding 10% and a maximum recovery time from drawdown not exceeding 3 months.
I am interested in two points of view:
1) searching for a benchmark (not a grail, but a benchmark) - what a good algorithm is capable of in principle, under which parameters it can be considered excellent and under which it is below average crap;
2) There is a real fund that is ready to invest in trading programs built on such agents, and the author gets a share in commissions.
Yuri, you don't have to invent anything. Everything has already been invented before us. The generalising indicator is the mathematical expectation. The turtle traders used it, and they must have taken it from someone else, too.
The profit from trading can be represented by this simple equation:
Profit = N x Av(P) - (1-N) x SL, where
Not Newton's binomial. (c) But in your expression you need specific trade sizes (mat. expectation of size). Imagine that they don't exist. You need to evaluate (compare) the effectiveness of several TS. And what do you care who trades with what lot. You want profit, and I want efficiency - different objectives. The system with a smaller profit, and other conditions being equal, may be more effective.
It is not difficult. But I haven't done it yet. )