With G-20 upon us and trade deal uncertainties abound, and the largest short position in NG futures in 3 years, on top of a surprisingly warm short-term weather prediction boosting prospective air conditioning demand that was later tempered, NG is certainly in a position to move higher.
After breaking through the 20,2 Bollinger lower band on the highest down-day volume since early Feb (which also created a tradeable intermediate-term bottom), UNG gapped higher yesterday leaving fresh large short positions from the previous week under water.
I would expect UNG to test $20.50 in the next few days/weeks.
Tuesday, June 25, 2019
Thursday, June 20, 2019
CL to short-cover to mid 59s
A combination of global-macro factors have converged on WTI creating a scenario with good probability of further short squeeze to mid 59s, the confluence of the upper Bollinger Band (20,2), 50-SMA, and 200-SMA.
GLD/Aug GC extremely bullish
It doesn't require a highly sophisticated technical analytic system to see that gold is aggressively pursuing multiyear highs. A combination of changing perception about increased likelihood of a Chinese trade deal as well as Fed Fund futures pricing nearly 100% odds of a rate cut in July have created a "goldilocks" scenario for GC. Deep OTM calls are the only way to play.
Tuesday, June 11, 2019
Metaphysics of daytrading
CME markets facilitate trade through the Centralized Limit Order Book (CLOB). CLOB is a matrix of limit order amounts that constantly change based on idiosyncratic constraints of the participants, inside bid/ask changes, news items, and events in correlated markets.
Scenario 1: Inside Bid/Ask Changes
When the amount of posted limit orders at a price level reach 0, the next best price becomes the "inside" bid or offer. This "race" to 0 is what causes dynamic shifting of the bid and ask which translates into the best prices that market orders can transact against leading to the last quoted price up (at the offer) or down (at the bid). Generally, HFTs and market-making firms are aware of how much size they contribute to the overall posted liquidity and assume inventory risk to be able to capture spreads. But when an outlier/skew of aggressively-sized and one-sided initiative market orders consume that posted liquidity, this sudden impulse may cause large disturbances in posted limit orders at nearby and even further out levels. HFTs interpret this data as a possible change in volatility and must modify their mass quotations to account for increased risk and to make sure they do not overly contribute proportionally to posted liquidity.
Scenario 2: External Events
News items and price moves in correlated markets are externalities that cause participants to modify exposure through limit or market orders based on these events' consequences on their valuation models with optimal execution as a secondary consideration and the primaries being recency of the catalyst and potential price impact. At this point, price performance behaves as a Random-walk with drift or a Martingale with little-to-no autocorrelation, meaning previous price movements don't predict future performance.
The statistical profile of the endogenous supply/demand may not change significantly in the latter scenario though price performance may be significant.
Only in the presence of such a statistical shift viewed as a complex event and an outlier with meaningful context accompanied by or without the events that comprise Scenario 2 can a clear daytrade be taken. This is a situation where high price discovery power within the internal supply/demand of the market of interest leads to a high probability of a Partially Observable Markov Decision Process (POMDP), as opposed to the Martingale of the external event case.
In the rare case that the market behaves as a POMDP, an outlier in endogenous supply/demand is occurring with a high probability of impacting nearby participants in a similar fashion (impacting their inventory constraints) and causing a chain reaction/mini-contagion. In this situation, price performance exhibits autocorrelation with respect to recent trading. As HFTs and other participants reduce their exposure, market facilitation rises with a spike in potential price movement. Specifically, more price levels are attainable per volume unit as the race to 0 at each level occurs more quickly.
If a market is experiencing low price discovery power, it is likely trading with a Martingale or Random-walk with drift component and the endogenous supply/demand will not provide strong trade signals as autocorrelation may be low or non existent.
Finally, it would seem that AI methods especially reinforcement and meta-reinforcement learning would only apply to the sparse reward cases of POMDP. Automated and intelligent methods that identify huge imbalances leading to high price discovery power in the near- and short-term can reduce cognitive load on the human trader. Similarly, automated tools can highlight times where the market exhibits low price discovery power and a Martingale process.
Scenario 1: Inside Bid/Ask Changes
When the amount of posted limit orders at a price level reach 0, the next best price becomes the "inside" bid or offer. This "race" to 0 is what causes dynamic shifting of the bid and ask which translates into the best prices that market orders can transact against leading to the last quoted price up (at the offer) or down (at the bid). Generally, HFTs and market-making firms are aware of how much size they contribute to the overall posted liquidity and assume inventory risk to be able to capture spreads. But when an outlier/skew of aggressively-sized and one-sided initiative market orders consume that posted liquidity, this sudden impulse may cause large disturbances in posted limit orders at nearby and even further out levels. HFTs interpret this data as a possible change in volatility and must modify their mass quotations to account for increased risk and to make sure they do not overly contribute proportionally to posted liquidity.
Scenario 2: External Events
News items and price moves in correlated markets are externalities that cause participants to modify exposure through limit or market orders based on these events' consequences on their valuation models with optimal execution as a secondary consideration and the primaries being recency of the catalyst and potential price impact. At this point, price performance behaves as a Random-walk with drift or a Martingale with little-to-no autocorrelation, meaning previous price movements don't predict future performance.
The statistical profile of the endogenous supply/demand may not change significantly in the latter scenario though price performance may be significant.
Only in the presence of such a statistical shift viewed as a complex event and an outlier with meaningful context accompanied by or without the events that comprise Scenario 2 can a clear daytrade be taken. This is a situation where high price discovery power within the internal supply/demand of the market of interest leads to a high probability of a Partially Observable Markov Decision Process (POMDP), as opposed to the Martingale of the external event case.
In the rare case that the market behaves as a POMDP, an outlier in endogenous supply/demand is occurring with a high probability of impacting nearby participants in a similar fashion (impacting their inventory constraints) and causing a chain reaction/mini-contagion. In this situation, price performance exhibits autocorrelation with respect to recent trading. As HFTs and other participants reduce their exposure, market facilitation rises with a spike in potential price movement. Specifically, more price levels are attainable per volume unit as the race to 0 at each level occurs more quickly.
If a market is experiencing low price discovery power, it is likely trading with a Martingale or Random-walk with drift component and the endogenous supply/demand will not provide strong trade signals as autocorrelation may be low or non existent.
Finally, it would seem that AI methods especially reinforcement and meta-reinforcement learning would only apply to the sparse reward cases of POMDP. Automated and intelligent methods that identify huge imbalances leading to high price discovery power in the near- and short-term can reduce cognitive load on the human trader. Similarly, automated tools can highlight times where the market exhibits low price discovery power and a Martingale process.
Subscribe to:
Comments (Atom)