I am trying to understand how a ticker price is determined for an asset, in this case I am specifically looking at BTC data from GDAX. The question stemmed when downloading historical BTC data from GDAX and seeing the following (truncated):

1516099829,11921.070000000000,0.008100000000 1516099829,11921.070000000000,0.037873260000 1516099829,11921.070000000000,0.031519050000 1516099829,11921.060000000000,0.037464080000 1516099829,11920.000000000000,0.001000000000 1516099829,11912.530000000000,0.001678900000 1516099829,11909.000000000000,0.000100990000 1516099829,11908.360000000000,0.001000000000 1516099829,11908.360000000000,0.001000000000 1516099829,11900.630000000000,0.001680580000 1516099829,11900.000000000000,1.293924550000 1516099829,11900.000000000000,0.132108480000 1516099829,11895.000000000000,0.000300000000 1516099829,11891.170000000000,0.000727720000 1516099829,11890.190000000000,1.400000000000 1516099829,11890.160000000000,0.000838510000 1516099829,11890.000000000000,0.058736000000 1516099829,11888.740000000000,0.001682260000 1516099829,11885.310000000000,0.252173310000 1516099829,11882.920000000000,0.841910000000 1516099829,11876.860000000000,0.001683950000 1516099829,11876.310000000000,0.000728620000 1516099829,11871.000000000000,0.000100990000 1516099829,11871.000000000000,0.000100990000 1516099829,11869.590000000000,0.010000000000 1516099829,11868.000000000000,0.000100990000 1516099829,11865.630000000000,0.010000000000 1516099829,11865.000000000000,0.001685630000 1516099829,11864.490000000000,0.900000000000 1516099829,11864.480000000000,1.113098000000

As you can see, the price fluctuates wildly in a single second. I understand that this data is accurate and shows all of the executed orders for the second, but I do not understand how the ticker value is determined for this second.

I am going off of the (possibly incorrect) assumptions that getting the ticker value from their REST API over and over would not show these wild fluctuations for a single second. I draw this conclusion because I have ran a loop that gets the ticker data over and over with just 300 ms delay and the price has never changed in the same second. As a matter of fact, it shows the same timestamp until the price changes, in which case it shows the new timestamp. This leads me to believe that there is something behind the scenes that is determining the ticker price per second based off all of the trades happening in that second.

If it helps, I am asking this because I want to back-test GDAX data for a program that I have written. In my program, I get the ticker data for the current point in time every two seconds, and receive one value back (the ticker price). Now, when back-testing and 1516099829 comes up, I do not know which price to use. My program will behave very differently depending on which data point I use because the prices vary so heavily.

  • Welcome to the volatility problem. May 10 '18 at 20:51
  • Very simply it's the last most recent market order which took the highest/lowest buy/sell order on book. May 11 '18 at 11:59

you subscribe to the wss feed you can see most the subsecond stuff is opening and canceling limit orders which doesn't effect the ticket price. Only market orders represent the exchange of bitcoin. Market orders are liquidity takers so get a fee. They represent actual market movement. The limit orders often will never get matched. I often market orders may only come in every minute to five minutes under normal trading atmosphere. So ticker price more constant. Hope this help.

  • thanks for the reply. So am I correct in understanding that the data that I posted is just orders on the book, not necessarily matched orders that moved the ticker price? If so, it sounds like it's not possible to tell what the actual ticker price is at a time from the data that I have? I would need to find another data set that shows the actual ticker price per second if I want to back-test a program that functions on the ticker price? I got my data from api.bitcoincharts.com/v1/csv if that helps.
    – NZHammer
    May 13 '18 at 20:26
  • Market orders never go on the book, it would be open limit orders leaving book, but how can you tell if it matched or was simply canceled. Most transactions are open/closing orders. Basically algorithms jockeying for positions etc. Much fewer are exchanges of bitcoin. So really it's not as active as it may appear. Sometimes though it can speed up like in sudden market rise or dip. Usually spurred by a large market order filling. May 14 '18 at 6:48
  • You should use the gdax websocket feed. It relays the most granular information nearly as soon as it hits the market. Delay depends on your physical location and market conditions. Also for different order books from the rest API, especially lvl 3, your explicitly not supposed to periodically request. Instead start wss feed and then get book and proceed updating the book with the feed using the book and wss transaction sequence number, kind of like bleeding breaks for air tight book. May 14 '18 at 6:53
  • Ok, I've been checking price by getting the ticker value every 30 seconds and checking my order statuses every few seconds to see if any of them have filled (Both REST API). I avoided web sockets because I figured that was more than I needed for my simple program, but I will look into them more. Thanks for shedding light on what I was actually looking at with the raw data I posted. It makes sense now. I'll up vote / accept when I have the points to do so.
    – NZHammer
    May 16 '18 at 0:58
  • Yes the rest API is easy, but if you can handle authenticated rest API wss subscription should be easier. May 17 '18 at 17:07

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.