ADDENDUM: As you know, the wisdom of crowds requires diversity, decentralization and independence (among a total of 4 criteria). Confirmation with this study.

Pfizer Spigit 24 Hours of Innovation Event Video from Stephen Jordan on Vimeo.

As has been pointed out here, here and here, the SmarterCities site is not a prediction market at all. It is an “opinion market” in which a variety of participants invest in multiple approaches based on their view of effectiveness of those approaches. They do not have mutually exclusive outcomes, nor can the different outcomes be verified as occurring when the markets close. Consider the site as an initiative in assessing people’s sentiment on different options related to the future infrastructure needs of cities.

Companies such as InTrade and BetFair operate prediction markets as central destinations where people participate in different markets. Spigit provides prediction markets for companies to use for their own purposes, whether internally for employees or externally for participants. As such, it ultimately is up to each client to decide how they will use Spigit’s technology.

Spigit offers two types of prediction markets:

• Traditional markets based on the stock market paradigm
• Token investment-based market

With that as background, presented below is a description of Spigit’s stock market form of prediction markets.

Basic Prediction Market Set-up

The intent of Spigit prediction markets is to make it easy to set one up and get going quickly. Below is a screenshot for setting up a prediction market on the Spigit platform.

Prediction market question, close date and description: The market starts with the question posed to participants. The standard advice to companies using the prediction market service is to create a “question that can be definitively answered within the set timeframe.”

Close date is when the market closes, and should relate to when the answer to a question can be definitively known.

The platform allows additional description to be provided for a question. This can include information such as the definitive source of record for defining which outcome occurs.

Create possible outcomes: The prediction market will have several possible outcomes. Keys in selecting outcomes include:

• Each outcome is mutually exclusive from the others. Only one of the outcomes will be realized upon closing of the market.
• The outcomes provided should cover 100% of all possibilities. This means that one of the outcomes will be realized upon the closing date.

Initial prices: In addition to listing the outcomes, the maker of a prediction market also sets the initial probabilities for each outcome. These probabilities take the form of prices in the prediction market. For many prediction markets, setting the initial probabilities as equal for all outcomes makes sense. All outcomes may be truly equally possible, or the market administrator is fine with letting probabilities naturally find their own level by participants’ actions.

Alternatively, different initial probabilities may be set for the different outcomes. This is useful in situations where financial rewards are part of the market. If a particular outcome is known to have a higher probability of occurrence, setting the initial probability higher can reduce the financial reward if that outcome does indeed occur.

Prediction Market Mechanics

Once the prediction market is live, participants will be able to buy and sell positions in different outcomes. Participants start with a given amount of currency. They are then free to trade different outcomes in the prediction market, based on their own views and sources of information.

Each participant will be presented with a list of outcomes, and their initial probabilities of occurrence:

From this page, each participant can make trades on the different outcomes by clicking the “buy/sell” button. The screenshot below shows a trading page for the outcome “9.4 – 9.5”, where participants can buy and sell positions in a fictional currency dooklas:

Buying and selling: Participants buy shares of an outcome when they believe it has a reasonable chance of occurring. If their view is that the probability of an outcome exceeds the current market probability, they may buy as well. Participants sell positions when their view of the outcome changes, or they believe an outcome has a lower probability than what the market is showing. Sales are only for positions already held by a participant.

Updating probabilities: The probabilities of each outcome are affected by the trades in the prediction market. The bigger the trade, the more it affects the probability of the outcome, up or down. In addition, because the probability of all outcomes must equal 100%, the probabilities of all other outcomes are affected by the trade. Assume a participant buys 100 shares of the “9.4 – 9.5” outcome. The share price is higher than the initial price of 17.000, going for 19.008 dooklas. After making the purchase, all probabilities adjust, as shown below:

Spigit allows prediction market administrators to plug in different pricing functions including the most popular logarithmic function in use by a number of other prediction market vendors. The platform market maker uses the algorithm to adjust the probabilities after a trade.

In this way, the market stays up-to-date with the latest probabilities as represented by the participants’ collective view.

Alerts: As the prices (probabilities) of each outcome change, participants can elect to receive alerts. These alerts can be set for when the probabilities drop below or above specified thresholds.

Leaderboard: Participants’ currency positions are maintained by the Spigit platform. The Leaderboard shows the top participants based on their net worth. Net worth changes as prediction markets are closed out, and winners and losers are determined.

Choose a winner: Once the outcome is known on the defined date, and the market closes, the prediction market administrator chooses the winning outcome. All holders of shares for that outcome are paid out at a price of 100.000. Gains are based on the different share prices purchased in the winning outcome. Share prices of other outcomes go to zero, becoming worthless.

Prediction Market Use Cases

Several companies are using or will be deploying Spigit prediction markets for various purposes. The table below shows these:

Spigit prediction market - enterprise use cases

You can contact Spigit to learn more.

[T]he “truth grounding” of prediction markets serves to avoid the self-reinforcement described above.

If a market, grounded on a real outcome, says A=90% and B=10% one day before expiration, and I believe that B is the real winner, then I will bet on B.

In this poll-like question, it makes no sense whatsoever to bet on B. The A will be the winner just due to the fact that the winner will be determined by which contract is leading at the expiration (in this case it will be A), not by which technology is going to be the true winner.

So, as a player I have no incentives to reveal my true beliefs. Instead I have all the incentives to follow the crowd in a bubble.

Remember: The crowd gives good decisions when each player thinks for itself and has the incentives to do so.

Panos Ipeirotis’s blog

-

ADDENDUM: Five hours after we published on the IBM Smarter Cities (tipped by a Twitter Direct Message from Hutch Carpenter), InTrade borrowed our info without thanking us for the tip, and published a tweet, labeling them as “prediction markets” (as Midas Oracle did originally, since I made the stupid mistake to put more weight in Spigt’s unscientific marketing prose than in my own critical reasoning). But Midas Oracle corrected the original post, thanks to David Pennock and Panos Ipeirotis, and Midas Oracle published another post stating my mistake. As for InTrade, they never published a correction, and they never published a new tweet establishing the truth, namely that the IBM Smarter Cities system is actually not powered by a market mechanism. InTrade has a history of not acknowledging the mistakes they make. They also have a history of siphoning Midas Oracle without crediting us, but that’s another story.

UPDATE: And InTrade post another tweet about IBM Smarter Cities “prediction markets”…

UPDATE: Jed Christiansen’s post

20 prediction lines

Only one socially generated prediction is over 50%.

UPDATE: The first version of this post used the term “prediction markets”, because that is how Spigit brands these mechanisms, but, on a closer look, each participant can’t choose the amount to bet/trade. –> FAIL. On top of that, their system is a beauty contest. –> FAIL2.

UPDATE: Jed Christiansen’s post

Prediction Market

Spigit implements traditional prediction markets, a tool for determining likelihood of future events and trends. In this model, market creator asks a question relating to a future event and multiple mutually exclusive answers that cover all possible outcomes. The creator also sets initial probabilities or prices. Spigit provides a simple trading model in which all transactions are executed between system’s market maker and the end user. Administrators can configure share buckets that users can choose to buy. The price is determined by a configurable price function; possibilities include quadratic, exponential, and linear. Irrespective of the actual function used, the price goes up as more shares are bought and goes down with shares sold. Users can set alerts when the stock price goes above or below user-specified values. When the market is closed, investors in the chosen answer receive 100 virtual dollars per share. All other investors lose their money.

Betting Market

This market inherits its basic structure from traditional prediction markets. Each market consists of a question and mutually exclusive answers relating to a future event, however, it does not involve stock market-like trading. Each user is given a fixed amount of tokens/chips for each open market. Users are free to distribute their tokens among possible answers as they see fit. They can change their bets at any time. Each market is assigned a budget by the market creator. When the market closes, the budgeted amount is uniformly distributed on per-token basis to all users that have placed their tokens on the accepted answer. This form of the market has several advantages over traditional prediction markets. It’s simpler to use since users do not have to deal with stock market formalism. Late entrants in the market benefit equally well when they bet their tokens on the right answer. The markets can be configured to allow users to play “blind” (i.e. not know what the current probabilities are when they place their bets), resulting in removing mob-effect bias.

Global Trends and Patterns

Activity Indexes–Indexes can be charted over a period of time to detect trends in participation levels.

Market price charts showing trends over different aggregation intervals and periods.

Trading volume – Overall and market-specific trading volume tracked hourly, daily, weekly, monthly.

Drill-Downs

Group Predictions – Market predictions made by user groups identified based on user attributes (location, position, age groups), etc., roles (executives, experts, etc.).
Reputation Weighted Predictions – If full spigit platform is utilized, predictions can be computed based on topic-specific reputations

Group Interests – Trading activity levels across active markets based on user roles and attributes

Bias Detection and Noise Removal

Two types of biases exist in any rating/betting system: proclivity of some individuals to over/under-rate an answer, and the tendency to favor outcomes that are beneficial to the individual or a group. The second one is applicable and discoverable for prediction markets. Spigit can compute prediction deviations for a group (based on user attributes/roles).

Spigit can reduce bias and the impact of day trading by computing probabilities after excluding trades made by members of the group that exceeds a configurable threshold.

Read & Write Web

Twitter

UPDATE: They have just brought the Best Buy quote down.