Ward 6 - page 3

 
DmitriyN:
Let's look at a specific situation. 2008, trending down by who knows how many points.
Then a 50% pullback to the resistance/support level. It took 2 months or something like that.

There is no such thing as a "trend down for who knows how long". There was a 100% upward pullback (correction) with a scale (scope, price move) of over 50 pips, which is what I use for TP=SL. And since that's the case, I don't care about the direction of the larger move.
 
Dr.Drain:

Ah this. That's easy. It doesn't matter where price is going, it's how it's going. I can generate a chart that will have a stupid unbreakable upward trend, but all 100% of trades with TP=SL=50 pips will be closed by SL. Especially since in practice you cannot specify a "trend" without specifying the scale on the price axis in question. Until you have specified the scale, i.e. the value of the expected TP at the opening of a trade with equal TP and SL, you cannot make a decision. For at the same moment, both the decision to sell and the decision to buy can be equally true. And both will close at the TP. At different times (first a trade with a lower TP, then a higher one, of course). If you could separate the trend and the flat by setting the scale in advance (otherwise you have no right to say these words at all) - it would be another representation (formulation) of the Grail. But you can't. So let's not talk about "long trends".

Lots of letters. Unfcooled ))))
 
Dr.Drain:
Roman100, if your algorithm really works, it can be reformulated into a more obvious one in the form of a non-lagged non-drawing filter . If such a transition cannot be made - your algorithm has no predictive power to shift the probability above 50%.

Approaches can be different). Judging by this, my filter even "determines";;) as it exclusively uses pending orders.
 
Dr.Drain:

It doesn't look very presentable solely because of the small number of deals.

Well, you can't draw any more or less adequate conclusions from so many deals yet...
 
No, no. To understand that your filter is not ahead of anything, consider it on simple signals instead of prices: meander, sine, triangular sawtooth, and most importantly: step. That is where you will see all the lag. For miracles do not happen, and you cannot predict in advance where the jump will occur and that there will not be a pullback.
 
вот блок открытия ордеров
==========
/ ордера есть - значит выйдем
if (TotalOpenOrders > 0) { return(0); }
// ордеров нету - выставим случайные
   Cen1 = MathRand();
   Chi = DoubleToStr(Cen1,0);
   //   
   Lots = 0.1;
   // через жопу определяем чётное ли число - вообщето решается одной строкой типа if (Cen1&1) {нечет}
   // но пишу так чтобы Енот смог разобраться :)
   if(StringLen(Chi)>1) {
      Chi = StringSubstr(Chi,StringLen(Chi)-1,1);
   }
   if (Chi == "0" || Chi == "2" ||Chi == "4" ||Chi == "6" ||Chi == "8") { // чётное число - Байки ставим
      OrderSend(Symbol(), OP_BUY, Lots, Ask, 3, Ask - BaSL*Point, Ask + BaTP*Point,"My order #"+total,16384,0,Green);
      return(0);
   }
   // число Нечётное
   OrderSend(Symbol(), OP_SELL, Lots, Bid, 3, Bid + SeSL*Point, Bid - SeTP*Point,"My order #"+total,16384,0,Red);
// выходим   
return(0);

what is it in 2008? here's an example Random TA entry = SL the result is..:


\\\\\\\\\\\\\\\\\

SymbolEURUSD (Euro vs US Dollar)
Period5 Minutes (M5) 2008.01.02 10:00 - 2008.12.31 17:55 (2008.01.01 - 2009.01.01)
ModelAll ticks (most accurate method based on all smallest available timeframes)
ParametersLots=0.1; BaTP=500; BaSL=500; SeTP=500; SeSL=500;

Bars in history74639Modelled ticks4156824Simulation quality90.00%
Chart mismatch errors0




Initial deposit10000000.00



Net profit2182.38Total profit40017.71Total loss-37835.33
Profitability1.06Expected payoff1.40

Absolute drawdown1984.76Maximum drawdown2760.01 (0.03%)Relative drawdown0.03% (2760.01)

Total trades1558Short positions (% win)772 (51.81%)Long positions (% win)786 (51.02%)

Profitable trades (% of all)801 (51.41%)Profitable trades (% of all)757 (48.59%)
Largestprofitable trade50.00losing deal-50.84
Averageprofitable deal49.96losing trade-49.98
Maximum numbercontinuous wins (profit)8 (399.79)continuous losses (loss)9 (-450.63)
Maximumcontinuous profits (number of wins)399.79 (8)Continuous loss (number of losses)-450.63 (9)
Averagecontinuous winnings2continuous loss2



 
Dr.Drain:

There is no such thing as a "trend down for who knows how long". There was 100% upward twitching (correction) with a scale (scope, price move) of over 50 pips, which is what I use for TP=SL. And since that's the case, I don't care about the direction of the larger move.

Roughly speaking 2100 points down and 970 up (I didn't count exactly). The procession stretches over 10 weeks.

 
Roman100:
Well, on such quantity of transactions it is impossible to make more or less adequate conclusions yet...


You can. There are two ways to draw conclusions. 1. Hypothesis - experiment - confirmation or refutation of the hypothesis and 2. many-many-experiments - statistical processing - hypothesis-sucking.

If we go by path 1, hypothesis (theory) -> experiment, that for its (hypothesis) confirmation it does not need many experiments to reveal statistics and suck the hypothesis out - it is enough even that small amount of data and fixed upward slope to reasonably assume that the theory works.

 
DmitriyN:

Roughly speaking 2100 points down and 970 up (I didn't count exactly). The procession stretches over 10 weeks.


So there would have been a sustained Kick in that pair. Let's assume so. But there is not one or two pairs in trade, but a dozen or more. On average, there is still an effective averaging at this level as well. I'm off to bed. I'll be back tomorrow night, Moscow time.
 
Dr.Drain: On average, there is an effective averaging still going on at this level too.

Who decided that?