At University of Bristol, 4th year MEng
These are not detailed notes! This isnt all you need to pass
- Internet Econ and FinTech
- Lecture 3&4 Economics of the internet
- Lecture 5&6 Economics of the Internet (MORE ECON)
- Lecture 7: Cloud Computing - Ignored
- Lecture 8: Market & Economic Agents
- Lecture 9: Market and Economic Agents II
- Lecture 10: Empirical Methods (Mathsy)
- Lecture 11: Economic Agents and Market-Based Systems III
- Lecture 12: Economic Agents and Market Based Systems IV
- Lecture 13
- Lecture 14: Social Finance / The Crowd Economy
- Lecture 15: Sentiment Analysis
- Lectuer 16: Financial Technology
- Lectuer 18: Blockchain
- Lecture 19: Malware & Cybercrime
- Lecture 20: InsureTech and RegTech
Cloud is the future for tech.
Company's success has a lot of different causes, but Positive feedback and network externalities can help elevate a product a lot. Especially if you have competing superior rivals (think industry disrupting products)
The idea of success from The Long Tail: Ages ago when distribution, marketing and just getting products out was hard, companies preferred to only care about the 20% of the content (specific to media) which was where the majority of the consumers lie and use. However, there exists the long tail, that has 80% of the content, which was previously ignored as the old companies found it not cost effective to market the long tail to users.
Companies like amazon, google etc marketed this long tail and focused on it (imagine user suggestions). This is how their profits would increase so much. They'd grab the 20% user share from the peaks, then spread that over to the 80% shares for less marketable content.
They used things like customer reviews and recommendations and ratings to learn what groups of people prefer what, and pushed this recommendation down the long tails allowing other users to buy these too.
Debates about the long tail:
- Does it really exist?
- What % is it, and does that change over time?
- Is it a % or a sliding window
- The actual shape of the curve is debated as well.
Some answers: People have shown that the tails have either gotten longer or are worth more, where you have more niche content basically.
Situation: You're an old big boi company, and there's a market demand that you're not only meeting, but exceeding in terms of resources and skill. This allows you to mostly set the price and value of your services and products.
After a few years, a new company launches as a competitor in your field, maybe not the same quality as yours (yet) but seems more lightweight and a bit cheaper.
Overtime this company can grow while you grow, but it will be seen as the underdog and cheaper solution. At some point this will then grow enough to meet market demand and start taking away at your customers / bottom line.
Two types of innovation:
- Sustaining innovations: Incremental improvements on existing products and services. It's attractive to existing customers and businesses, and eventually you will offer more than a customer wants (is that bad or good?)
- Disruptive innovations: Perform less well than existing products, with lower quality and possibly less sophistication. However it's going to be simpler, cheaper and possibly more user-friendly.
Dilemma with innovation: Will just sticking to your roots bring you down? Sometimes it's worth trying new things or thinking differently before someone else does it!
Old internet companies were full pipelines, that started with something and gave you an end result.
However new tech companies prefer to be platforms rather than services. For example Uber and AirBnB are platform businesses and create an online venue for matching demand and supply to workers. This allows them to get money from both sides if they can.
An advantage of platform businesses is that they remove the idea that gigs in a gig market are hard to get. They opened up the market by:
- Creating low barrier to entry
- Aggregated supply together, providing an easy place to find demand
- Large number of customers join the platform makes the consumers/suppliers very glad and more interested in joining.
This allows the workers who worked on commision, worked on "gigs", to be able to find their gigs a lot easier and make finding work more simple rather than spending time looking around.
Gig Market: Short term, single contracts (cars, concerts, photographers etc) rather than fulltime job focused. how uber/airbnb helped: Helped people find their demand and helped demanders find their supply and took a little bit from both along the way.
Recommender Systems: Basically different examples of algorithms that recommend you data to pull you into the long tail.
- Raw Data Based: People's likes, ratings, followings and user surveys lead to better recommendatiosn
- Passive / Implicit based: People's behaviour, their viewing history and viewing times, having lists of user interactions and grouping "types" of users together.
Macroeconomics: Large-scale or general economic factors like interest rates and national productivity
Microeconomics: Behaviour of individuals and firms in making decisions aregarding allocation of resources, and the interactions among these individuals and firms.
You can base this on
- Consumer's desire and willingness to pay a price for the service.
- market demand is the total of what everyone in the market wants.
The value of an object changes per person, for example buying a falafel when you're hungry vs buying one when youre full. the worth to you changes.
We can plot the market demand as a function of quantity vs quality.
If you drew a graph of Quantity(x) vs prices(y), the supply would increase linearly in a straight line at a 1-1 trend. More supply would be available at higher prices.
The slope of demand on a Quantity(x) vs Price(y) scale would be a line sloping linerarly downwards, highest prices at the lowest quantity and lowest price at the highest quantity Imagine this as the opposite direction of the Supply line.
The point where demand and supply cross is the market equilibrium. This is the point when the quantity demand equals quantity supplied Q'. The corresponding Price is P'
In an ideal market, we'd expect Q' products to be sold at Price P'.
In the cases of excess supply:
- If there is a product surplus
- Goods will be supplied but not purchased
- This leads to the prices falling.
- Until they reach an equilibrium price, P' where the quantity supplied = quantity demanded.
In the cases of excess demand:
- There's a product shortage
- The goods will be demanded but will not be supplied
- prices will rise
- Eventually the prices will settle at a point P' where the quantity supplied = quantity demanded.
A market shock may occur:
- Would lead to a shift in demand (which affects the equilibrium price and quantity available in the market)
- There's a negative shift in the demand, i.e the average price paid for some X of product will fall such that if the market would pay £20 per product at a quantity measurement of "2" in the market, they will now pay £10 at the same quantity of measurement.
- This usually occurs during a technological disruption where the demand for the service/product could fall drastically.
New Definitions:
Combinatorial invention: A technology whose rich set of components can be combined and recombined to create new products / services
The social inertia to new ways of doing things are overcome (people get used to doing new stuff), this affects the demand.
Innovators and manufacturers play with a set of components and come up with new ways of using them, this means the supply is affected.
An example of a key factor of the "combinatorial innovation" is open source software. This allows for a lot of innovation in new ways.
- Digital goods: Are delivered online, therefore cheap (mp3s)
- Information Goods: Value of the goods is in the information it provides (google, news websites etc)
- Online purchasing of physical goods: This moves the business to business supply chain (amazon)
- Online provision of services: Affects the impact of demand and the consumer's internal pricing of theproduct.(eg spotify)
For digital technology these are the key qualities
- Fixed upfront costs, which essentially is high costs for development
- Cheap to distribute and cheaper variable costs (any additional cost per sale is variable cost)
- Most of the production costs are sunk, i.e cannot get manhours back unlike factory which can be sold.
- No capacity constraints, which would limit the number of times something can be reproduced.
Non monetary differentiaters between these :
- Digital goods are often experience goods, which means that a customer doens't know the value without actually using them. This means their internal pricing could be higher or lower. But also means they cannot estimate your costs on it that easily
- Positive network externalities such as the network effect are often very strong. This values the product higher depending on the impact of the positive network externality.
- Search costs for a consumer are low, specifically customer aquisition costs are low.
If two companies are competing in terms of digital cots, then their variable costs are near zero. This competition will drive the prices down to zero.
This however means that new companies might not be open to joining as the setup cost could be too high and they wouldn't be able to recover anything spent.
There's usually a tendency to gain a monopoly and or zero pricing.
Zero Pricing: The demand for a good/service/commodity is significantly greater at exactly 0, than compared to any price even slightly higher than 0.
When the dotcom boom was happening, people predicted that the costs of small businesses selling products and services to consumers would be very low as they wouldn't require intermediaries.
However, as the costs of reaching customers went lower, the cost of getting their attention (converting views to clicks?) increased. This meant that a new kind of intermediary company emerged, for example Ebay and Amazon.
Price discrimination occurs when a seller finds a way to charge different buyers different prices, depending on their willingness to pay.
for example:
- 1st degree pricing: Personalise the prices per person, which is hard in the physical world but easier online
- 2nd Degree pricing: Offer the same product at different versions at different price points
- 3rd Degree pricing: Offering different prices to different groups, for example Students.
1st degree is per person 2nd degree is focused on options 3rd degree is focused on the type of person they are.
The idea is that a seller would love to charge the buyer the price that gets closest to what they're willing to pay for a product. However the seller doesn't actually know the internal pricing of the buyer, therefore cannot price the object as such.
However, they should be able to learn from previous transactions and histories and provide an estimate pricing to ensure that it works.
Ex: Amazon, Orbitz that priced you more based on if you were accesing it through a Mac. This is not technically illegal.
Basically provide the option for the user / customer to pick what kind of services they want. Offer these services at different prices by reducing features or adding limitations in the product.
For example, you can have games without items or you can have Spotify with no ads.
This is great as it allows a customer to build a true valuation of your product. For example they could start with a free version, then pay $1 for a small feature and could then get hooked and pay a subscription fee of £5. This allows each user to externalise their internal pricing, but through the options available.
However, note that if there are multiple options, sometimes people willing to pay a higher price would end up taking the lower options if it does exactly what they wanted to do.
- Sell the product at different price points to different groups of users
- Example: Student Discounts, Rail Cards Etc
Personalisation is the much fancier version of 1st degree pricing where the customers decide their own versions.
They pay based on the features they accumulate, and they allow the customers to create exactly what they want. People tend to choose the higher tier options when offered with different options in general leading to an economic benefit for the company.
Eg: Dell comptuers / build your own pc sites
Bundling is selling a number of services and products together for one price
Like versioning it's used to sell more to customers who otherwise woulnd't have purchased.
Adding products for a company to a bundle doesn't change their internal costing, as the variable costs are minute, however to the customer who has gained more products, they would be okay paying a higher price than they wanted to, due to the added features.
Any competitor aiming to join the market and compete would have to beat the bundle prices, or find a reason to attract customers away from the bundles.
If two parties trade, it has an effect on the others who had no say in the trade. This is known as an externality.
- Negative externalities impose costs
- Positive ones impose benefit.
A network externality is when buying a product or services confers indirect cost on all those who have already bought the product. This can be positive or negative
- Positive: More recharge station as electric cars increase
- Negative: Less space on roads making it worse for other car owners.
In technology: Network effect is a lot more useful. Ex: - MS office, Forums, Social networks, Paypal accounts etc.
As the network increases, the demand for the network increases until a peak. At this point the demand starts reducing due to the saturation of the market or some other negative impacts (think facebook if you paid). The peak is when the users are willing to pay the highest, and this is usually around the midpoint of the supply/growth/population of the network.
People who join in the network early wouldn't have a high internal pricing of the product, requiring the companies to have a lower valuation. As the network and demand increases, that's when the internal pricing for the nproduct increases and people are willing to pay more.
A good way to build a quick network effect is the idea of using free tools and free versions. Once this network is at it's peak, you can introduce other charges or versions that could enable users to stay and capitalise on the network size. (Tinder premium // PayPal for businesses etc).
Switching between different products and services can have a switching cost. This is things both in monetary and non monetary form, for example controls on a DVD player // buying an adapter.
Switching between things in digital world are a lot more common, and usually costs a lot (for example switching from Windows to Linux).
When switching costs are very high, buyers are said to be locked in to one provider.
Some sources of the switching costs are:
- Training how to use the system
- Network effect (leaving snapchat but all your friends are there.)
- Setup Costs (Reset your amazon cards to use paypal isntead)
- Reduced service or loss of information (won't get spotify / youtube recommendations on amazon video/music)
How do you enforce a lock-in?:
Proprietary formats are a great way of enforcing monopoly over a network of users. For example the .doc format!
However, this means that a competitor just needs to break in the key feature and can break in the lockin effect, For example open office with the reverse engineering of .docx
Adopting an industry-wide standard allows a user network to be shared amongst providers.
- Standards allow people to share networks
- Usually leads to a bigger market but smaller market share, (more customers, but less of the monopoly on them).
- Standards Leader: One player, often the major ususally stards the standard by exposing their proprietary format. (ex Adobe and pdf)
- Standards War: When two competitors compete to determine which standard is to be adopted.
- Standards negotiation: People negotiate a standard collectively
- Paywalls (blocks users, but increases overall profits)
- Capitalising on brand, making people want to pay.
- Sell multiple and complementary goods. (ex: people selling album CDs with tshirts on tour)
Old school marketing on the internet Measured by :
- Cost per impression : Displaying
- costs per click: interaction
- costs per lead: selling details and hints to target customers.
Online marketing with auction market mechanisms:
- Providers specify where they want their ads and some budget restrictions
- The advert placement is based on page content, demographics of vistors and time of day etc.
- Advertisers bid for ads (in a robotic auction)
- Earned media: Anectodes on a site (ex reddit)
- Paid : Advertising
They provide a mechanism for buyers and sellers to meet and trade
Some Types:
- English Auction: Open Ascending Price auction
- Dutch Auction: Open descending price auction
- Sealed bid Auction: Bidders enter a bid to the auctioneer, and the highest bid wins. In the first price sealed, winners pay the price they bid, in the second price sealed (vickery) winner pays the second highest price.
- Continous double auction (CDA): Buyers and Sellers enter a bid/offer at any time. Commonly usesd in fin markets.
- A first price sealed bid auction is not incentive compatible : Bidder bids less tha nthey value the item at.
- A sealed bid auction is incentive compatible, where bidders can bid their true valuation knowing it wont affect the price they pay.
- Ebay uses the english auction system.
- Ebay ended up using the vickery (second sealed) auction system with all the sniping.
- Suppliers bid to meet a contract
- Suppliers need to be qualified to participate
- Largest number of auctions
- Google assigns them a price based on Bid price X quality
- Quality is dependent on: historic click rate, relevancy, and landing page quality and load speed of the ad.
Quality of the ad: Aims to score the probability of a click through for the ad, and ensure that those who click aren't dissapointed in the ad as that would mean they wont click other ads in the future.
- Direct impact: Environmental impact
- Indirect: Changes in behaviour that technology enables (home shopping)
- Systemic: Changes in society, people take remote jobs
- Social Impacts: Social media , privacy etc.
But how do big companies with less employees impact small businesses?
- Cost has decreased to start a business
- Businesses can reach customers faster
- Niche businesses can make more profit through online portals and payments
- Gig work (uber, deliveroo etc) can lead to more jobs.
- Tax avoidance isn't illegal
- People do fancy weird stuff to save money being spent on tax.
- Tax Evasion: Not paying tax
- Tax Avoidance: Finding loopholes and reasons to reduce tax payments
- Low Value Consignment Relief: No VAT on goods under £18 sold fomr EU to Channel Islands.
Play.com, Amazon and Tesco just sold their products under £18 via the channel islands! That however was shut down :(
They did this by round tripping their DVDs to channel islands then mailing it back individually.
- Tax loophole that allowed a lot of companies to avoid tax.
- It's closed to new companies, and will close to old companies in 2020.
The steps are as follows:
- US company licenses it's IP to an Irish / Bermudan holding company (with no employees, and has tax residents in bermuda with management and control in bermuda)
- Sublicenses this IP to a dutch holding company (no profit here, so no taxes!)
- This sublicensies it to an irish operational company (which has no profits either!)
- This irish company gets business revenue from Non-US companies, and pays royalties to the dutch company which doesn't actually pay tax by EU regulations
- The dutch company pays royalties to the other (parent, step 2 ) irish company and doesn't pay tax by EU regulations
- The irish/bermudan company just pays royalties to the american company that gave the original IP.
Better explanation of the double irish dutch
Market economics is a new way of looking at how to share / allocate limited or scarce resources. There are two aspects
- So called Market Based Control (MBC) or Market Based Resource Allocation (BMRA).
- Provides fast, robust and distributed control for varying problemss.
It requires software versions of traders and marketplaces.
Dynamics from microeconomics can help build automated dynamic resource allocation and control in engineered systems.
- Demand > Supply: Sellers' market / Shortage
- Supply > Demand: Buyers' market / Surplus.
- Price is a function of the quantity supplied or demanded.
- Price that buyers are prepared to pay at each possible quantity is a demand, and is plotted on a demand curve.
- The price that a seller is prepared to sell for each possible quantity is a supply curve. (as price increases, quantity supplied increases)
Free Markets are self correcting: If the prices and transactions are taking place at prices away from equilibrium, the competition amongst buyers and sellers will move prices back towards equilibrium.
At equilibrium, traders won't change their prices as it's good for both of them.
This kind of market can give effective allocation of resources without a central controller or external intervention into the area.
A common ideal of efficient allocation is the notion of Pareto Efficiency (PE):
- Allocation is PE if no-one can be made better-off without someone else being worse-off
- The PE allocations can arise from free markets even if they have characters just dealing in their self interest
- Free markets are not guaranteed to reach optimal allocations. (imagine a seller having a monopoly)
Traders in CDA deal with a lot of data from multiple sources and must act in real time to achieve their maximum utility.
There's the idea that a smooth supply and demand curve gives a theoretical equilibrium price. (imagine two lines \ and / meeting and making a tilted X).
However in reality:
- the CDA auction leads to a stepped supply and demand "curve".
- Each step in the supply curve represents an additional unit available at the indicated price
- Each step in the demand curve represents an additional unit desired as the indicated price.
- Buyers and Sellers will lie in real life meaning that the percieved supply/demand may be different to the actual ones.
- Dynamic variation in both curve of the Supply and Demand. This happens if a buyer and trader in a small market leave the market
- This happens after every transaction.
- Even without transactions, prices change
- Fake buyers and sellers by giving them some cash and limits on how low they can sell or how high they can buy.
- Giving different limits to people meant that different types of supply/demand curves could be introduced into the market.
- Vernon Smith won a nobel prize for his lab style studies of human market-trading.
- Smith's alpha : Root mean square deviation of transaction prices around the theoretical equilibrium price, expressed as a percentage.
- Allocative Efficiency: Total utility earned by all traders, ... measures how effective the market is at extracting 'gains through trade'
- Single Agent Efficiency: measures how well an individual agent performs!
The allocative efficiency of a market cannot be over 100%, but one of a single agent can!
Gode and Sunder set a series of experiements with humans and software to see whethe the allocative efficiency came from the traders or from the market.
- ZI- U (Zero Intelligence) Unconstrained traders quote random prices and an overall maximum price to trade. They didn't reach close to equilibrium and the market were not effective
- ZI - C Constrained traders, quoted random prices but were told to not make a loss producing deal, and that meant they had similar results to those of humans
this proved that the intelligence of allocative efficiency came mostly from markets, but not from the humans specifically. And that allocative efficiency isn't the best measure of intellegence for traders in a market!
So, are ZI-C traders all you need to make a useful market?
Nope, Dave Cliff the man had these findings in a report:
- Analysed the PDFs (Prob Density Funcs) of ZI traders and proved that Gode&Sunder's results were artefactual
- Predicted conditions when ZI-C traders would fail to equilibrate
- Implemented the ZI-trader system to empirically demonstrate failures.
He then went on and just invented ZI-Plus traders or ZIP traders!
- They aim to have a profit margin
- They demonstrated to succeed in markets that ZICs failed
- And they had a human like behaviour in a CDA / Retail auction
- ZIC sellers generate offer prices at random, in the range from min selling limit to the max price in the system.
- ZIC buyers generate an offer price at random, from the max selleerr limit down to min buying price in the system.
- The intersection in the area under the slope of these two curves gives us the expected transaction price estimates. The point where the two functions meet gives us an equilibrium price.
The above works well in a situation where the transaction-price prob density func is symmetric. In a world where this is not true, you can see a place where the expected transaction price would differ significantly from the equilibrium price.
this means you'd definitely need another algorithm: Zero Intelligence Plus (ZIP)
- For sellers, you have a limit price L that's the lowest you can sell for.
- The asking price P is L plus some profit margin M.
- If you have something to sell, and your current price is P and if sellers are accepting below P or are making offers below P, then decrease M (but not below 0).
- If trades are happening at prices above P, then increase M.
Buyers run the inverse of this, and the amount you update the profit margin M is determined by a learning rule which makes the ZIP behaviour adaptive!
- It's adaptive, i.e. sets the margin of profits up and down base on simple ML rules!
- Quote price
Pis set by the limit priceLand the marginM
P = L ( 1 + M)
The amount by which the margin is raised or lowered is determined by
- Target price (stochastic)
- learning rule ( adjust towards the profit margin )
For a target price, the ZIP algorithm aims for a number slightly below the last quoted if it's trying to reduce, or aims slightly above if you're trying to raise.
ZIP was actually better than humans, same with GD.
- Outperformed all competition in it's contest
- Do nothing until:
- The bid offer spread drops to a sufficiently small value
- Offer is less than the smallest transaction price
- there's not much time until market closes
If any of these are valid, then jump in and "steal the deal" so long as the sniper makes a profit greater than the minimum threshold.
However, it's not the best when:
- there's only snipers (since they do nothing)
- Unable to see the market's equilibrium price P_0, and therefore will snipe any deal.
- Sophisticated algorithm, uses belief function from recent market activity.
- Calculates the belief function from a history H of n trades.
- Calculates for each bid or offer, that a bid or offer would be accepte at that price
- GD interpolates to smooth the belief function for prices that don't exist in the history
- chose a quote price that maximises a trader's gain. It's calculated by a product of utility gain (expected profit) and the probability of acceptance.
- A MGD is a modified version of GD that works for limit order books!
- Uses real time dynamic programming to formulate agent bidding strategies.
- States are represented primarily by an agent's holdings and transition probabilities are estimated from the market event history.
- The above estimation is done similar to GD hence the name
- Uses past history to estimate the current Equilibrium Price (P_0) and estimates the volatility.
- Has an aggressiveness value, and its taken into account for the calculation of a bid/offer. More aggrisvely placed ones are more likely to be accepted.
- Aggression function: Updated based on price volatility in the market. In a more volatile market, a small change in aggressiveness will lead to a greater change in bidding behaviour.
Testing these different systems is hard analytically, so everyone just runs experiments
- In MBRA - Market Based Resource Allocation models, you can assume that people are using the same trading algorithm.
- In Financial Markets, that's implausible / not really happening.
So, someone actually tested it on a CDA type:
- Homogenous population tests: Everyone runs the same algorithm. This provides a good baseline for performance.
- One-in-many: Explore vulnerability to defection / invasion in the trading algorithms. Such as imagine in a MBRA model, one of the bots swapped algorithms.
- Balanced Group Tests: Buyers and sellers are evenly split between the two types of algorithms, and each has a counterpart with the other type's identical limit prices.
Some results of the algroithms being tested:
- Kaplan is only good if there are small number of trades amongst a larger population as it relies on equilibrium in the market
- ZIC lost to both ZIP and MGD
- It's hard to tell a winner directly.
- AA beat most traders (including GDX)
Here's a really good link for the same notes by the amazing Aleena Baig!
- Understand what's normalisation of deviance
Market Terminology:
- Go long: Buy X, expecting the price to rise (bullish)
- Go short: Sell X, expecting the price to fall (bearish)
Short-Selling: (How to make money as the price is falling)
- Borrow n units from a lender
- Sell n units at sometime from now at price
P1( to a different buyer) - Price of XYZ falls to
P2. - At time
t2, buy n again atp2and return to lender. (at a pricep2<p1) - Profit =
P1 - P2 - LenderFee - Costs
- Contract derives its value from other assets.
- Those assets are known as underlyings.
- Future contracts:
- WILL be executed on a certain date aka a delivery date.
- At a forward price
- Buyer of this is bullish (long) and the seller is (bearish)
- Buyer thinks prices will rise, seller thinks they will fall.
- (Forward price > current Price) < priceAtDeliveryDate (if buyer wins)
- Option Contracts:
- MAY be executed on/by the expiry date.
- The contract specifies at strike price, or the exercise price
- Options to sell are puts, options to buy are calls
- Both of the above are standardized, and exchange-traded.
- Standardized: Size of contract, delivery/expiry date are prespecified
- Traded in a secondary market, or an exchange like stocks and shares.
- Right to buy or sell N shares.
- Price of option depends on the underlying risk premium
- Strike Price: Price at which you can buy / sell
- Expiry: Last date on which you can exercise the options
- American-Style: exercise can happen any date up to expiry
- European-Style: exercise can happen only on the expiry date.
- Option is written by the seller, and held by the buyer.
- Option prices are determined by
- Intrinsice Value: what would you get if it was exercised now?
- Volatility: Premium that depends on the underlying's price volatility
- Time Volatility: Potential risk-free return on money saved.
| Calls | Puts |
|---|---|
| Right to buy N units of underlying | Right to sell N units of underlying |
| Option-holder can buy at strike-price | Optionholder sells at price |
| "in-the-money" if the underlying price > strike price | if the underlying price < strike price |
| "out-of-the-money" underlying < strike | underlying > strike |
| "at the money" if the underlying = strike | underlying = strike |
Calls like the underlying price being greater than the strike price (they buy for cheaper than market)
Puts like the underlying price being lesser than the strike price (they sell higher than the market )
note: definitely NOT sure if this is correct
- Long Call: "I think that the price is going to go up, so I'm BUYING the option to BUY more stock at some price K" (I will buy at a lower price than the future)
- Long Put: "I think the price is going to go down, so I'm BUYING the option to SELL you stock at some price K" (Will sell at a higher price than the future)
- Short Call: "I think the price is going up, so I'm BUYING the option to SELL stock at some price K" (I will buy at a lower price than the future)
- Short Put: "I think the price is going down, so I'm SELLING the option to BUY me stock at some price K"
- Electronic marketplace for gambling.
- MUD - Multi User Dungeon
- Twinking - Experienced player helping a newbie.
- Ninja - Stealing / looting from a monster
- Shard - One or more servers that make up a single shared game world.
- Bind on Equip: Player has used BoE it can't be used or sold by another player
- Bind on Pickup: Once a player has picked up a BoP item, it can't be used by another player.
- Multi Unique User Logins - MUUL - Monthly active players / subscribers
- PCU - Peak Concurrent Users: Highest number of players online at the same time.
For a proper economy to exist in the virtual world, there's gotta be a balance struck between the currency sources and sinks.
If any imbalances exist, this is MUDflation.
High skill players can create their own metagame currencies, such as Dragon Kill Points?
Virual crime is totally a thing, PvP as an example.
- first example was seen in Everquest, which had an exchange rate going between EQ gold and US$
- EQ's GDP was actually more than the game.
- People started producing this at scale, so there would be genuine bot farms and human farms that had people playing the game just to get cash.
- September 13, 2005: Virtual plague in WoW
- Boss would cast a hit point draining spell "Corrupted Blood"
- The spell spread accross the world.
- Pandemic ensued that killed lower-level characters, and was carried by higher-levels.
- Studied by epidemiologists for real world reactions.
- Anti-terrorism officials took notice of the event, noting the implication of players to join and spread the epidemic.
- Mechanisms in gameplay that drive players to keep playing the games due to the addictive nature of the games. Some examples: Visible goals, rankings etc.
- This has been adopted by a lot of different companies, for example Location Based- Foursquare, MyTown.
- Fitness: WiiFit, Nike+, FitBit.
- Education: DuoLingo, KhanAcademy.
Web 1.0: Few content creates, personal web pages are common. Often static pages.
Web 2.0: User contributions, user generated content (youtube, wikipedia), SaaS started becoming a thing. Dynamic programming languages became a thing. Social networks etc.
- Trading units that pay based on real world events (such as an election):
- Winner Take All Systems
- % payment, based on the quantitative outcome like shares or votes.
- Market is run like a countinous double auction, where participants state the prices they're buying and selling at.
- Trade price is considered a measure of probability of a given event by collective of traders.
- Money speaks louder than opinion polls.
Essentially it's just people placing "bets" on the chances of something happening in the future, and the uncertainty of it will help predict the cost / place the cost of the bets.
- Opened in 1988 @ university of iowa
- Small-scale real-money futures market, where contract payoffs depend on economic/political events
- Non-for-profit
- Stakes are small, with limits per trader
- Low numbers of traders.
- It's not considered gambling, though most Electronic markets are.
- Only place in the US that allows you to use real money on markets
- You can use fake money / points in other prediction markets.
- Binary answers (mutually exlusive)
- Create two futures contracts, C_a, C_b
- Each share of C_a pays out £1 if A wins, same for C_b.
- Contracts initially offered by allowing any trader to buy one C_a and C_b for £1.
- You can only sell shares that you own.
- You can only buy shares with money you have.
- CDA is used to trade securities.
- Enlist a crowd of humans to help solve a problem defined by the system's owners.
- Can build some artifact together, like wikipedia articles.
- Can execute one task (amazon turk)
- Can contribute to some evaluation: Reviewing and voting at amazon.
- Can be implicit: Ad Behaviour!
You need to recruit and retain contributers:
- Pay them ££ -> Mechanical Turk
- Crowd Source as a side effect, like using reCAPTCHA
- Recruit Volunteers!
Twitter sentiment is good because it is relatively live and can show the group / crowd sentiment / opinions / views faster than predictive markets. It can be automated from user content.
Twitter Monitoring of Flue Pandemic:
- Monitered twitter for "flu marker" words, such as Temprature, throat, etc.
- Weight of the tweets depended on how many stems there were in it.
- True positive results, but there are also false positives.
Disaster Relief:
- Helped people figure out which parts of the city and world needed support during a distaster.
-
Document Level Sentiment Analysis: For a given document, identify it's overall attitude towards an object for discussion
-
Sentence or Phrase Level: For a given sentence, identify if it is +ve or -ve.
-
Aspect Level: Identify all opinions expressed with regard to any aspect of any object.
- +ve or -ve
- Supervised learning using bag-of-words.
- Outperformed human generated feature set of opinion words
- Unsupervised learning:
- Uses interesting pair of words, like "part-of-speech" tags.
- Look at the probability of the terms appearing in the doc / probability of them appearing in total.
- Tested on search engines.
- Multiple expressions of sentiment with regard to different objects / aspects.
- Starts positive, ends negative
- Given a document, identify every "quintiple" of sentiment using
- Object being discussed
- Aspect of the object being disucssed
- Sentiment expressed about it
- Who has the Sentiment
- when was it said
Problem Decomposition for Aspect Level :
- Aspect Extraction:
- Extract the target ("Phone")
- Grouping Aspects into categories:
- Find synonyms for aspects ("battery", "signal") etc.
- Aspect sentiment classification (the actual algorithm)
- Find meta data for entity holders and time extraction.
Algorithm Example:
- Mark sentiment words using a lexicon (+1 for +ve, -1 for -ve)
- Identify sentiment shifters (negations like "not, never, cannot") and swap sentiment for shifted words
- Identify "but" phrases, this helps mark sentiments for phrases that can't be identified (since they're usually the opposite of the other side of the but)
- Sum sentiment scores, weighted by word distance to aspect word.
Automated Trading:
- Montioring blogs, tweets and posts in investment forums, identify "bullish" and "bearish" statements, and use that to buy/sell.
- Identify experts in these (based on trend prediction) and weigh their opinions more.
The Crud Factor is described as the phenomenon that ultimately everything correlates to some extent with everything else.
Act of organisations or individuals influencing the sentiment tools via fake reviews or "sock puppet" commenting.
This is mostly done on review websites, where the sentiment can affect a business, and therfore has a monetary incentive.
- Yelp has fake reviews, they filter about 25% of the posts as they think it's fake.
- Governments have gotten involved in fining companies that post false reviews.
When one person uses a lot of fake accounts to post fake reviews. This can be both positive and negative.
Often these can be found out by just looking at how similar their tweets are, or meta data about location / time.
They aren't meant to shift the sentiment, just convince others to shift it instead.
Software built to help do sock puppertry easier.
- Consists of diverse plausible, and geographically consistent online personas.
- Static IP addresses are assigned to personas, so it seems real.
- Puppeteers are randomly selected, and the traffic is blended with outside users.
- Can be (and has been) used by goverments to influence the crowd sentiment in foreign nations.
- Spotting sentiment spam is difficult and requires effort.
- It's relatively easy to spot duplicate / very very similar reviews, and mark them as fakes.
- Someone used a ML approach to find fake reviews / spam sentiment via dataset of "removed reviews".
- They used word n-grams, subject focused data.
These researchers observed:
- Negative outlier reviews are heavily spammed, positive outliers less so.
- Singleton reviews are often fake.
- Top ranked reviews are more likely to post fake reviews.
Some spam detection on atypical behaviours is :
- Promiting or victimising only a few target products
- Targetting a group of products in a short period of time
- Often giving v high or v low scores
- Giving ratings wihch deviate from other reviewrs of a product.
Financial Regulation:
- Authority with rules and directives to control and manage services.
- Protects actors (by providing insurance and deposit protection)
- improves efficiency (by monopoly regulation and antitrust laws)
- reduces risk (differentiates between retail banking and other banking systems)
- Barcode Scanners:
- Helped retailers, for inventory keeping and labelling prices
- It's used in customer databases with loyalty cards.
- Credit Cards:
- Diners Club (intro)
- VISA mailed 60k cards to everyone
- BoA had 1mil cards in circulation in 1960
- Magnatic strip made it a lot easier to use.
- ATM (Automatic Teller Machine):
- Barclays started it, 24/7 access to money
- 80s networks appeared, and added more functionality
- Consolidated ATM network (LINK) started in the UK
These are examples of how the technical world grew in finance: First the data was needed to see how technology would work in this world, barcodes provided a great entry point into it. Second the network effects and scaling occured, where the card companies may have incurred a loss, but it allowed to build a network that helped in the long run. Third we came to ATMs that enabled the customers to get access to their resources more often using technology they wouldn't have before. This enabled the lock-in environment and helped the economy to work at scale. It even lead to regularisation and inter-operability amongst companies.
- Traditional incumbent finance ( big banks )
- Technical startups (monzo, etc)
- Incumbent big tech - Google, Microsoft apple etc.
Traditional banks:
- Added tech to streamline their business
- lead to a form of incremental innovation
Start-up tech:
- Used tech to introduce new services
- Finance designed from the ground-up (eg cryptocurrencies)
Incumbent tech companies:
- Use current tech dominance to move into FinTech
- Streamline tech and lock in users (apple pay)
- Leverage +ve network externalities
- Leverage Customer data.
There's no direct definition, for techfin, and it's basically just giant technologies company , "BigTech" moving into finance as part of their service offerings.
For example, Apple's Apple Pay, Amazon Pay, Android Pay etc.
Altering defintions:
There's two main definitions here, fintech and techfin.
FinTech: When companies take the financial processes, and apply it to technology o make something new. for example building the app to be good first then building the financial models etc. (monzo)
TechFin is when companies take the technology they have, and move it towards finance, for example creating an app on an old backend system. (banks making apps for themselves, ex: Barclays app)
The visuals make no sense (Why are they a cube?) But the descriptions are:
- Blockchain: Ordered and timestamped distributed public ledger, contains history of verified and valid transactions
- Social Networks: Web-based services, that allow individuals to construct public porfiles, share with their other users and traverse connections within the system
- NFC: Near Field Communication, short wireless point-to-point interconnection.
- Peer to Peer: Self organising system of equal autonomous entities. Aims for shared usage of resources in a networked environment. Avoids central dependencies
- Big Data Analytics: Characterised by enormous volumes of data, high processing and a lot of data sources to be taken into account.
- Further Enablers: Things like Ai, Mobile internet worldwide, UI and Security.
There are many factos that contribute towards the success of a company. There are a few outlined key factors:
LASIC:
- Low margin
- Asset light
- Scalable
- Innovative
- and Compliance Easy.
- Internet age, services are expected to be basically free
- High network effects, so build critical mass ASAP.
- Once critical mass achieved, monitise as quickly as possible.
- Ensure customer lock-in.
- Profit margins will remain low at the user level, so you need high margins
- Enables a low profit margin and is asset light
- it's able to scale without incurring large fixed costs
- This means that the marginal costs are low again
- Free ride on existing tech (using SMS paymetns)
- Technology must scale to reap benefits of network externalities.
- Ensure scale doesn't comprimise efficiency / increase costs
- Products and operations must be innovative
- Widespread use of mobile phones and internet service enables much innovation in the FinTech / TechFin space.
- Low Regulatory complains enables innovation
- businesses that recieve incentives to aid inclusion have an advantage.
- development led regulation is better than development lagged regulation.
There are a lot of case studies, check AleenaCodes' notes
Evolution of payments:
- From barter & direct exchange to single goldsmith banks with a central bank and a central ledger.
Flow of Money:
- There are intermediaries between the buyers and sellers, which means there's a transaction cost at each step, often meaning that the final seller gets lesser than the item is being sold for.
- The risk intermediaries take, is offset by imposing penalties on the seller when transactions go bad via a system of chargeback.
- Having a centralised bank can expose customers to risk.
The steps for a blockchain / bitcoin based transaction are:
- Agree on the transaction
- Create the transaction message
- Sign the transactions message
- broadcast the transaction
- verify the transaction
- success!!
Agreeing on the transaction:
- Accept the price of the object
- Add a price above that for any small fee that's needed to be made to incentivise the bitcoin miners.
Create the transaction message which has 3 components:
- A reference to the earlier transaction that paid her the 10 btc
- a list of addresses of recipients to the payments, i.e addresses of those involved on both sides, since the transaction pays the sender money back
- a list of amounts to pay the addresses.
Sign the message:
- Encrypt it with the private key.
Broadcasting the transaction
- Broadcast the encrypted message along with the public key.
- Initiate the broadcast by spreading message to the sender's peers in the peer2peer network
- Before broadcasting it, they verify the message so that inputs >= outputs.
- Sender's transaction still hasn't been included in the blocckhain yet.
Verifying the transaction
- Miners compete to be the first to "mine" a "block".
- Winner keeps the fee from every transaction.
- This is meant to incentivise the transactions.
- Each block has a header that contains
- timestamp and version number
- hash of the previous block
- merkle root that summarises the new transactions.
nonce(randomness) value that the miners header is hashed with sha256.
- After the header is a list of transactions recorded in the block, and it includes the miner's fees.
--- THIS IS TOO BORING ---
There's an economic market that is focused on ensuring that there's no computer crime. and there are companies making money from combating it.
Ransomwear is no longer lucrative, so attackers have shifted to cryptomining now, where they infect computers with bitcoin and other mining software, and mine that instead.
Email is still one of the biggest, if not the biggest ways that Malware is delivered on computers. The most common method of email malware is phishing attacks using fake websites and fake login forms.
Terminology:
- Virus: Program tha copies itself via attachement to a host code.
- Worm: Self replicating program that needs no host and no user intervention
- Trojan: Malware that masquarades as something else, invites the users to run it.
- RootKit: Malware concealed from normal user-processing.
- Backdoor: root into the system that bypasses security measures.
- Zombie: Machine controlled via backdoor/rootkits
- BotNet: network of zombies / bots.
- Randomware: malware that encrypts files and demans a ransom to unencrypt it.
- Phishing: fake emails that appear to come from trusted sources.
- Whitehat: good boi hackers.