Can accuracy be increased in algo trading from 40%?

stoch

Active Member
#21
An algorithm is a set of instructions which a computer is programmed to follow in order to carry out a particular task. In the case of trading, these defined instructions pertain to placing trades with a speed and precision which a human trader would not be able to achieve. The instructions set the timing, quantity, price limits or any other relevant criteria that would ensure the most profitable trade.


It is important to keep in mind that algotrading is a market strategy and not meant for long-term investors.
Yeah and it relies on price data a lot while investment strategies process fundamental data which help you to foresee possible shifts in supply and demand and hence predict price movements.
 
#22
Hello All,

I recently started algo trading using python with Zerodha. I tested multiple strategies and whenever I backtest any of these strategies I noticed that in a month the accuracy never goes above 42% to 44%. Are you able to achieve a better accuracy rate?

My algo setup is simple:

Go long when 10 EMA crossed above 20EMA and recent close should be greater that PSAR.
Exit condition is if 10EMA cross below 20EMA or PSAR is above the last close.

The same is reversed for short. Do you have any suggestions on how the accuracy can be improved?

Thanks in advance.
Simply reverse the strategy and your accuracy will also reverse. Means from 40% to direct 60% - if its 30% then on reversing the strategy it will be 70%

I often see people complaining about worst accuracy in tips from so and so strategies or companies. I think that, if accuracy is that much poor of that company, then why don't you just do the opposite. if they say buy then you sell and if they say sell then you buy.......in following strategies or tips you are making loss, then why don't you do the opposite.......?
 
#23
The accuracy of algorithmic trading engines is fantastic. When well implemented, a marginal error as low as zero is attainable. However, the lack of enough training data is a big blow to the implementation of such algorithms.