Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Lessons learned building an ML trading system

tradientblog.com

Post reply on HN
Print view View original

41–50 of 160 posts

Re: Lessons learned building an ML trading system

#41

Post by Akababa »

onlyrealcuzzo wrote: post #7

Is this the author? I would love to know how this fared recently in the large sell-off. What he says about some markets possibly being predicable rings true to me. But the article was far from convincing that the BTC market is actually predicable. The natural assumption should be that the author was in the right place at the right time. Although he went through great lengths, I'm not convinced this is anything other…

Through hypothesis testing you can estimate the probability that this was due to luck is very low. Assuming that a monkey would have a 50% chance of profiting on a day, the chance of going a month without a losing day is less than 1 in a billion.

Re: Lessons learned building an ML trading system

#42

Post by proverbialbunny »

echelon wrote: post #20

Can this same strategy be leveraged on zero-fee stock exchanges? Why is crypto the target here?

Low trading fees is what originally caused me to change gears from etf programming trading to btc programming trading in the early days. I imagine it is the same for most.

Re: Lessons learned building an ML trading system

#43

Post by pjc50 »

garmaine wrote: post #33
pjc50 wrote: post #27

Earlier quoted context omitted.

Binance is in Malta, a somewhat corrupt tax haven; are you sure it's illegal there?

Malta is in the EU.

OK, so which EU directive are we dealing with here? Which regulator does Binance answer to?

Re: Lessons learned building an ML trading system

#44

Post by turk73 »

megaframe wrote: post #40
nickreese wrote: post #11

After having spent an insane amount of time in late 2017/2018 building an HFT bot for Binance I can say this is a pretty solid article. In our case we were doing triangle trading between BTC/ETH/USDT pairs and had our buys/sell delay down to 3-7ms. At one point moving 0.3-0.7% of Binance’s daily volume. Few notes: * Finding an objective point of truth for value when all of the currencies are floating is hard but vita…

Ran an ML model years ago had a number of great months then out of no where no trade I or the ML would make, would work. Looked like someone was front running my orders and messing with my trades. Weird delays, trades would take to long to go through all, and all sorts of odd events on Level 2. Ended up shutting it down took a good 2 months before my manual trades started going through at a normal rate again. This is…

Take it from me, they are fucking with you. There are algos out there that detect "real" traders like you and me and endlessly fuck with you on equities long/short and all options. I assume also on commodities and Forex, but have no direct knowledge. No clue about cryptos or any of that trash, but you're swimming in shark-infested waters with cryptos anyways. That stuff is all completely manipulated trash.

My suggestion if you want to beat the "market" is to develop some kind of algo that can trip up other algos and cause them to show their ass and then buy that dip.

Re: Lessons learned building an ML trading system

#45

Post by nickreese »

carlsborg wrote: post #18
nickreese wrote: post #11

After having spent an insane amount of time in late 2017/2018 building an HFT bot for Binance I can say this is a pretty solid article. In our case we were doing triangle trading between BTC/ETH/USDT pairs and had our buys/sell delay down to 3-7ms. At one point moving 0.3-0.7% of Binance’s daily volume. Few notes: * Finding an objective point of truth for value when all of the currencies are floating is hard but vita…

> the orderbooks were slightly different between bots That sounds like a big deal. If this is repeatable, you should document it better. Unregulated doesn't mean a license to do blatantly illegal things. Crypto exchanges certainly get taken to court. Sure it wasnt your book building algo and a snapshot retrieval race?

After 2 weeks of digging through our code I’m 100% confident it wasn’t our code. Was able to hook up to a total of 3 additional accounts via api, all 3 books matched, and were started at different times.

We wrote support and they told us they would investigate. Never heard back.

Re: Lessons learned building an ML trading system

#46

Post by symplee »

Akababa wrote: post #41
onlyrealcuzzo wrote: post #7

Is this the author? I would love to know how this fared recently in the large sell-off. What he says about some markets possibly being predicable rings true to me. But the article was far from convincing that the BTC market is actually predicable. The natural assumption should be that the author was in the right place at the right time. Although he went through great lengths, I'm not convinced this is anything other…

Through hypothesis testing you can estimate the probability that this was due to luck is very low. Assuming that a monkey would have a 50% chance of profiting on a day, the chance of going a month without a losing day is less than 1 in a billion.

How many monkey bots are flipping the coin daily? Selection bias just needs one.

I very much look forward to the author's follow up post at the end of 2020 to see if another 5k turns into 200.

Re: Lessons learned building an ML trading system

#47

Post by throwaway_e4WNi »

nickreese wrote: post #11

After having spent an insane amount of time in late 2017/2018 building an HFT bot for Binance I can say this is a pretty solid article. In our case we were doing triangle trading between BTC/ETH/USDT pairs and had our buys/sell delay down to 3-7ms. At one point moving 0.3-0.7% of Binance’s daily volume. Few notes: * Finding an objective point of truth for value when all of the currencies are floating is hard but vita…

> the orderbooks were slightly different between bots

As somebody who still runs a profitable bot on Binance I find this hard to believe.

Also all order book related endpoints/streams are public, so queries/subscriptions are not tied to a specific account.

Re: Lessons learned building an ML trading system

#48

Post by wpietri »

jjoonathan wrote: post #36
jrockway wrote: post #21

Earlier quoted context omitted.

I feel like commercial activity finds value and removes it from the system. To create value, you need to give someone else money. If I wanted to build a better society, I'd make sure everyone had the best possible education. Paying for this would bankrupt me, maybe even bankrupt the entire country. But with every single person in the country walking around with a deep understanding of music, art, mathematics, enginee…

The reason why you feel this way is because there is a large difference between value† as defined by markets (which I'm writing with a dagger †) and value as defined by everyone else (which I'm writing without a dagger). Value† is weighted by wealth. Value is not. Feeding a starving African village generates value, but feeding a starving African village does not generate value†, because the village has no money. Figu…

For what it's worth, I use "makes money" for creating value†, and "creates value" for the common sense.

For those interested in the topic, I've found the Lean Manufacturing literature helpful. They're usefully obsessed with creating customer value, meaning doing or making something valuable for their customer. Which has some of the problems you mentioned, in that traditional commerce mainly serves people in proportion to the money they have. But you can pretty easily extend the Lean approach to value to non-commercial situations.

Re: Lessons learned building an ML trading system

#49

Post by wpietri »

cco wrote: post #9

And in the end, what value was created? "Liquidity in the BTC market"?

Exactly. Regular speculators at least have the (hazy) claim that they're reducing the spread for people trading markets with actual utility. But given BTC's lack of economic meaning, that doesn't work here.

If anything, providing BTC liquidity might just be increasing net societal harm.

Re: Lessons learned building an ML trading system

#50

Post by nickles »

milesvp wrote: post #37
nickreese wrote: post #11

After having spent an insane amount of time in late 2017/2018 building an HFT bot for Binance I can say this is a pretty solid article. In our case we were doing triangle trading between BTC/ETH/USDT pairs and had our buys/sell delay down to 3-7ms. At one point moving 0.3-0.7% of Binance’s daily volume. Few notes: * Finding an objective point of truth for value when all of the currencies are floating is hard but vita…

Even in regulated markets there are problems. I have a friend who built an HFT algorithm that he was using for trading stocks. He had some really good results with early testing. But at some point he was confused why many of his buy/sell orders weren't being executed despite being open for several minutes (hours?) his algo would make bids far away from the current spread, anticipating movement. He finally concluded t…

> But at some point he was confused why many of his buy/sell orders weren't being executed despite being open for several minutes (hours?) his algo would make bids far away from the current spread, anticipating movement

* Did his levels actually get crossed? It could be that he never had the opportunity to get filled at the price he hoped for.

* What was his queue position? Stock exchanges tend to be price time ordered. If others had submitted orders at the same price before he had, they would have priority when getting filled.

* Apropos the prior point, was his broker actually submitting orders when he sent them? Some brokers may avoid showing deep out of the money orders, which could have affected his queue position.

> He finally concluded that some institutional traders must have access to sub-penny ordering, despite it being against regulation.

There's nothing wrong with this. Reg NMS rule 612 permits sub-penny price improvement [0].

[0] https://www.sec.gov/divisions/marketreg/subpenny612faq.htm#q...

Post reply on HN