Api (over) regulation - SEBI


I agree with any and all solutions once they “know” there is an issue with the retailer’s algorithm trading…. But they really don’t yet…

They speculate this is the issue… but deep down they know they want to stop the advertising that misleads people… instead of attacking that directly, they are trying all these BS means…

Prudent regulators cannot be reactive. They cannot police and run after 200 advertisers or bring down 2000 websites “after” a mess up. Right now there is no way to know differentiate a retail algo and a marketplace algo. When this is the case, there is indeed an issue.

For example, if an algo market platform with 1 lac clients subscribed to the same Algorithm1 is tweaked to trigger 1 lac buy orders of a stock, while the founder of the same algo company is ready to dump a million of this stock from his personal portfolio… and does this for a month across all caps… this will lead to banning all API orders after this event.

The technical details provided in the earlier largely solves the problem, unless of course SEBIs intention is to willfully target retail API clients… especially as the public has been left to trust the good intentions and has neither been notified about the antecedents of the “experts” group that made the proposals nor the professional relationship they maintain with SEBI.

Love you for your thoughts…

But IMO every regulation is always reactive… it has to reach a certain level of awareness until the regulators are aware of the problem, right?

I would love that regulators consider second order and third order effects of their decisions… but I have not seen any proof of this in any recent judgements by SEBI or the Supreme courts… hence my questions… nothing personal… pls understand that…

Love intellectual debate (as opposed to emotional debates).

I’m not.conceding…. (though I will easily, if you win me over, intellectually)… I’m fundamentally a trader, right? If proven wrong, I’ll reverse my opinion :joy::joy::joy:

Limit orders at circuit can easily simulate market orders. If limit orders are alllowed there is no reason why stop Limit should not. There are many ways on how people trade and invest, i dont think its right to come up with arbitrary restrictions like this.

Why ? Markets abroad have much more volume and do not have such restrictions. Why limit the capability of a retail trader/investor to that of someone who can place only 1 order a minute. Where does this time limit come from? I can go to kite web and send more orders per minute through shortcuts.
Api is not the only way to automate. I could send 3 orders per second through ahk over Nest, and that had limitations of GUI. A reverse engineered solution will be even faster. So this does not solve the problem but again adds arbitrary restrictions that limit the capabilities of normal users.

Some sane Limits make sense but these are too low. Its easy to consider only one’s own use case and disregard others as invalid.
Also, you can have people sharing a single ip address. So they as a group get restricted to one limit.
And again, i can send many more orders through hacky solutions that simulate desktop/web interactions and so you will only restrict normal people and not algo sellers. Unless ofc you mean to apply restrictions on all interactions - web or api - but again these seem too limiting. Higher limits might make sense to ban algo sellers.
But they could just buy multiple network cards/ multiple pcs connected to separate networks.
They could use multiple aws type remote desktops, i dont have experience with them but i assume they will have unique ip addresses.
They could install software on client machines to send the orders, and this easily bypasses ip limits.
It adds overheard but these rules can also be worked around by system sellers who would have the resources to do so and you limit everyone else.

Also algo sellers can provide a valuable option to people, beating Mutual funds on net/dd is relatively easy in trading. We need ways that do not discourage growth of ecosystem, but yeah prob more bad actors because people chase high returns disregarding risk so bad guys become more visible.

There really should be some space for proprietary solutions too although i think many wont accept this. As a trader who built something himself, why would i disclose details and allow them to leak. I do not provide algos to others but these kind of rules imply that people like me wont.
And so you remove some of the people who do have capability of deriving a good edge. So you will have more no-edge or mediocre edges in the market.

Anyway, US / EU markets are mature and algos and apis are widely used and they could be used as reference.
I think nithin has covered all of the points very well in his interviews.

We need reasonable and practical regulations. There is aways some risk in the market, it cannot be eliminated but bad actors have to be discouraged/penalized.

Obviously, am speaking of a framework… specifying limits only as an example… perhaps your algos will get affected by it… perfectly understand. But right now, it is about slowing down the API machinery rather than getting it removed. Referencing mature markets doesn’t help… we need to have mature regulator for that. Do we?

  1. MARKET ORDERS: Are really notorious. So are SL-L, SL-M orders. Precisely why “approved” algo features bracket order din’t allow them, until the infrastructure got ready. Runaway trades, loop trades you name it… market order is the culprit. Simulating market order using Limit orders is not same as Market orders as market orders never doesnt even reach the exchange order book (HFTs pounce on them before that). It “looks” same due to abundant liquidity.

  2. LIMITS: It is not about one use case. It is about making algo trading reduced to the level of automated trading. If a person can place 10 orders (MANUALLY) in a min, then a multiple of that needs to be the limit for API. Same thing for day wise limit.

  3. TRACKING: By ip addr, I did mean all devices. But only counting API orders (not manual orders from web, mobile, terminals, etc). If people share same ip addr (not sure how as even docker containers are assigned separate ips), still their client id are unique right? That combination is good enough to put a security rule.

  4. SELLERS: Really not bothered abt Algo sellers, they hv deep pockets and know how to get things done… it retail investors who has no one to bat for them… if their algos of these sellers were so good they shd start an quant fund, follow PMS compliance, get RI/RIA or a new AA certificate, get auditing done, etc… instead of asking a fee. Perhaps asking to disclose trading logic is not right, but they need to demonstrate that all risk management controls are adhered to with periodic appraisal. SEBI does not have a IT cell capable of bringing down sites, but it has control over RI/RIA, Brokers, Exchanges. Regulator can only work with the resources it has.

  5. API services: We know there are alternatives to API, they will always be. But taking a stance against restrictions (even unreasonable ones) will ensure that the bad apples across the industry from Institutional players to Exchanges to Regulators, who RESENT retailers using API for disciplined trading and investing, will SUCCEED in making brokers stop API Services altogether. Some of us had to wait for a decade for API services to begin in India, but if removed now, it will take another decade to come back.

I do not understand what you are saying. Perhaps i m mistaken but i think you are wrong. All orders go to exchange books whether we can ‘see’ them or not. A limit order at circuit or with 0 price is same as market in terms of impact. There is no extra risk from SL-L orders and they are a valid and needed way to manage risk. Removing market orders forces people to define limits and that can help in avoiding mistakes. So that can make sense esp in illiquid markets.

People behind a router will have same public ip.
yes you can make a combination of ip and client id but then that does not help much does it ? There are easy workarounds as i said and a major risk of system sellers is the impact of distributed orders that escape RMS rules (which are tailored to individual accounts). 10000 people can send 1 market order in same market - that is the risk of system sellers.

I am only pointing out the logic flaws. Cannot do much about diplomacy, i do not consider that my input will have any bearing on the result. SEBI babus will do their own thing and we have to adapt. Only Influential people might be able to show reason to SEBI.

One positive out of all of this for me is that now i will be forced to look at HTF systems - something that i have neglected so far - always planned for future but not yet acted on as i focused on scaling up my current system.
Just hoping that i can continue trading current one, probably will have to do jugaad again …

  1. You are thinking of SL-L orders only from the point of reducing risk for the position taken. But I am speaking of the risk they pose when used as triggers for entering a new position, which was preciously why it was denied by Exchange for Bracket Orders feature (which is an approved Algo product). Also, regarding liquidity, API orders are allowed for not so liquid assets as well, which is when NOT allowing Market orders help. If API orders are restricted to specific stocks or indices then perhaps SL-L/M, Market orders could be less risky.

  2. True, 10000 client account can send a single order for the stock, which is why “speed bumps” to delay 1st leg orders (that is, entry orders) upto 1 minute should be enabled by the broker. Limit orders DELAYED randomly will reduce operational and concentration risk. This is the one of techniques considered in HFT arena (Ref: Asymmetric speed bumps: A response to high-frequency trading | VOX, CEPR Policy Portal), my suggestion is to port this to our context so that algo sellers’ trading PnL output will get broken down.

  3. Regarding SEBI, like any other corporate entity, there will be good and bad elements in there, ranging from utterly corrupt to extraordinarily visionary, not that these are mutually exclusive. The idea is to do things that is good for the industry as a whole, develop fintech ecosystem based on technology, perhaps even expand our products to the world… or else we will end up depending on whatever foreign players innovate and be end users of twitter, facebook, whatsapp, google and whatever is invented abroad for the next 10 centuries.

  4. Regarding looking at HFTs, less said the better… one could look, look and get nostalgic at best. Just a passing comment, thats all. All the best, many things are doable.

agree, that is why i insisted that the algo review process by SEBI should be in the price range of small trader.
the review process can be free too, if the developer somehow proves that the algo can’t be run any other a/c’s and only on self account.

Atleast this move will force these parttime algo creators to get their code reviewed.

Just to save few retail traders, we cant have such a big regulatory hole.
SEBI needs to address both small & big parties

I don’t really agree with most of the points. You are considering only your use case and disregarding others.
I lot of these restrictions can be easily bypassed by not using api and if applied to all - it is much much worse than what SEBI is proposing. Completely draconian.

So its not solving anything, but creating restrictions for everyone. Some wont mind and for some it will be game breaking. No reason why US/EU regulations should not be looked at, dont have to copy but they have apis widely available.

This is my final reply here as there is nothing more to add. Whatever will happen, will happen …

  1. SLL is not any more risky vs limit orders, perhaps it can concentrated but this applies to limit too when sent together - and reasonable limit prices will limit the risk. That is how i enter over 1000s of trades without issue.

  2. 1min delay is like sending us to the dinosaur period. It might be fine for people investing or trading Higher timeframe but not for intraday.

  3. api usage is nowhere as risky as HFT where RMS systems apparently have tough time catching up as explained by Nithin in an interview.

  4. HFT is beyond me. I meant trading Higher Timeframe - sending orders at the end of the day. Less work, perhaps lower returns too and no impact of api issues as workaround solutions will be acceptable.

  5. I don’t need my code reviewed by anyone. I will not be leaking system info - my hard work - to others, it makes no sense. These people cant even keep phone numbers secure.

Public facing people may need some kind of regulation, oversight/monitoring/enforcement but even there api rules seem too restrictive without solving anything.
Maybe best to make sure that end users are also educated on how trading works, but greedy people will always take the bait.

I understand your position and agree with them. And I am not speaking from my use cases. It cant even be adapted for higher timeframes. In all probability will have to move to alternative non api solution like many of us.

Still, I am convinced that API services need to be retained even if restrictions are applied that make them almost manual entry for all purposes except for the automated part . As 3rd party algos sellers are downright unethical and dangerous, the impact of restrictions is not important. the important thing is nullify the control they get by operating their clients’ accounts. besides, they already know how to make great profits like 300% and 1500 % etc, right? let them use their own money or upgrade to quant funds.

let us see, but do keep the inputs pouring in.

1 Like

we are planning to re-enable Zerodha Connect API and start programming some trading systems, is it a bad time to start spending time and resources on this project ?, as brokers may stop providing API services?
your suggestions are highly appreciated.

I would think so…. Based on Nithin’s response in point 2 below…