Reasons for failure of Algo from backtesting to forward testing

oilman5

Well-Known Member
#2
algo is not failing . its the logic is faulty.So good trader can make money.On condition of future market it wont work. Its like computer chess. Av programme I win 80% time . against chessmaster higher mode I lose 60% , against Fritz I lose 98%.Yes computer chess far better than me.
But market has SURPRISE element -its in future, so if that condition can properly programmed ,u will get profit consistently.
Some sophi players under pro - give consistent return using algo - only that is not available to all.
,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,
bye
 
#3
How are you forward testing? Manual or automated.

you certainly can't automate forward testing, unless you become a dealer.

In backtesting , there are no emotions. if forward testing is done manually, emotions are involved. Hence Algos fail.

you need to completely automate your algos.

Get certified from NSE and become a dealer.
 

jagankris

Well-Known Member
#4
algo is not failing . its the logic is faulty.So good trader can make money.On condition of future market it wont work. Its like computer chess. Av programme I win 80% time . against chessmaster higher mode I lose 60% , against Fritz I lose 98%.Yes computer chess far better than me.
But market has SURPRISE element -its in future, so if that condition can properly programmed ,u will get profit consistently.
Some sophi players under pro - give consistent return using algo - only that is not available to all.
,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,
bye
One of the Professional algo told by an user.
No TA no levels,no Trend,no Support Resistance,no breakout/Breakout failure :)

Minimum Capital required - 25 Lakhs.

Rules
1.Buy Nifty whenever there is a 10 points dip.
or
Buy Nifty when the Nifty doesn't fall 10 points with in 8 Minutes.

No of lots to be bought - 40.
If Nifty is currently at say 6400.
It will buy Nifty till 6000.

Book profits when ever there is 10 point profit from the purchased price.

When asked about any crashes sudden fall the algo will buy OTM puts.
He didn't share that.
 

Algo

Active Member
#5
Hi

Reasons for failure of Algo from backtesting to forward testing:

1.Crap Datafeed.
2.Backtested data and Real market data having mismatch

or what are more reasons for this??

how to do real time forward testing without using own margin?

I completely agree. Some remedies for both these reasons are:
1. Algo should be back tested/forward traded with exchange broadcast data. It is also called level II data. Also, the dummy exchange where it will be tested should be well recognized & behaving like a mirror of real exchange.
2. If the whole algo is made over a single platform/system where broadcast data is taken, these issues will not arrive.

The main purpose of Forward testing or paper trading is you do not require to open an account with broker, hence, no real money invested. This question should not arise at all..
 

Algo

Active Member
#6
How are you forward testing? Manual or automated.

you certainly can't automate forward testing, unless you become a dealer.

In backtesting , there are no emotions. if forward testing is done manually, emotions are involved. Hence Algos fail.

you need to completely automate your algos.

Get certified from NSE and become a dealer.
I dont think you have to become a dealer for forward testing. Even a broker is not required since, you are not sending order to real exchange.
 

Algo

Active Member
#7
One of the Professional algo told by an user.
No TA no levels,no Trend,no Support Resistance,no breakout/Breakout failure :)

Minimum Capital required - 25 Lakhs.

Rules
1.Buy Nifty whenever there is a 10 points dip.
or
Buy Nifty when the Nifty doesn't fall 10 points with in 8 Minutes.

No of lots to be bought - 40.
If Nifty is currently at say 6400.
It will buy Nifty till 6000.

Book profits when ever there is 10 point profit from the purchased price.

When asked about any crashes sudden fall the algo will buy OTM puts.
He didn't share that.
Again, i do not know how you got the information of Rs. 25 lakh investment. 2 years back i got my algo developed, tested, and went live with a budget of 1 lakh. Now, i think it will be done even below the price. Algo trading is now not that expensive.
 

jagankris

Well-Known Member
#8
Again, i do not know how you got the information of Rs. 25 lakh investment. 2 years back i got my algo developed, tested, and went live with a budget of 1 lakh. Now, i think it will be done even below the price. Algo trading is now not that expensive.
The capital is for
For 40 lots Nifty Margin + drawdownes + Capital for Buying Nifty Puts.

Not for algo development.
 

Algo

Active Member
#9
The capital is for
For 40 lots Nifty Margin + drawdownes + Capital for Buying Nifty Puts.

Not for algo development.
Hi, great!! Which platform are you using. please gives a brief about performance etc.
 
#10
I guess after reading all these posts, I would provides bit of my insight.... It's not the data feed nor will any higher level of broadcast improve your results in live market. As a general rule of thumb.. Any TA based Algo can be at best right at 60% if the times. Usually in back test, we look at being trading always same quantity for buy and sell as and when the signal comes. Then usually to get better results we keep optimizing the parameters and do not realize that we are back fitting the data... The biggest difference that you can really make in live trading by using the simples of the algorithm is to focus very strongly on position management. Position management is the key differentiation in being a successful trader and an unsuccessful one.
 

Similar threads