> The order also finds that over a six-week period—August through September 2016—a former Coinbase employee used a manipulative or deceptive device by intentionally placing buy and sell orders in the Litecoin/Bitcoin trading pair on GDAX that matched each other as wash trades. This created the misleading appearance of liquidity and trading interest in Litecoin.
We always knew that wash trading activity was rampant in the cryptocurrency space, but I didn't actually expect Coinbase employees to be doing it within Coinbase. If employees of the most regulated exchange in the world were wash trading with some regularity, makes you wonder how bad it must be at some of the more Wild West exchanges.
And I literally worked on a Dark Pool Crossing Engine myself at the the time, and here's my take:
Having multiple algorithms running at the same time was...unremarkable. And since they're all running different algorithms and all trying to game the market, it wouldn't surprise me if they ended up trading with each other by accident, especially if they had multiple accounts to trade with. What's honestly more surprising is that they only had two running at once. And further, if they both traded through a dark pool, then there's no way they could know they were trading with each other, because that's one of the intended features of a dark pool.
So accidental seem plausible, but then again so does malice.
Now if it wasn't through a dark pool, and the algos could see each other's accounts, then naughty naughty, and they deserved every cent of the fine.
Having two separate algos directly and actively trading in the wild is highly unusual.
There are obvious reasons for which you don't want your algos to trade one against the other on the market.
First, it's very dangerous as you can end up in the situation described here where you essentially self make the market.
But more directly for simple execution costs/fees reasons.
I don't know of any single player that has concurrent algos where there is not a "netting" layer between algos and the market, so that concurrent trades are matched internally before reaching the market.
No it's not. ATD had like 50.
>There are obvious reasons for which you don't want your algos to trade one against the other on the market.
Who said they're trading against each other? Just look at GME. Swapping shares back and forth makes trading volume go up.
Every imbalance is gameable, and they have armies of people whose job is to find every gameable imbalance and make money off of it.
>First, it's very dangerous as you can end up in the situation described here where you essentially self make the market.
As we've seen when traders have bankrupted themselves.
>I don't know of any single player that has concurrent algos where there is not a "netting" layer between algos and the market, so that concurrent trades are matched internally before reaching the market.
Well I know of one that had 15+, and coinbase just got busted with two so...
They're not evil, just another tool.
I don't (and won't) work for them again, but there were non-evil reasons for a dark pool.
Anonymous trading (dark pools) also prevents favoritism and collusion too. You're trading based on nothing but the price and public information about the company. So it's less gameable.
In addition, the order finds that while Hedger and Replicator had independent purposes, in practice the programs matched orders with one another in certain trading pairs, resulting in trades between accounts owned by Coinbase."
Doesn't sound very accidental.