#### augubhai

##### Well-Known Member
Control losses, Limit losses, Avoid losses... Eliminate losses

My average monthly return is 10%, in the first month I earned 120%, in the second -100%. Ending capital = ₹0.
My average monthly return is 100%, in the first month I earned 200%, in the second -100%. Ending capital = ₹0.
My average monthly return is ∞%, in the first month I earned ∞%, in the second -100%. Ending capital = ₹0.

When stated like this, average monthly return means nothing, big losses mean everything. Or, monthly average return is more meaningful than average monthly return. D Did i confuse u on this one?). Better still, state the compounded rate of return.

Take 0% average monthly return, with initial capital of ₹100.

Case 0: In the first month I got 0%, in the second 0%. Capital at the end of month 1 was ₹100, at the end of month 2 ₹100. Net change 0%.
Case 10: In the first month I got 10%, in the second -10%. Capital at the end of month 1 was ₹110, at the end of month 2 ₹99. Net change -1%.
Case 25: In the first month I got 25%, in the second -25%. Capital at the end of month 1 was ₹125, at the end of month 2 ₹93.75. Net change -6.25%.
Case 50: In the first month I got 50%, in the second -50%. Capital at the end of month 1 was ₹150, at the end of month 2 ₹75. Net change -25%.

Except in case where u lose 100% and have no more dough, the result is the same if the order of return is changed.
For example, reverse case 50: In the first month I got -50%, in the second 50%. Capital at the end of month 1 was ₹50, at the end of month 2 ₹75. Net change remains same at -25%.

I have taken 2 months... but as per my maths teacher, same logic applies even if there are more months than my fingers.

And my moral science teacher tells me the following morals from this story:
1. Eliminate big losses.
2. Bank deposits are a relatively risk free, always +ve return system. But get a rate that can trump inflation.

#### jamit_05

##### Well-Known Member
Control losses, Limit losses, Avoid losses... Eliminate losses

My average monthly return is 10%, in the first month I earned 120%, in the second -100%. Ending capital = ₹0.
My average monthly return is 100%, in the first month I earned 200%, in the second -100%. Ending capital = ₹0.
My average monthly return is ∞%, in the first month I earned ∞%, in the second -100%. Ending capital = ₹0.

When stated like this, average monthly return means nothing, big losses mean everything. Or, monthly average return is more meaningful than average monthly return. D Did i confuse u on this one?). Better still, state the compounded rate of return.

Take 0% average monthly return, with initial capital of ₹100.

Case 0: In the first month I got 0%, in the second 0%. Capital at the end of month 1 was ₹100, at the end of month 2 ₹100. Net change 0%.
Case 10: In the first month I got 10%, in the second -10%. Capital at the end of month 1 was ₹110, at the end of month 2 ₹99. Net change -1%.
Case 25: In the first month I got 25%, in the second -25%. Capital at the end of month 1 was ₹125, at the end of month 2 ₹93.75. Net change -6.25%.
Case 50: In the first month I got 50%, in the second -50%. Capital at the end of month 1 was ₹150, at the end of month 2 ₹75. Net change -25%.

Except in case where u lose 100% and have no more dough, the result is the same if the order of return is changed.
For example, reverse case 50: In the first month I got -50%, in the second 50%. Capital at the end of month 1 was ₹50, at the end of month 2 ₹75. Net change remains same at -25%.

I have taken 2 months... but as per my maths teacher, same logic applies even if there are more months than my fingers.

And my moral science teacher tells me the following morals from this story:
1. Eliminate big losses.
2. Bank deposits are a relatively risk free, always +ve return system. But get a rate that can trump inflation.
This is a thin line. In this pursuit, one gets fixated on losses and in the process the mind starts to make "eliminating losses" as the prime goal. Whereas the fact is that out of 10 trades 5 to 7 will be losers for any high frequency trading system. If you make the mistake of trying to avoid losers, then you will end up with a "curve fitted" trading system, which is a huge loss of time.

Needless to add, eliminating big losses which are more than the size of average profits is plain benefit.

#### augubhai

##### Well-Known Member
my PM to someone in response to his PM was forwarded by him to someone else who posted it in his thread.

I am copying that post here in toto to remind me of thing to check while looking at backtest results... will need to add a few more points to the checklist later.

Thinking about it again, that system result showed about 250% return every year without compounding. My current systems like the ORB should be able to that with a reasonable MM plan... 250% p.a just needs about 100 points on NF

some comments i got for ur system performance from senior member.

the system -having win rate >50% and Profit Factor 2 to 3, over a period of 4 years - is very difficult, but not impossible. (Anyway, I don't know of a system with similar results )

Whenever, I see a hugely +ve backtest, i look out for few things:

1. Bars repainting - systems looking into future. But Dell tells that it does not look into future.

2. Irresponsible Money Management, where u take maximum risk everytime. But Dell tells that he has a position size of 4 lots throughout. So, system is good on this one also.

3. Not considering costs+slippage, especially in lower timeframes. But Dell says he has take 0.6% for cost, which is good enough.

4. Curve fitting. Though the system tested for 3 years - no way to confirm whether would perform same way in other time periods or in the future.

5. Simple nonsense like considering Buy Price at Low of bars, or all Sell Price at High of Bar. But I think Dell is aware of this, and he mentions that he buys on Close, (but someplace he also mentioned Open)

6. Impractical stuff, like trading the Day Open (first tick), or trading too many signals in a minute, which would not be possible without advanced automation. But the 3 min timeframe seems reasonable.
i got a question from my friend , want to clarrify things clearly so posting here , no other intentions here ;;;;

first i want to clarrify that how trades are done on 15m tf and trade taken on 9;02;59......as my signal taking tf is 15m which is associated with daily and hourly but here i play a trick , system looks around 3mtf to finetune setup trades and for profit booking it will lead to upscenario 3>15>60>daily ,
so it is possible to take trade for finetuning on 9.02.59...........but system is of 15m /60/daily......

second thing i have already attached ami report showing that it doesnot look into future

third thing if we trade constant 4 trades overall time with than where question of irroponsible mm comes place.

fourth thing , let us assume i don't deduct brokerage than lets calculate 800 trades are there so how much brokerage will be there , it is not make much difference

regarding curve fitting , i posted here all years report seperately also , will post all months report in excel , with trade details , and my mm rules details with yearly wise , month wise and day wise ,

one question , all systems are backtested and using curve fitting , is there any system which is not curve fitted ?
if we get only half performance than also we survive ....isn't it ?

trades are completly on closing price of bar , there is no question of using opening price ....that concept is used in abhi report , i am aware of defects using opening price ,so i not used that ........

plz throws ur questions and query's as it helps in understand things clearly ......

it is person trading system , who makes money , it's his phycology , mentality , emotion handling capacity that makes him fit for trading , i am well settled with my earlier system and makes money there also , all is the thing , u have to know where it will perform better , u have to take double quantity there and when u know that this is not favourable setup around , u have to trade less quantity at that time , all main thing is .............let profit running trades as much as and shorten ur loosing trades as much as possible , and first thing is protect ur capital ......market is always there for trade , if u have no money than how u trade later ?

#### augubhai

##### Well-Known Member
Discussion on mech vs. disc systems in another thread. Posting it here since the Bakwaas thread is my gyaan and agyaan thread.

The posts are just my attempt to define a mechanical trading system. I am not claiming that mechanical systems will result in profits. However, since they reuse design and save mental/emotional energy, they are definitely more efficient than discretionary systems. Mechanical systems may be profitable or unprofitable, depending on many factors, just like a descertionary system may be profitable or unprofitable.

Just for the heck of it and for sake of debate, I'd say that all mechanical systems must have a discretionary system embedded within.

Conversely, all discretionary system run manually by a trader is just another mechanical system in which the trader is just a mechanical system's component!!!

In a well designed mechanical version, trader's ability to think/see/interfere is severely limited. In latter, most important thing that impacts profitability, the most essential trait of the trader's is to "act without thinking".

If I was going to ultimately build a mechanical system, I'd much rather have a Gigahertz computer than me doing the grunt work!!

I have fully automated developed mechanical system that is based on Gann/Fibonacci system and all I have to do was turn machine ON at 9AM and walk away. This worked nicely for 2 months but in 3 days, market asserted itself and ruthlessly trapped the system wiping out fair chunk of profits.
Right now, I am firmly in the mechanical trading camp. I think that if a discretionary system is a "system" at all, then it should be possible to turn it into a mechanical system. http://www.traderji.com/trading-diary/88102-bakwaas-trading-4.html#post829863

My father still believes that his tailor stitches better clothes than the readymades available in shops. But look at the effort and skill involved in buying, measuring and stitching the cloth. And you also need the expert tailor to do the work, else the result may not be better than the readymades.

On the other hand, the readymades are made by machinery and unskilled/semi-skilled workers. The skill part has been captured in processes and machines that can repeatedly produce the dresses more efficiently without expert supervision, and in volumes unimaginable by the tailor. It should not be very difficult to incorporate the logic/skill and customizations of my father's tailor into this mechanical process, if needed.
The biggest hurdle for traders, pros and newbs alike, is Drawdowns. In its face, traders start to bail-out on their methods. This effect is most pronounced in Mechanical trading as its drawdown is sharper as one has not developed the skill.

Surely consistent success in trading is NOT a function of ones mediocre skill in trading and exception skill in programming. There is no way around developing the skill to trade with ones eye.

Mechanization of the trading process is a dead end. To taste consistent success, one will have to lay himself bare to the elements of risk without the help of any machine, person or deity, but relying on his skill and skill alone.
Why are u inferring that a mechanical trader would not have developed the system though the same process as a discretionary trader? Why wouldn't a mechanical system handle drawdown as good or as bad as a discretionary "system"? In fact, one of the focus of a good mechanical system should be handling drawdown. Are you trying to relate mechanical trading with stupidity? For the record, mechanical trading ≠ stupidity.

What I meant was that if u have a experience based/discretionary "system", then you should be able to translate it into machine language (though sometimes it may be complex, and not worth the effort). A "system" consists of different components (MM, bias, price action, probablility, news, whatever), that interact with each other within boundaries (system rules). So if u really have a discretionary "system", then it should be possible to document it, and translate it into machine language. If it cannot be documented (machine translation may be more difficult and not worth the effort), then probably u do not have a system.

We are in an age where people are decoding genes, nano-particles, and the whole universe. If there is a discretionary "system", translating its rules will not be rocket science. Let me stress that I am not saying that it is necessary to translate every discretionary system into mechanical. I am just saying that if a discretionary "system" is really a system, then it should be possible to create a mechanical system out of it.

In other words, a discretionary system is a system whose rules have not yet been captured/documented.

And while I am at it let me also clarify a few more things that u seem to have assumed:

1. A mechanical system does not mean a code based system. You can be trading a mechanical system even if u have not seen a computer in your life.

2. A trading system - discretionary or mechanical - need not be based on the eye. There are many systems that are traded without ever looking at the charts.

Long winded post, but I was pissed off because there were so many false assumptions in the post. And I am not replying about the Sai stuff, though it may be apt...
I am sure that you see that I didn't say anything personal. I just presented my views somewhat frankly on the subject.

I have not tried to make trading mechanical as I believe if that were the case then the best computer making companies would saturate the market. They'd use the code of best designs to extract profits on largest volumes possible. And it may be happening who knows.

Cheers.
Nothing personal form my side either . I just thought that some of the assumptions in ur post were incorrect.

Any system - mechanical or discretionary - can be profitable or unprofitable. Even if profitable, it may not be profitable in all time periods and all markets. And, systems will continue to be impacted by actions of other systems (and traders). Check out the quants that make huge profits for some time - and then have to update or scrap the systems - when they start making losses... may be because other quants interfere.

Here is one simple example of a mechanical system - not code based, not visual - it is just something that worked, and there was no guarantee that it would continue working for ever.
@Augubhai

(I am doing this purely to earn an insight or two out of this conversation.)

Lets take a common pattern like a 15 min pivot, which has several variation depending on shapes, sizes, ATR, occurrence in a trend, momentum and selective cases of breakout failures.

Its nothing exotic yet I'd say that it would take considerable amount of programming skill to code it. The programmer must know this pattern to come up with a code. A rare combination i'd say.

It is important to get the code right, because after a few dull days and drawdown the entire responsibility would fall on the code to spot the pattern and take a position to end the week in profit.

To this, one may argue that this pattern entry cannot be coded as it is not a method, in a stricter sense, at all. However, there are several profitable combinations of emas, indicators and other formula which can make a good trade-able system.

For that, I can confidently say that such a system would have a subdued performance as there is nothing special about it. This system would just be a huge load of number crunching, which almost anybody has access to. In fact, whatever little profit that it may earn would only be because of the traders resolve.
Amit,

I think we maybe saying the same thing, though I am not sure about it....

If u say that a system uses 15 min pivots (maybe in addition to a lot of other patterns/data points), then do you have a consistent way of defining 15 min pivots - can u define 15 min pivots in a non-arbitrary way? If yes, then we can code it (maybe, even i can code it). U could have n number of qualifying rules for the pivot. For example, that it should L1 fractal pivot, or L2, or L3, or it should be a certain % off the previous pivot high/low, or that it should be considered only in the direction of the trend (define trend clearly), or RSI, or u might filter to ignore signals on wednesdays, or after 3 PM, or news days, or when volatility is high/low, or if previous signal was a winner, or risk is within x points, etc... whatever. shapes, sizes, ATR, occurrence in a trend, momentum and selective cases of breakout failures. If there is a clear definition, then it can be part of a mechanical system.

Else, if u say that patterns cannot be coded, then it means that the pattern is not yet defined. Taking the 15 min pivot, we might call one occurrence a pivot, and another not a pivot - for reasons like it was not visually clear, or that it was an inside pivot, etc. Then the question is whether u will consistently call it a pivot/non-pivot the every time you see it in the same context. If yes, then u have well defined rules that u just need to extract and document - and that can go into a mechanical system.

However, if there is a chance that the next time u look at the pivot, u might call it a non-pivot or vice versa, then this is subjective, and cannot form part of a consistent mechanical system (but can be part of a random mech system . It just means that u do not have a "system" at all - that it's just a gut feel approach. Such a "system" is still in the "design" phase, and not ready for implementation/execution. It means that every time u see the pattern, u mentally run a quick design iteration, and decide whether it is a pivot/non-pivot.

Just my opinion... like is it hot or cold now? why? because i feel it is hot, and i will switch on the the AC now? absolutely fine.... but if u can attach numbers (above x°C is hot), then u can have a thermostat in the AC do the work for you.

As I type, I have 2 mechanical systems running with open positions. I don't have to think about the patterns on the charts. That brainwork is not needed now, because the brainwork was done during the design phase of the systems - and I am just using the results of that brainwork...

(I can blah more, but already lot of blah...

#### augubhai

##### Well-Known Member
Can the past predict the future?

Everything’s tested in historical markets. The past is a pretty good predictor of the future. It’s not perfect. But human beings drive markets, and human beings don’t change their stripes overnight. So to the extent that one can understand the past, there’s a good likelihood you’ll have some insight into the future. - James Simons

#### OMSAI

##### Banned
augubhai..........NIFTY march series(Positional) ka bakwas target keya hai aap ka....:lol: regards RG

#### augubhai

##### Well-Known Member
augubhai..........NIFTY march series(Positional) ka bakwas target keya hai aap ka....:lol: regards RG
rajbhai... target to aapke thread me hota hai. mujhe to market target karta hai.

#### dell

##### Well-Known Member
augubhai ,
can u suggest me , how to hedge our positions infront of overnight adverse gaps ?
i have no knowledge of options ......
i have seen that inspite of our sl , say assume 60 points , trades with loss of 100,200 and 450 points are there and this all are due to adverse overnite gaps ,
so how to face this situation ?and is there any solution for this ?