Criterion for automatic selection of optimisation results. - page 8
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
Извините, господа, но здесь вроде бы другая тема обсуждалась(хоть и в прошедшем времени, тем не менее).
Agreed, but no one but me has posted their version of the auto-selection criteria for the results... )))
Это могут быть самые разные варианты сигналов, например:
первый - ослабевание основного, появление плохих признаков, отклонение с разрывами, и т.п. ...
второй - пересечение важного уровня, разворот основного тренда, пробой канала, или какой-нить сильный сигнал, отменяющий или отсрочивающий возможные возвраты...
третий - восстановление основного сигнала, отскок от трендовой или горизонтальной линии
i.e. these three signals, especially the first one (weakening of the main one, appearance of bad signs, deviation with gaps, etc.) are more related to the essence of the main signals, to the trading strategy itself.
I guess I know what I'm talking about. For my own ideas, I approach it differently - I know from the start that a major trading signal may not be followed by the expected development, what I take it you call "weakening of the major, appearance of bad signs, deviation with gaps, etc.". And since the expected development didn't follow, it means that the mechanism I was counting on is no longer working in the market. The main trading signal failed and the money in the market went the other way. I suppose that like everyone else in this situation, I am well aware that, yes, it is very possible that the market will come back, that it is possible to win back, etc. But I restrain myself from doing such a thing, as a wagering is not based on the mechanism of the underlying trading signal, but only on chance. I can say this because I have also pre-calculated the stop signal - the level from which further market behaviour has no stat advantage. You see, I open where I see the advantage and close at the level where there is no advantage. And wagering, based on pain of own drawdown, from the level, where the statistic advantage is gone, in my opinion is a very pernicious way, because it has nothing under it but desire to win back. So if the main trading signal is dead, then it is dead. I am experiencing a loss. Perhaps if you calculate your stop level in a statistically significant way, then you don't have to stitch a strategy of wagering on drawdown to your main trading strategy. I've noticed that this "wagering differently if it didn't work out the first way" approach is widespread. In my opinion this is a clear sign that the basic strategy has not been finalized and instead of correcting the basic strategy "fixing max drawdown with extra profit" is widespread, i.e. treating not the cause but the consequence - which is the same problem, only why not solve it from the start? This is what I do - from the beginning the main trading signal is healthy and there is no need to correct it. This is my ideology. This is why I'm against complex superstructures, fixes, enhancers and the resulting indicators - they are most likely a substitute for admitting to oneself that the main strategy has failed. The initial minimisation of drawdown to a level where no extras are needed, where only the main trading strategy is wagering money, gives me the feeling that everything is built correctly.
Согласен, но ни кто кроме меня не выложил свой вариант критерия автоотбора результатов... )))
There you go, begging for my super simple but super effective criterion ;)
And to be fair, StatBars, we are just the two gentlemen who posted not references to reasoning that has reached an inconclusive dead end, but criteria. And now we are discussing their ideology. I doubt we are preventing the rest of us from bringing anything of value to the table.
Vita писал(а) >>
I can say this because I have pre-calculated the stop signal - the level from which further market behaviour has no statistical advantage. You see, I open where I see a statistical advantage and close where there is none. And wagering, based on pain of own drawdown, from the level, where the statistic advantage is gone, in my opinion is a very pernicious way, because it has nothing under it but desire to win back. So if the main trading signal is dead, then it is dead. I am experiencing a loss. Perhaps if you calculate your stop level in a statistically meaningful way, you don't have to stitch a strategy of wagering on drawdown to your main trading strategy. I've noticed that this "wagering differently if it didn't work out first" approach is widespread.
There are strategies where losing trades are 1-5% and drawdown is 60-90%. This is not the drawdown of closed positions, but of profitable ones inside open trades. It is this drawdown that makes sense to cut.
If one believes that the profit margin decreased not so much as to close the position permanently, in this sense it is worth to close this position, wait for its restoration and then open it again. It does not matter if there was a slippage or not, if it raises doubts it's better to wait it out and continue as planned. And if you close the signal, it may still be alive, but not for us any more. Another thing if the signal is dead for a long time, then it makes sense to wait for it to work again.
In general, I believe the safest strategy is the one based on a group of signals and their classification, because trends are changing constantly, but they repeat all the time too. If we divide them not only into trend and flat of the current TF, but also into higher TFs, the flat on the higher one is a trend on the current one, both are flat there and there, and in one after the growth, or vice versa on the other... It is even better to consider three TFs. For example I am working with 15 and taking into account trend or flat on H4 and D1. This is a great help for my TS. This kind of orientation is like a breath of fresh air.
About the sorting criterion...
Somebody write their own formulas. I'd like to try it out for comparison...
Есть такие стратегии, в которых убыточных сделок 1-5%, а просадка 60-90%. Это просадка не закрытых позиций, а внутри открытых сделок прибыльных. Именно эту просадку и есть смыл урезать.
I didn't even think of touching profitable trades with my hands. ;) I mean, I'm not so sensitive that I care about the drawdowns within a profitable trade. Unless, of course, I fully understand what I'm talking about.
We opened a position and lost, losing as much as 5% of the balance, but eventually everything came back and the position was closed with a profit. The drawdown turned out to be 95%, i.e. we were on the verge of losing the whole deposit... Profitable trades should not be like walking on the edge of an abyss, that's what it's all about...
Открывается позиция, уходит в минус, да такой, что баланса остается 5% от того что было, в конечном итоге все возвращается и эта позиция закрывается в плюсе. Просадка получилась 95%, т.е. мы были на волосок от того чтоб потерять весь депозит... Прибыльные сделки не должны быть как хождение по краю пропасти, вот о чем речь...
I get it. I have reinforced for myself the expression "Profitable deals should not be like walking on the edge of a precipice". Any trades should not be like walking on the edge of a precipice. A single trade that can result in a 95% drawdown is unthinkable to me. It's wrong from the start. As before I am inclined and will repeat - to rescue a trade that leads to a margin call is to treat the consequence, not the cause. You can do that, but to me it is a futile way to go, because the logic here is simple - why should a strategy based on "you have to save the deal" lead to success rather than to an even bigger collapse? Why not enlist such a rescue strategy from the outset and not take things to the point of "hanging by a thread"? Because, and I firmly believe in this, there is no basic strategy or rescue strategy in such cases. Putting illusions aside, all we have is exorbitant risk per trade, the pain of a drawdown in that trade, and the hope of a "rescue strategy". Sometimes this works, the intermediate ending is delightful and the illusion is backed up. But not for me. And you are highly discouraged from betting your entire deposit on a single trade. If it hurts, it means you're doing something wrong. Translated as, if you have to save your deposit (trade, basic strategy), then you are wrong from the start.
I'm shocked - 95% drawdown from a single trade. Amen.
95% is me for example, of course there is no point in even fighting it.
But back to the topic at hand. I have highlighted for myself the most important indicators:
PipBar - pips/bars (sum of points of all trades divided by the sum of bars of these positions, i.e. quality of time used)
PF - profitability (Profit Factor)
SdDay - number of deals per day (sum of deals divided by number of days tested)
ProcDay - the percentage of profit per day (calculated using a special formula, those who think that the total percentage divided by the number of days is the same - error!!!)
MD - Maximum drawdown
SrD - Average drawdown (total drawdown of all orders divided by number of orders)
Ust - Sustainability (profit in $ divided by the maximal drawdown in $)
PrMD - Percentage of used deposit funds divided by maximal drawdown. (With a growing lot, if we play with 5% of the deposit and get a drawdown of 20%, the game at 10% of the deposit corresponds to 40%, the indicator serves to eliminate errors in comparing the criteria of different strategies, if the used funds percentage diverges)
Other values only disturb me. I'll send you my revised version of the formula later. I'll be glad to compare it with someone else's.
And here is the formula, or rather the code right away:
if(MathAbs(Loss)==0 && Profit>0) PF=10;
else PF=Profit/MathAbs(Loss);
if(PF>=3) Vigoda=2*SdDay+((PipBar+Ust)/10)+((ProcDay*10)/(MD+(SrD*4)));
else Vigoda=(PF-1)*SdDay+((PipBar+Ust)/10)+((ProcDay*10)/(MD+(SrD*4)))
if((SdDay*5)<1) Vigoda=0; // filter from random deals (limit at least 1 deal per week)
I am waiting for your comments and critics...
А вот и формула, вернее сразу код:
if(MathAbs(Loss)==0 && Profit>0) PF=10;
else PF=Profit/MathAbs(Loss);
if(PF>=3) Vigoda=2*SdDay+((PipBar+Ust)/10)+((ProcDay*10)/(MD+(SrD*4)));
else Vigoda=(PF-1)*SdDay+((PipBar+Ust)/10)+((ProcDay*10)/(MD+(SrD*4)));
if((SdDay*5)<1) Vigoda=0; // фильтр от случайных сделок (ограничение сделок не менее чем 1 в неделю)
Жду Ваши коменты и критику...
It is difficult to comment on it without an example or code, which could be felt on your example, looking at the formula. Here's what it looks like to me:
The coefficient (PF-1)*SdDay, up to the point where PF>=3 is bounded from above, is: (PF-1)*SdDay = Average payout per day / average loss per trade. You can see the attempt to ration the profit and loss amount by days and trades themselves rather strangely. I doubt we are adding anything of value here. What is the point of limiting this Average Payout per day / Average Loss per trade to twice the number of trades per day at PF>=3? Firstly, you are clearly favouring something else in the rest of the formula. Secondly, you have the wrong design from the start - throw out the results with unsatisfactory simple criterion first (number of trades per period, profit factor, profit, etc.) and then there will be no need for such strange and, I believe, inexplicable restrictions in your formula.
Ratio ((PipBar+Ust)/10) = Profit / (1 / Point value / Number of bars + 1 / Amount of maximal drawdown) /10 - we normalize the profit again, but using the point value and number of bars. What is the meaning of such normalization?
And the (ProcDay*10)/(MD+(SrD*4)) part - is, as I understand it, again an empirical attempt to combine return for the period with risk. Why do you like this ratio and the previous parts better than the Profit / Drawdown? Curious, can you give examples - for these results Vigoda shows so and so, which reflects this and this well?
The criterion formula cannot be so difficult (in writing, not in understanding). We should try differently to put the evaluation of profitability, risk, payoff for a period and threshold on number of deals for a period into one criterion formula. In my opinion, by design, the criterion should look as follows: Profitability Score * Risk Score * Return Rate Score * Threshold Score or something else. The scores should be multiplied, not added up. It's probably the way the world works that you don't add up days with pieces, height with courage, jump height with running speed, because such metrics don't make any sense. And neither should you.