Data driven insights about trading accounts on GDAX

I’ve recorded all the trades done on GDAX for the last few days expecting to identify the whales or high frequency trading bots. However, I found none of those. The largest volume accounts were “single-shot” buying or selling. I couldn’t find any large accounts trying to manipulate the price.

For those who are curious about it, here are the data I gathered : http://docdro.id/xjOhomW

EDIT : Made a mistake in the analysis. The IDs are not users IDs but orders IDs. Will update that post with additional findings.

3 Likes

Wow. How far back can you go? I think if you want to find price manipulation, you probably need more than a few days of trades. Also, pick the days with highest volumes (2/5 - 2/6 is one that comes to mind).

I looked through a few of the highest volume traders. Even though it shows a lot of trades, according to the timestamp, the trades (for more than 1 trader) happened at the same time. My guess is GDAX list them as separate trades, but it really is just someone putting a very large single trade. I’m thinking of this scenario:

Seller A sells 0.02 BTC at $8500
Seller B sells 0.01 BTC at $8500
Seller C sells 0.1 BTC at $8502

The buyer puts in a taker trade for 1 BTC, and GDAX fills that trade from Seller A to C (and a whole bunch of others) to fill the order. Even though seller C is selling at a higher price, the buyer made a taker trade and bought at market price (best price available), which included seller C’s order. In the data you pulled, even though Buyer placed 1 trade, it looks like he placed a bunch of trades because GDAX filled it with multiple sellers, so it looked like he traded multiple times.

That’s my understanding. Love to hear your thoughts on this.

2 Likes

Only a few days back because that’s when I started to record the individual trades from GDAX. They are only accessible in “live mode”.

Yes, I see what you’re saying. After reviewing the code and how I interpreted the GDAX documentation, I might have made a mistake. I thought that “maker_order_id” was “the ID of the order’s maker” but because of your observation, it’s most likely “the ID of the maker’s order” (sorry my english isn’t great :joy:). So, if that’s true, there is no way to track users.

However, I can still get some valuable insight from the last few days. Any ideas of what information I could extract ?

1 Like

I briefly looked at the GDAX API just now. I haven’t really played with it so I’m not sure if you can pull these info for all accounts or specific to a single account. If you can pull these for all accounts, it would be valuable info:

Total daily volume for each trade pair
Total fiat transfer out of Coinbase/GDAX over X days
Total crypto transfer out of Coinbase/GDAX

1 Like

My platform is only recording the “BTCUSD” pair right now. So I can get you that info for this pair. You don’t trust the daily volume published on GDAX ?

There is no way to get that info. I was more thinking about computing statistics on the price/volume to get a sense of how the price moves at an atomic level. If we understand how the price moves better, we might come up with techniques to trade it and accumulate bitcoins from exploiting the volatility.

1 Like

That would be interesting too. Although I have to see how you implement this to understand how it will work.

1 Like

I’m doing some work on the infrastructure. I’ll give you the numbers and the part of the code that’s doing it once I have some free time.

1 Like

Sound great. Looking forward to it.

2 Likes

Here it is : https://ufile.io/hf6v0
It contains the python script to run the numbers and the data extracted from my DB.
The numbers I got are coherent with what is shown on GDAX.

2 Likes

Cool! Can’t wait to check this out.

2 Likes

You guys might be interested in this tech i use - https://bookmap.com/bookmap-ceo-explains-the-key-feature-of-the-trading-platform/

Havnt used it for Crypto yet but helps with fingerprints around algo trading :slight_smile:

1 Like

💰 YEN · YouTube ·️ YEN.CAMP 🧠