One of the most requested feature from retail traders is to fully automate their trades. It's understandable because automation just makes life simpler. Companies like Zapier, IFTTT are quite successful and are built on the premise of automating tasks for individuals and institutions.

Yobee Research's core offering is deep learning based quant strategies, specifically intraday. However, providing recommendation is half the work done -- timely and accurate execution of orders is equally essential. If we generate 10 trading signals, the trader has to place 30 orders (1 entry + 1 stoploss + 1 target order against each signal). Few technically savvy traders might write code to automate this, but majority traders are stuck with manual order entry.

There are a couple of ways we can resolve this problem.

Uploading a file to broker

Prepare the orders in a file and upload it to your broker. This was a fantastic solution 10 years ago but is now too cumbersome.

Sharing password

Some companies ask for a trader's internet trading password. They automate the login process and place orders. However, this is very insecure. Imagine giving away the password to your bank account. Our broking account is also sensitive and private.

It is also illegal. In May 2020, National Stock Exchange (NSE) issued a circular to stockbrokers, asking them to create awareness amongst their clients. Giving away your password means giving up control of your trading account. It can lead to unauthorised trading and losses.

Additionally, companies asking for passwords are exposing themselves to complaints and lawsuits.

So, we immediately discarded this approach as an option. Someone might do this for a select few traders like friends & family but not for the entire retail traders' community. This approach will not scale well, and regulators will shut it down.

Partnering with Stock Brokers

When designing our app for retail traders, we had many productive discussions with legal, product & compliance teams of stockbrokers in India on what is permissible.

Now, a customer of our partner broker can log in to the app, using his broker's credentials. This is just like "Sign in Google". We never get to know the user's password and security is not compromised. This approach requires additional effort from brokers and us but is the correct way to do things.

A successful login gives us the ability but not the authority to automate orders. There are a few restrictions we enforce. To name a few:

• Traders need to re-login after 24 hours
• Consent from the trader to automate trades

In our earlier example of 10 signals, we take approval from traders to place 30 orders. It is exact 30 orders, not 29 or 31. We maintain careful logs and ensure that no one in the company can abuse this trust. All this happens in a single click for the trader though.

Almost all companies do not take this consent. They are, again, opening themselves to complains and lawsuits for unauthorised trading.

The nerd inside me is still not satisfied with the solution. There's manual intervention required of the trader. However, it's a necessary compromise for now. We want to establish trust & safety with our customers, partner brokers and regulators.

At least, now, orders are always accurate because they are system generated in a single click — but this is just one part of the solution. The other challenge is the timely generation of orders — that is timely consent from the trader. In intraday trading, signals have a short shelf life and consent is required as soon as possible.

Currently, and traditionally, strategies generate signals randomly throughout the day. Considering we need to take consent every time, a trader has to be always alert. I have not observed any company, that offers algo solutions to retail traders, working to ensure that traders do not execute state signals. They do not design their own strategies and have no quality checks or control.

Overcoming this problem is a core undertaking at Yobee Research. Our initial approach is to generate a basket of signals at pre-defined intervals. To achieve this, meant a complete re-thinking when designing quant trading models and neural networks.
This is another detailed discussion that I'll probably cover in a future essay.

To conclude, a vertical integration — from creating unique strategies to controlling order execution is crucial. It allows us to design strategies that fit into the trader's behaviour and current regulation.