Hi, I'm a data scientist by trade, don't work in the quant industry, but nonetheless enjoy exploring market data. I have young children so it's a brief bit of escapism for me. I've recently purchased a small amount of level 2 data (for ES); circa 3.5GB per day worth of data. I like the idea of monitoring the balance between volume of bids versus the asks (10 levels each side), to see if this provides any indication about future price movements. I am not trying to anticipate the next tick but rather, perhaps the path of price for the next few minutes or so. My instinct tells me the signals I am hoping to find will be rare (if exist at all), perhaps occurring a handful of times in a month. With my limited spare time, so far I have been assigning a sliding scale of weights to each level (highest weight for bids or offers closest to the trading price), computing the weighted sum of each side, and then a rolling average of each side (eg rolling 15 second average to smooth the volatility). From there I've computed an imbalance metric that varies between +100% and -100% (-100% here would indicate ALL of the volume on the inside 20 levels is on the sell (ask) side, 0% would indicate equally balanced bids Vs asks). Then... I just eyeball the charts, imbalance versus price. Generally my imbalance metric bounces around between -20% and +20% but there are few periods where it may even exceed 40%. I tend to think the best teacher is the data itself but nevertheless I felt it might be worthwhile posting these thoughts here incase any experienced members would be generous enough to nudge me in the right direction with my explorations. Many thanks
Take your skills and point them at the implied vols of the mega cap group. Level II is listed cash market liquidity, but equity trades using modern financial engineering now. Off exchange, unlit, OTC, synthetic, etc. Index returns are insured by the vol market, and for this reason, the vol market (for mega cap) is the most important thing for immediate direction.
You mention "a rolling average of each side." I don't know if you are aware of the concept of group delay (ie lag) in averaging. Perhaps take some time to learn about digital signal processing (DSP) and learn about reducing lag. John Ehlers (Technical Papers) is the man whose work you want to look at.
These PDFs ""might"" be helpful ? Trade arrival dynamics and quote imbalance in a Limit Order Book and... The price impact of order book events Hope these are helpful... Have Stop Will Trade Edgy...
Do you create your own charted Index of these MegaVols (IV of GOOG, IV of AMZN, IV of META, etc., etc., with each equal weighted or do you CAP wt them or Price Wt them... Do you chart this versus the CBOE VIX for SPX to see if there is a lead lag EDGE you can exploit. Have Stop Will Trade Edgy
I just normalize the IVs, and equal weight 'em. The norm script I use this one (can't find a link - abbreviated) Code: # by linus on the ThinkScript Lounge #: Multiple instruments comparison ## Method used = NormalizedSymbols ## linus, 2014-03-16, v0.3 script normalize { input data = close; input Min = -10000; input Max = 10000; def ha = HighestAll( data ); def la = LowestAll( data ); plot normalize = (((Max - Min) * (data - la)) / (ha - la)) + Min; } Picture example (IVs are inverse to track move down with move UP on index). This is what the data suggests. If you're a good enough quant shop, you can replicate NDX vol with mag7 vols (vol arb?). I included the last three days of relative moves for the two. Worth looking into I guess, but I just use it because it's in a persistent lead/lag relationship with the D1 contracts.
Thanks, this is an interesting line of enquiry and seemingly takes me away from my current focus on bid ask volumes. My interpretation and prior assumptions about your post are: - the implied volatility of various megacorps is presumably underpinned by tradeable instruments including Options. - if the implied volatility of these megacorps significantly changes, it provides reliable(?) clues about short-term movement in the corresponding stock index. - presumably a sudden increase in implied vol foreshadows a decline in the corresponding index. Assuming I've followed correctly(?) I am surprised this holds true (I am not disputing it), as I would instinctively expect the stock index to "price in" the same information immediately. I wonder why /How such lead/lag relationships can persist..
Vol market maker flows are passing through and the dealers/option holders are continuously adjusting to greeks. If you learn market making (quant finance) and index basis pricing, you start to get a feel for how the different venues interact. MM flow and large spec flows (FOPs, VAR SWAP, basis extremes). Generally speaking, at least one side of an arb has to be a retail venue, lol.
You could name it... Wait for it... Real Mega Volatility Index (RMVI) ... Sound out: RemVee anyway... Right... it makes more sense to plot against its own Tech Index - NDX instead of SPX... and yes.. You could look for volArb trade setups. Where there was plenty of liquidity in both stocks and/or futures and you have a big lead or lag in RMVI to NDX and you go long or short Mega Stock stock options vs long or short QQQ options (or do the futures options, etc.) --- and Bobs Your Uncle. Nice work Real... Edge... Have Stop ~ ~ Will Trade