Taleb | Fooled by Randomness

• Results from luck are more prone to be taken away by luck; results achieved with less luck are more resistant to randomness.
• Skewness issue: it does not matter how frequent a person succeeds if failure is too costly to bear.
• Retired Dentist Checking Portfolio Returns
• Difference between noise and meaning: all observations are at best a combination of return (meaning) and variance (noise); and on a short time increment, it is entirely variance.
• News is full of noise, where history is largely stripped of it; however, it does not mean that histories are good indicators of the future. Length and breadth of history considered matters.
• We do not react to positive and negative noise equally. Even if the distribution of noise is symmetrical, the emotional torture from short term negative noise is felt far greater than their positive counterparts. Paying attention to short term noise, on balance, lead to emotional deficit. Hence, stay away from noise.
• Importance to understand the mean and variance of any field: dentists and pianists face much less career randomness; the jungle where traders live, on the other hand, is ruled by randomness.
• Cross sectional problem or survivial of the least fit: at a given time in the market, the most profitable traders are likely ones best fit to the latest cycle (eg “dip buyers” of EM and HY bonds 1992 to 1998); if the latest cycle dynamics are random, these traders are also by definition most prone to a randomness reversal or regime change
• Traits of bad trader: married to position; switching narrative (become “investor” just to hold on to losing bets); no game plan for losing times as such periods are deemed too improbable
• Bull & Bear Zoology: bullish and bearish only suggests view on direction, but not magnitude... I can be bullish on a stock yet shorting it, because of asymmetry of payoffs. What informs a trading decision is a prob distribution table. TV commentators use these terms because they are rewarded only by the frequency of their correctness but not magnitude
• Separating risk taking and risk management decisions where risk taking is informed by past data but risk management is not ie prob of some events occuring are not derived from past data.
• Randomness does not mean patternless and most often it does not look random; purposefully creating randomness is a paradox and it never works
• Humans have evolved as a separate species for 130,000 year, majority of which is spent in the African savannah where field of probabilities is narrow – information transmission is limited by physical distance; limited number of people we meet over a lifetime etc. Later on, even when probabilitic fiend widened over history, religious suppression of any thought contrary to determinism has delayed the development of probability studies. Hence, our brains have never acquired the proper probabilistic depth to deal with the complexity of modern world. E.g. when only one of two outcomes will occur, we can not imagine an outcome that is the linear combination of the two – a vacation spend 50% Bahamas and 50% Paris, or a cancer patient 28% death and 72% survival. To the patient, he only sees himself dying or survive. This inability causes us to make irational choices.

Lewis | Flash Boys

• Flash Boys is a book about the different insidious ways financial intermediaries (brokers, banks, HFTs), which by definition the conduit supplying societal capital to productive enterprises, extract ilegitimate profits from investors order information.
• Electronic Trading: bank department in charge of programming and selling algorithms that execute buyside trades when they submit orders.
• Dark Pool: private exchanges ran by large brokerage banks where large buy and sell orders are matched first internally before broadcasted on public exchanges. Conflict of interest: banks prefer executing first in their dark pools rather than seeking out best price in the market. Banks can also sell access to their dark pools, and thus information of large orders, to HFTs.
• Payment for order flow (PFOF). Exchanges (NYSE, BATS) and market makers (Citadel Securities, Virtu Financial) pay brokers to send over their client orders. This revenue source has allowed brokers to slash commissions to zero, benefiting investors. However, conflict of interest: brokerages, instead of seeking for "best execution" in the interest of the investor, now maximizes kickbacks when making "routing" deicisions.
• Take and make liquidity. Company A is trading at $5.0 $5.5. A market order at $5.5 is said to take liquidity or cross the line; while a limit order of $5.4 is said to add liquidity and is rewarded when met with an opposing sell order.
• Spread Networks and CBSX exchange in Chicago. Spread networks build the straight fiber line connecting Chicago and New Jersey. The day when it went online, CBSX exchange flipped its fees and kickback structure – paying brokers for taking liquidity (market orders) and charging fees for adding liquidity (limit orders). Explanation: paying is to incentivize brokers to route market orders first to Chicago (to be frontrun by HFTs using the fiber line), and charging is to extract HFT firms fees for laying limit teaser orders (100 share lots) as exchange is now allowing them to profit.
• Latency Tables. HFTs can also decipher from which specific broker is an order made by identifying patterns in latency, i.e. the time it takes for an order to travel from a broker to all exchanges; and use the info to guess e.g. whether more of the same order is to follow.
• Reg NMS of 2007 updated the broker mandate from "best execution" to "best price" defined by NBBO (National Best Bid and Offer). "Best price" mandate caused orders to be routed to more exchanges (from most favourable price to next, regardless of volume) and therefore creating more HFT arbitrage opportunities. Less discretion for brokers also made their routing more predictable.
• NBBO is calculated inside SIP (Security Information Processor) where each exchange's prices are transmitted, compiled, and disseminated. Greater the number of exchanges, longer it takes for NBBO to be calculated. HFTs game this system by creating faster processers than the SIP technology to have a more "real time" view of the market, and therefore, frontrun investors who trade based on SIP prices. HFTs profit depends on 1) time gap between SIP and private prices; 2) volatility of stock during these milliseconds (or how many times the SIP price and private prices differed). This is why HFTs is incentivized for more market volatility, and more exchanges.
• HFT Strategies
1. Electronic Frontrunning: obtaining an investor's order information at one exchange and race him to next
2. Rebate Arbitrage: seize the kickbacks exchange offered without actually providing the liquidity that the kickback is meant to entice
3. Slow Market Arbitrage: seeing price changes at one exchange, and race to act on another exchange before its price can update
• HFT profits were made possible by fiduciary failures of banks and brokers. Banks, upon receiving an order to its dark pool, willingly ignores offers from rest of the marketplace so that can charge HFTs access to bridge the gap. This creates unnecessary layers of middleman and increases transaction costs.
• The chance of offsetting orders (200 shares buy and 200 shares sell at $80) coming to the same exchange at the same time is extremely low. Therefore HFTs are said to benefit investors by providing liquidity – being the buyer on one exchange and seller on another. However, this bridge is itself absurd as why was the order not brought to the right exchange in the first place? Becuase brokers respond to rebates.
• "The original false note struck by the big Wall Street bank – the act of avoiding making trades outside of its own dark pool – became the prelude to a symphony of scalping."