4 ms·

This is some interesting analysis, but all of the causal language is unsupported -- and I think mostly inverted from the reality. Here is an equally supported

This is some interesting analysis, but all of the causal language is unsupported -- and I think mostly inverted from the reality. Here is an equally supported description: - Retail and futures traders create instability by placing leveraged trades and stop orders that amplify swings. - Market makers are aware of instability and design their bots to turn off so that they don't end up on the wrong side of a liquidity cascade. - People with large orders often cancel them in order to improve their orders when chasing the price. (This happens in non crypto markets too, but some of those markets have incentives and regulation to force market makers to provide stabilizing liquidity.) The most explicit manipulation is the news outlets designed to amplify positive news. But even that can be explained by desire for clicks as much as short term market shifts.

VHRanger 5y ago
Author here. I considered this, but rejected most of those hypotheses. > Retail and futures traders create instability by placing leveraged trades and stop orders that amplify swings. True > Market makers are aware of instability and design their bots to turn off so that they don't end up on the wrong side of a liquidity cascade. Algorithmic traders, yes. Market makers absolutely not. MMs want to be there as much as possible in liquidation cascades, because bid/ask spreads are huge. MMs effectly print riskless money in these situations (which is why you see Alameda and DRW issue so much USDT in these events). > People with large orders often cancel them in order to improve their orders when chasing the price. It's absurd to think this is the case when the order cancellation pattern is precise to ~3ms and repeated >5 times in a 30 second span. What you see on 26/7/2021 0:59:20-1:00:45 is intentionally done by bots designed to do this. > But even that can be explained by desire for clicks as much as short term market shifts. Agreed, there's a footnote that posits other actors than the momentum ignition traders could have planted the fake news because they saw the same opportunity
- stepanhruda 5y ago
  Market makers don’t print riskless money, they are affected by a very real risk of divergence loss.
  - VHRanger 5y ago
    You dont have a risk of divergence loss if you're simultaneously buying and selling the same product in the same order book. Bid/ask spreads were as wide as double or triple digits during the liquidation event. At this point the only limit to riskless profits is your liquidity and the speed at which you can execute.
    - iakh 5y ago
      Market makers can't just conjure up a counterparty. The spread widens because lack a liquidity increases directional risk. If it were truly riskless, others would jump in to close the spread.
      - wickoff 5y ago
        This is exactly what market makers do. During violent moves prices vary wildly between exchanges and market makers become takers. They instantly buy and sell the same amounts on different exchanges without risk, then re-balance their accounts.
        evrydayhustling 5y ago
        No. To buy and sell on your own schedule, you must take liquidity, paying the spread. This is not what market makers do, they sell liquidity. And it becomes less possible to find efficient clearing trades during rapid moves, because the spread widens exactly because market makers don't want others to arbitrage them.
    - gruez 5y ago
      > you're simultaneously buying and selling the same product in the same order book. How's that even possible? Suppose the bid is at $95, the ask is at $105, and you receive a buy order for $100, you can't "simultaneously" sell that and make a profit.
    - stepanhruda 5y ago
      You can do arbitrage without being a market maker and you can be a market maker without doing arbitrage.
- kikimora 5y ago
  Market makers likely know how thing works and react to a volatility spike. I wonder what was the traded volume during the event and how it matches normal order book disposition. If only retail orders left in the boom then there will be very low volume. If there were pre-planted orders from the manipulator then all liquidations should hit them. Remaining orders should disappear shortly after the event, as soon as liquidations complete.
- evrydayhustling 5y ago
  > Algorithmic traders, yes. Market makers absolutely not. MMs want to be there as much as possible in liquidation cascades, because bid/ask spreads are huge. MMs effectly print riskless money in these situations (which is why you see Alameda and DRW issue so much USDT in these events). This is a deep misunderstanding of market structure. Market makers are algorithmic traders. In times of price stability they benefit from capturing the spread many times. But it is not risk free: they lose money when offering liquidity to "informed trades" (those followed by a price move) because they are holding the wrong position through the price move. For that reason, every market makers implements fail-safes that stop offering liquidity when they are not confident they will be able to clear their position at a break even price.
huac 5y ago
I agree with most of this and would add that the article could be organic and the subsequent trading opportunistic (rather than being tied together). I am somewhat surprised that bots react to news within 5ms -- your execution delay on crypto exchanges is quite a bit longer than that, since the exchanges are generally hosted in public clouds and you are subject to network latency to get in there. Even within the same cloud and even within the same k8s cluster, you should expect 2-4ms for an inter-pod hop. That being said, maybe there is a hedge fund literally in the same k8s cluster as FTX...
- mthoms 5y ago
  Well, Alameda and FTX were founded by the same person. Up until very recently Sam was CEO of both companies at once. https://en.wikipedia.org/wiki/Sam_Bankman-Fried https://en.wikipedia.org/wiki/Sam_Bankman-Fried Edit: I suspect Alameda uses data from FTX but might not trade there: as per the article, the manipulation needs thin order books to work.

VHRanger 5y ago

Author here. I considered this, but rejected most of those hypotheses. > Retail and futures traders create instability by placing leveraged trades and stop orders that amplify swings. True > Market makers are aware of instability and design their bots to turn off so that they don't end up on the wrong side of a liquidity cascade. Algorithmic traders, yes. Market makers absolutely not. MMs want to be there as much as possible in liquidation cascades, because bid/ask spreads are huge. MMs effectly print riskless money in these situations (which is why you see Alameda and DRW issue so much USDT in these events). > People with large orders often cancel them in order to improve their orders when chasing the price. It's absurd to think this is the case when the order cancellation pattern is precise to ~3ms and repeated >5 times in a 30 second span. What you see on 26/7/2021 0:59:20-1:00:45 is intentionally done by bots designed to do this. > But even that can be explained by desire for clicks as much as short term market shifts. Agreed, there's a footnote that posits other actors than the momentum ignition traders could have planted the fake news because they saw the same opportunity

stepanhruda 5y ago
Market makers don’t print riskless money, they are affected by a very real risk of divergence loss.
- VHRanger 5y ago
  You dont have a risk of divergence loss if you're simultaneously buying and selling the same product in the same order book. Bid/ask spreads were as wide as double or triple digits during the liquidation event. At this point the only limit to riskless profits is your liquidity and the speed at which you can execute.
  - iakh 5y ago
    Market makers can't just conjure up a counterparty. The spread widens because lack a liquidity increases directional risk. If it were truly riskless, others would jump in to close the spread.
    - wickoff 5y ago
      This is exactly what market makers do. During violent moves prices vary wildly between exchanges and market makers become takers. They instantly buy and sell the same amounts on different exchanges without risk, then re-balance their accounts.
      - evrydayhustling 5y ago
        No. To buy and sell on your own schedule, you must take liquidity, paying the spread. This is not what market makers do, they sell liquidity. And it becomes less possible to find efficient clearing trades during rapid moves, because the spread widens exactly because market makers don't want others to arbitrage them.
  - gruez 5y ago
    > you're simultaneously buying and selling the same product in the same order book. How's that even possible? Suppose the bid is at $95, the ask is at $105, and you receive a buy order for $100, you can't "simultaneously" sell that and make a profit.
  - stepanhruda 5y ago
    You can do arbitrage without being a market maker and you can be a market maker without doing arbitrage.
kikimora 5y ago
Market makers likely know how thing works and react to a volatility spike. I wonder what was the traded volume during the event and how it matches normal order book disposition. If only retail orders left in the boom then there will be very low volume. If there were pre-planted orders from the manipulator then all liquidations should hit them. Remaining orders should disappear shortly after the event, as soon as liquidations complete.
evrydayhustling 5y ago
> Algorithmic traders, yes. Market makers absolutely not. MMs want to be there as much as possible in liquidation cascades, because bid/ask spreads are huge. MMs effectly print riskless money in these situations (which is why you see Alameda and DRW issue so much USDT in these events). This is a deep misunderstanding of market structure. Market makers are algorithmic traders. In times of price stability they benefit from capturing the spread many times. But it is not risk free: they lose money when offering liquidity to "informed trades" (those followed by a price move) because they are holding the wrong position through the price move. For that reason, every market makers implements fail-safes that stop offering liquidity when they are not confident they will be able to clear their position at a break even price.

huac 5y ago

I agree with most of this and would add that the article could be organic and the subsequent trading opportunistic (rather than being tied together). I am somewhat surprised that bots react to news within 5ms -- your execution delay on crypto exchanges is quite a bit longer than that, since the exchanges are generally hosted in public clouds and you are subject to network latency to get in there. Even within the same cloud and even within the same k8s cluster, you should expect 2-4ms for an inter-pod hop. That being said, maybe there is a hedge fund literally in the same k8s cluster as FTX...

mthoms 5y ago
Well, Alameda and FTX were founded by the same person. Up until very recently Sam was CEO of both companies at once. https://en.wikipedia.org/wiki/Sam_Bankman-Fried https://en.wikipedia.org/wiki/Sam_Bankman-Fried Edit: I suspect Alameda uses data from FTX but might not trade there: as per the article, the manipulation needs thin order books to work.