Conversation

1) Ratelimits, part 2: FTX's data

Quote Tweet

1) How I learned to stop worrying and love rate limits

Show this thread

4

5

65

2) NOT INVESTMENT ADVICE. NOT TRADING ADVICE. (Sorry for the deleted thread -- had a key typo at the start...)

1

5

3) People constantly want higher ratelimits. But not all orders are the same. The core metric here: $ volume per order sent. In other words, how efficient are your orders?

1

3

4) What do FTX's users look like? Well, here is $ traded / orders sent for the top 20 order senders on FTX.

2

7

5) So, one note is that some takers have fewer orders per trade, which makes sense. But even when selecting for liquidity providers, there's a huge range. Some customers are 75x more efficient than others!

1

5

6) What's happening? Well, this, basically:

Quote Tweet

5) Anyway... So what about the other mistake? Well, let's say you want to provide liquidity. A common strategy is: Send a bid 0.20% below the best bid, and an offer 0.20% above the best offer.

Show this thread

1

4

7) And, in case you're wondering who reaches out to us asking for higher ratelimits: it's not the people around $100.

1

10

8) So anyway, we are working on increasing throughput and decreasing latency, which is long overdue. Also, check yo orders.

4:01 PM · Apr 8, 2021Twitter Web App

Replying to

and

Sam Sir, CME uses a messaging efficiency program (MEP) that punishes overzealous order activity without commensurate trading volume, this will help the longer term health of FTX infrastructure rather than spinning up more and more compute power. penalty could be less harsh tho

1

Replying to

and

Tbh FTX is head of basically every other exchange regarding rate limits. I try to keep the amounts of req as low as possible. I think every dev should be respectful of 3rd party APIs. It's hard to implement gainzybot on other exchanges due to their way stricter rules.

1

12

Replying to

<3 actually a bit (pleasantly) surprised to hear that!

1

9

Show replies

Joey Wong王祖兒

Replying to

and

How come so many 1 share trades (few dollars) from takers?

1

1

Replying to

dumb bots

6

LeslieisHODLing

@LeslieisHODLing

Replying to

and

Yes, we evaluated traders using the same metric back at an exchange I used to work for. You learn a lot from it. Sometimes the distributions correlate with geography. But yes, the guys who complain usually have much lower volumes per order.

3

Replying to

and

Hey Sam, would you consider adding feature like saving all PNL history for all pairs on basis of trades of subaccounts even trades are closed? is it useless data load for site?

1