Explained: What is the NSE colocation controversy; Ravi Narain's, Chitra Ramkrishna's role in it

Also read: NSE scam: Sebi can identify ‘yogi’ in 10-15 days, says cyber forensics expert

That is because some players could jump the queue and pounce on the share or dump it before your order was executed.

By figuring out the loopholes in the trading system, as a result of which their buy or sell orders would be executed ahead of the rest of the market.

Because the difference was not big enough to arouse suspicion. It could have been a rupee or two, or sometimes just a few paise more per share. It is like filling petrol for Rs 1,000 and later being told that you were given petrol for only Rs 990.

Like in the example of filling petrol, an individual motorist may have been shortchanged for Rs 10, but everyday hundreds of motorists were being ripped off. Likewise, in the market, there were millions of trades that happening daily.

Also read: NSE-Yogi saga: CBI grills former NSE CEO Chitra Ramkrishna

In August 2009, the NSE said that it would offer colocation facility to stock brokers who were willing to pay extra. Simply put, brokers could place their IT servers, right next to NSE’s servers for a fee. This meant that the prices broadcast by NSE's trading system would first reach the brokers whose servers were closest to NSE's servers.

Illegal, no, but you could say it was unfair to those brokers who could not afford it. Non-colocation brokers would receive the prices with a lag, and this lag would be anywhere between a few milliseconds to a second or maybe even more, depending on the physical distance from the exchange's servers.

Interestingly, NSE introduced this facility without SEBI first putting up a discussion paper on the subject, as is usually the procedure.

The whole controversy is referred to as the NSE algo trading scam or High Frequency Trading (HFT) scam.

Technically, it cannot be called algo trading scam or an HFT scam. Algorithmic trading or algo trading is a method of executing a large order using pre-programmed instructions. These instructions could be based on different factors such time, price, and volume.

High frequency trading (HFT) is a type of algo trading characterised by high speeds. The algo is rapidly shooting buy and sell orders at a speed that cannot be matched by humans. Besides sophisticated algorithms, fast access to order books and colocation too are vital to success.

In December 2009, NSE started offering its colocation members tick by tick (TBT) data on the price feed. Think of this as being able to watch ball-by-ball coverage of a cricket match, instead of only getting to know the score at the end of each over.

It was almost like getting a pair of glasses with X-ray vision. TBT would allow a broker to see every single buy and sell order sitting in the system. Say a stock is quoting at Rs 100. A regular broker can see the best buy and best sell orders from Rs 100 to Rs 98. In other words, he is getting a snapshot. On the other hand, a colocation broker was able to see every pending order, and he could see it granularly. Say, a regular broker can see cumulative buy orders for 10,000 shares at a certain price. The colocation broker could view every individual bid. He could see that there is a single bid for 5,000 shares, 10 bids for 100 shares, 20 bids for 200 shares, three bids for 250 shares and so on.

Interesting!

There is more. The colocation broker could also see every single order modification in the order chain. For instance, a buyer may decide to lower his bid. So he will still be in the system, but at a different price point. The regular broker may assume that the order has been cancelled, but the colocation broker knows that the buyer has just moved a few rungs down the ladder, but is very much interested in buying.

Knowledge of the depth of the order book can be misused to manipulate prices, as it is clear at what price points the maximum buyers and sellers are.

But colocation in itself cannot help pull off profitable trades; you need to have sophisticated algorithms that can process data and then come up with good strategies to capitalise on it. Brokers using the colocation facility had all of that.

Here's where the game becomes interesting. While in the check-out queue at a super market or in the check-in queue at the airport, you are always trying to figure out which line will clear faster, right? What if you knew beforehand which queue would clear the fastest? That is what some brokers were able to do.

The data was being disseminated to its brokers through the TCP/IP protocol. Without getting into technical details, the flaw of the TCP/IP architecture was that data was served on a first-come, first-serve basis. This meant that the broker who had logged in first to the server would stay in front for the rest of the day and could have the first shot at the trade.

Some brokers figured this out, most likely with the connivance of officials managing the system. They also got to know that all the servers were not started at the same time in the morning. So the brokers who knew which server would be started first would latch on to that server.

Additionally, some brokers also figured out that there was an advantage for being on the server with the least load (least number of trading members logged on it). That way trades would get executed faster than those of members who were logged on crowded servers.

When other brokers complained about it, NSE first introduced a load balancer whereby the orders would be automatically allocated to the least crowded servers.

By April, 2014, the NSE implemented the multi-broadcast TBT protocol. As the term broadcast suggests, the data feed would now be disseminated simultaneously to all colocation brokers, instead of on a first-come, first-serve basis. There was no longer any advantage in logging in first.

To an extent. But the smarter players still managed to game the system. The load balancer ensured that orders were automatically routed to the least crowded server. But NSE also had a backup server to which colocation brokers could log on if there was trouble with the primary servers. Some brokers would access this server regularly even when the primary servers were working fine.

Yes. Some brokers stopped accessing the backup server following the warnings. But some others continued to do so, and no firm action was taken against them.

SEBI has not been able to figure it out. Neither have other consultants who analysed the matter.

Because the brokers who gained from these illegal activities would have divided orders across front entities so as not to arouse suspicion.

Once someone had access to the entire order book ahead of the market, all he or she had to do was share that data with accomplices and split the trades.

According to whistleblowers, the players with this advantage were easily making between Rs 50-100 crore cumulatively every day.

Also, they allege that some of the players involved in this scam were foreign institutional investors. So a lot of money has "legally" gone out of the country.

The SEBI has banned OPG Securities from the market for five years and asked it to pay Rs 15.57 crore by way of disgorgement of illegal gains. The case is now pending with SAT.

Ravi Narain was the MD and CEO until March 31, 2013, and Chitra Ramkrishna, Deputy CEO during that period. Chitra then went on to succeed Narain as MD and CEO and was in office until December 2016 before being forced out.

The irregularities in the colocation facility happened between 2010 and 2015.

In their replies to SEBI's show cause notices, both of them have said that they were not familiar with the technology and that they had gone ahead with the advice of the functional heads. They also said that they were not involved in day-to-day operations of the colocation facility.

No. There were two parts to the NSE colocation scam. One was the flawed technology architecture that disseminated the tick by tick (TBT) price data in a way that the trading member logging in first onto the colocation servers would have an advantage compared to other trading members. Two, the brokers managed to log in ahead of others by figuring out which servers would be started early, and regularly getting on to the servers which had the least load. This, it was alleged, was by conniving with some officials of the exchange.

The employees could not be faulted for the technology architecture and SEBI was unable to prove the involvement of any IT staffer in helping some brokers get an unfair advantage.

In this case, too, the IT staffers could not be faulted because the exchange did not have clearly defined guidelines on brokers accessing the backup server.