Questions about calculating VWAP for /CL daily settlement

[ducatista]

I know its calculated during the settlement period (14:28-14:30 EST), but how? Are they polling every single transaction that occurs btwn 2:28:00 and 2:29:59 for both price and volume, and calculating the VWAP by weighing each individual transaction as part of the overall calculation?


In looking towards calculating VWAP on my own, my initial thoughts were to calculate it as the sum of 1m close @ 229 & 1m close @ 230, divided by the sum of the 229 and 230 volumes - before realizing its probably more nuanced than that. I dont think this method would be as accurate as the method indicated in P1 on my OP (which I'm assuming they use), but would it be a sufficient approximation?


How soon does CME publish the settlement prices? Is it only made available at the end of the previous trading day/ start of the next trading day? (6pm)

 

[Drawdown Addict]

The VWAP is an intraday indicator.

 

[ducatista]

VWAP is just a calculation. I dont know how you chartists choose to use it (nor do I want to, as thats not the point of this thread), but I'm talking about it from a settlement calculation perspective

 

[Sekiyo]

Volume Weighted Average Price
From my own thinking its price * volume.

101 * 2000
100 * 1000
099 * 500

VWAP is (101*3000+100*2000+99*500)/5500 = 100.454545455

Here you have access to the settlement prices.
Only updated after market close.

 

[SunTrader]

Settlement price is what matters, not a 1 or 2 minute VWAP of same.

After all how much can it vary from settlement itself? A tick or two (maybe three) most trading days.

 

[ducatista]

Settlement price is what matters, not a 1 or 2 minute VWAP of same.

After all how much can it vary from settlement itself? A tick or two (maybe three) most trading days.

What? VWAP is settlement. It's the primary method used by the changes to calculate settlement for the instruments I track. Problem is the data is not immediately published, so if your backtests rely on historical settlement data (which they probably do, as many futures datasets claiming to be 'close' prices are actually settlement prices), then you're basing your results on some unachievable marker that can't actually be traded - and thats not even including slippage, latency, etc.

Run an experiment comparing official settlement prices to LAST, CLOSE, or any metric of your choice. Sometimes its the same, sometimes its not, and sometimes its pretty far off. The longer your holding period, the less it matters. But if your strategy is execution sensitive, it matters a great deal.

 

[ducatista]

Volume Weighted Average Price
From my own thinking its price * volume.

101 * 2000
100 * 1000
099 * 500

VWAP is (101*3000+100*2000+99*500)/5500 = 100.454545455

View attachment 317342

Here you have access to the settlement prices.
Only updated after market close.

That was my initial thought process too. But I can't help but feel they get a little more granular than summative 'end of minute' data. CME doesn't elaborate unfortunately, but my thinking is that they're calculating the weighted sum of every transaction as it occurs during the settlement period. As opposed to the closing block for 2:28, the closing block for 2:29, etc. I'll run a few tests to compare the minute block VWAPs to the published settlement prices... if they're off, then they probably use transaction level data.


And yeah, that delay is problematic. Hence why I'm looking to see if I can approximate it on my own. I already have a slippage model (or rather, a model that accounts for differences in trades modeled on settlements vs trades executed on LAST), but Im hoping to narrow that gap btwn model/execution by better modeling the settlement on my own.

 

[Sekiyo]

That was my initial thought process too. But I can't help but feel they get a little more granular than summative 'end of minute' data. CME doesn't elaborate unfortunately, but my thinking is that they're calculating the weighted sum of every transaction as it occurs during the settlement period. As opposed to the closing block for 2:28, the closing block for 2:29, etc. I'll run a few tests to compare the minute block VWAPs to the published settlement prices... if they're off, then they probably use transaction level data.


And yeah, that delay is problematic. Hence why I'm looking to see if I can approximate it on my own. I already have a slippage model (or rather, a model that accounts for differences in trades modeled on settlements vs trades executed on LAST), but Im hoping to narrow that gap btwn model/execution by better modeling the settlement on my own.

What minute data ?
… Anyway … Good luck with your research.

Have you tried to VWAP the trades happening from 2:28:00 to 2:29:59 ?

There is nothing more granular than that …
I don’t get what’s your problem.

We’re not talking candlesticks here but a simple time and sales over the last 2 minutes.